Package 'EZbakR'

An EZbakRFractions or EZbakRKinetics object, which is an EZbakRData object on which EstimateFractions() or EstimateKinetics() has been run.

Character vector of the set of features you want to stratify reads by and estimate proportions of each RNA population. The default of "all" will use all feature columns in the obj's cB.

Parameter to average across replicates of a given condition.

What type of table is the parameter found in? Default is "kinetics", but can also set to "fractions".

If type == "kinetics", then kstrat specifies the kinetic parameter inference strategy.

Character vector of the set of mutational populations that you want to infer the fractions of. Only relevant if type == "fractions".

"Design matrix" specifying which RNA populations exist in your samples. Only relevant if type == "fractions".

If TRUE, then features and populations have to exactly match those for a given fractions table for that table to be used. Means that you can't specify a subset of features or populations by default, since this is TRUE by default.

If multiple kinetics or fractions tables exist with the same metadata, then this is the numerical index by which they are distinguished.

An R formula object specifying how the parameter of interest depends on the sample characteristics specified in obj's metadf. The most common formula will be ~ treatment or ~ treatment:duration, where treatment and duration would be replaced with whatever you called the relevant sample characteristics in your metadf. ~ treatment means that an average value of parameter should be estimated for each set of samples with the same value for treatment in the metadf. ~ treatment:duration specifies that an average value of parameter should be estimated for each set of samples with the same combination of treatment and duration values in the metadf. An example of the latter case is a situation where you have two or more treatments (e.g., drug treated and untreated control) which were applied for different durations of time (e.g., 4 and 8 hours).

NOTE: EZbakR automatically removes any intercept terms from the model. That way, there is no ambiguity about what parameter is defined as the reference.

What metadf columns should data be grouped by when estimating standard deviations across replicates? If this is NULL, then EZbakR will check to see if the formula_mean specifies a formula that cleanly stratifies samples into disjoint groups. For example, the formula ~ treatment will assign each sample to a single factor (its value for the metadf's treatment column). In this case, standard deviations can be calculated for sets of replicates in each treatment group. If such a stratification does not exist, a single standard deviation will be estimated for each feature (i.e., homoskedasticity will be assumed as in standard linear modeling).

If TRUE, an additional table will be saved with the prefix fullfit_, which includes all of the parameters estimated throughout the course of linear modeling and regularization. This can be nice for visualizing the regularized mean-variance trend.

Determines how strongly variance estimates are shrunk towards trend. Higher numbers lead to more regularization. Eventually, this will be replaced with estimation of how much variance there seems to be in the population of variances.

If TRUE, linear model will throw an error if parameters cannot be uniquely identified. This is most often caused by parameters that cannot be estimated from the data, e.g., due to limited replicate numbers or correlated sample characteristics (i.e., all treatment As also correspond to batch As, and all treatment Bs correspond to batch Bs).

Minimum number of reads in all samples for a feature to be kept.

If a label time variable is included in the formula_mean, convert its values to factors so as to avoid performing continuous regression on label times. Defaults to TRUE as including label time in the regression is often meant to stratify samples by their label time if, for example, you are averaging logit(fractions).

If TRUE, and if type == "fractions", then standard error will be regressed against logit fraction rather than magnitude of logit fraction. Makes sense to set this to FALSE if analyzing certain site-specific mutational probing methods when high mutation content things are likely low variance SNPs.

Certain formula lend them selves to efficient approximations of the full call to lm(). Namely, formulas that stratify samples into disjoint groups where a single parameter of the model is effectively estimated from each group can be tackled via simple averaging of data from each from group. If you would like to force EZbakR to fit the fully rigorous linear model though, set force_lm to TRUE.

Old parameter that is now passed the value force_lm and will be deprecated in later releases

If TRUE, conservative variance regularation will be performed. In this case, variances below the trend will be regularized up to the trend, and variances above the trend will be left unregularized. This avoids underestimation of variances.

Limit on the number of characters of the name given to the output table. Will attempt to concatenate the parameter name with the names of all of the features. If this is too long, only the parameter name will be used.

Table of effective lengths for each feature combination in your data. For example, if your analysis includes features named GF and XF, this should be a data frame with columns GF, XF, and length.

Data frame with columns ⁠<feature names>⁠ and nsamps, where ⁠<feature names>⁠ are all of the feature columns in the input to AverageAndRegularize(), and nsamps is the number of samples that samples from that feature combination needs to have over the read count threshold.

Data frame with columns "sample" and a second column of whatever name you please. The second column should denote scale factors by which read counts in that sample should be multiplied by in order to normalize these read counts.

If TRUE, identical, existing output will be overwritten.

An EZbakRFit object, which is an EZbakRFractions object on which AverageAndRegularize() has been run.

Name of factor from metadf whose parameter estimates at different factor values you would like to compare. If you specify this, you need to also specify reference and experimental. If type == "dynamics", this can have multiple values, being the names of all of the factors you would like to stratify a group by.

Name of reference design_factor factor level value. Difference will be calculated as experimental - reference. If type == "dynamics", then this should specify the levels of all of the design_factor(s) reference group. For example, if you have multiple design_factor's, then reference must be a named character vector with one element per design_factor, with elements named the corresponding design_factor. For example, if design_factor is c("genotype", "treatment"), and you would like to compare genotype = "WT" and treatment = "untreated" (reference) to genotype = "KO" and treatment = "treated", then reference would need to be a vector with one element named "genotype", equal to "WT" and one element named "treatment" equal to "untreated" (this example could be created with, c(genotype = "WT", treatment = "untreated")).

Name of design_factor factor level value to compare to reference. Difference will be calculated as experimental - reference. If type == "dynamics", then this should specify the levels of all of the design_factor(s) reference group. For example, if you have multiple design_factor's, then experimental must be a named character vector with one element per design_factor, with elements named the corresponding design_factor. For example, if design_factor is c("genotype", "treatment"), and you would like to compare genotype = "WT" and treatment = "untreated" (reference) to genotype = "KO" and treatment = "treated", then experimental would need to be a vector with one element named "genotype", equal to "KO" and one element named "treatment" equal to "treated" (this example could be created with, c(genotype = "KO", treatment = "treated")).

If you want to assess the significance of a single parameter, rather than the comparison of two parameters, specify that one parameter's name here.

Parameter to average across replicates of a given condition.

Type of table to use. Can either be "averages" or "dynamics".

NOT YET IMPLEMENTED. Will allow you to specify more complicated functions of parameters when hypotheses you need to test are combinations of parameters rather than individual parameters or simple differences in two parameters.

Same as design_factor, will be deprecated in favor of the former in later release.

Character vector of the set of features you want to stratify reads by and estimate proportions of each RNA population. The default of "all" will use all feature columns in the obj's cB.

If TRUE, then features and populations have to exactly match those for a given fractions table for that table to be used. Means that you can't specify a subset of features or populations by default, since this is TRUE by default.

If multiple averages tables exist with the same metadata, then this is the numerical index by which they are distinguished.

An R formula object specifying how the parameter of interest depends on the sample characteristics for the averages object you want to use.

Metadf columns should data was grouped by when estimating standard deviations across replicates for the averages object you want to use.

Character vector of parameter names in the averages object you would like to use.

If TRUE, then median difference will be set equal to 0. This can account for global biases in parameter estimates due to things like differences in effective label times. Does risk eliminating real signal though, so user discretion is advised.

Same as reference, but exclusively parsed in case of type == "dynamics, included for backwards compatibility.

Same as experimental, but exclusively parsed in case of type == "dynamics, included for backwards compatibility.

If TRUE, then identical output will be overwritten if it exists.

An EZbakRFractions object, which is an EZbakRData object on which you have run EstimateFractions().

Which dropout correction strategy to use. Options are:

• grandR: Described here. Cite that work and grandR if using this strategy. Quasi-non-parametric strategy that finds an estimate of the dropout rate that eliminates any linear correlation between the newness of a transcript and the difference in +s4U and -s4U normalized read counts.

• bakR: Described here. Uses a simple generative model of dropout to derive a likelihood function, and the dropout rate is estimated via the method of maximum likelihood.

The "bakR" strategy has the advantage of being model-derived, making it possible to assess model fit and thus whether the simple assumptions of both the "bakR" and "grandR" dropout models are met. The "grandR" strategy has the advantage of being more robust. Thus, the "grandR" strategy is currently used by default.

Which sample-detail columns in the metadf should be used to group -s4U samples by for calculating the average -s4U RPM? The default value of NULL will cause all sample-detail columns to be used.

Character vector of the set of features you want to stratify reads by and estimate proportions of each RNA population. The default of NULL will expect there to be only one fractions table in the EZbakRFractions object.

Mutational populations that were analyzed to generate the fractions table to use. For example, this would be "TC" for a standard s4U-based nucleotide recoding experiment.

"Design matrix" specifying which RNA populations exist in your samples. By default, this will be created automatically and will assume that all combinations of the mutrate_populations you have requested to analyze are present in your data. If this is not the case for your data, then you will have to create one manually. See docs for EstimateFractions (run ?EstimateFractions()) for more details.

If multiple fractions tables exist with the same metadata, then this is the numerical index by which they are distinguished.

If TRUE, then features must exactly match the features metadata for a given fractions table for it to be used. Means that you cannot specify a subset of features by default. Set this to FALSE if you would like to specify a feature subset.

Minimum number of reads for a feature to be used to fit the dropout model.

Maximum ratio of -s4U:+s4U RPMs for a feature to be used to fit the dropout model (i.e., simple outlier filtering cutoff).

Parameters passed to internal calculate_dropout() function; namely dropout_cutoff_min, which sets the minimum dropout value used for fitting the dropout model.

Character vector of the set of mutational populations present in your data. For example, s4U fed data with standard nucleotide recoding chemistry (e.g., TimeLapse, SLAM, TUC, AMUC, etc.) would have a mutrate_populations of c("TC"). Dual labeling experiments with s4U and s6G feeds would have a mutrate_populations of c("TC", "GA").

An EZbakRData object

Either "gene" (if deconvolving gene-level fraction news, i.e., for resolving fusion gene and component gene fraction news) or "isoform" (if deconvolving transcript isoform fraction news).

Character vector of the set of features you want to stratify reads by and estimate proportions of each RNA population. The default of "all" will use all feature columns in the obj's cB.

Mutational populations that were analyzed to generate the fractions table to use. For example, this would be "TC" for a standard s4U-based nucleotide recoding experiment.

"Design matrix" specifying which RNA populations exist in your samples. By default, this will be created automatically and will assume that all combinations of the mutrate_populations you have requested to analyze are present in your data. If this is not the case for your data, then you will have to create one manually. See docs for EstimateFractions (run ?EstimateFractions()) for more details.

If multiple fractions tables exist with the same metadata, then this is the numerical index by which they are distinguished.

If TRUE, then features and populations have to exactly match those for a given fractions table for that table to be used. Means that you can't specify a subset of features or populations by default, since this is TRUE by default.

Name of fraction estimate table to use. Should be stored in the obj$fractions list under this name. Can also rely on specifying features and/or populations and having EZget() find it.

Name of transcript isoform quantification table to use. Should be stored in the obj$readcounts list under this name. Use ImportIsoformQuant() to create this table. If quant_name is NULL, it will search for tables containing the string "isoform_quant" in their name, as that is the naming convention used by ImportIsoformQuant(). If more than one such table exists, an error will be thrown and you will have to specify the exact name in quant_name.

Table with columns transcript_id and all feature related columns that appear in the relevant fractions table. This is only relevant as a hack to to deal with the case where STAR includes in its transcriptome alignment transcripts on the opposite strand from where the RNA actually originated. This table will be used to filter out such transcript-feature combinations that should not exist.

If TRUE and a fractions estimate output already exists that would possess the same metadata (features analyzed, populations analyzed, and fraction_design), then it will get overwritten with the new output. Else, it will be saved as a separate output with the same name + "_#" where "#" is a numerical ID to distinguish the similar outputs.

Minimum TPM for a transcript to be kept in analysis.

Minimum expected_count for a transcript to be kept in analysis.

EZbakRData or EZbakRArrowData object

Character vector of the set of features you want to stratify reads by and estimate proportions of each RNA population. The default of "all" will use all feature columns in the obj's cB file.

Character vector of the set of mutational populations that you want to infer the rates of mutations for. By default, all mutation rates are estimated for all populations present in cB.

"Design matrix" specifying which RNA populations exist in your samples. By default, this will be created automatically and will assume that all combinations of the mutrate_populations you have requested to analyze are present in your data. If this is not the case for your data, then you will have to create one manually.

If you call the function create_fraction_design(...), providing a vector of mutational population names as input, it will create a fraction_design table for you, with the assumption that every single possible combination of mutational populations is present in your data. You can then edit the present column as necessary to get an appropriate fraction_design for your use case. See below for details on the required contents of fraction_design and its interpretation.

fraction_design must have one column per element of mutrate_populations, with these columns sharing the name of the mutrate_populations. It must also have one additional column named present. All elements of fraction_design should be booleans (TRUE or FALSE). It should include all possible combinations of TRUE and FALSE for the mutrate_populations columns. A TRUE in one of these columns represents a population of RNA that is expected to have above background mutation rates of that type. present will denote whether or not that population of RNA is expected to exist in your data.

For example, assume you are doing a typical TimeLapse-seq/SLAM-seq/TUC-seq/etc. experiment where you have fed cells with s^4U and recoded any incorporated s^4U to a nucleotide that reverse transcriptase will read as a cytosine. That means that mutrate_populations will be "TC", since you want to estimate the fraction of RNA that was s^4U labeled, i.e., the fraction with high T-to-C mutation content. fraction_design will thus have two columns: TC and present. It will also have two rows. One of these rows must have a value of TRUE for TC, and the other must have a value of FALSE. The row with a value of TRUE for TC represents the population of reads with high T-to-C mutation content, i.e., the reads from RNA that were synthesized while s^4U was present. The row with a value of FALSE for TC reprsents the population of reads with low T-to-C mutation content, i.e., the reads from RNA that existed prior to s^4U labeling. Both of these populations exist in your data, so the value of the present column should be TRUE for both of these. See the lazily loaded standard_fraction_design object for an example of what this tibble could look like. ("lazily loaded standard_fraction_design object" means that if you run print(standard_fraction_design) after loading EZbakR with library(EZbakR), then you can see its contents. More specifically, lazily loaded means that this table is not loaded into memory until you ask for it, via something like a print() call.)

As another example, consider TILAC, a NR-seq extension developed by the Simon lab. TILAC was originally described in Courvan et al., 2022. In this method, two populations of RNA, one from s^4U fed cells and one from s^6G fed cells, are pooled and prepped for sequencing together. This allows for internally controlled comparisons of RNA abundance without spike-ins. s^4U is recoded to a cytosine analog by TimeLapse chemistry (or similar chemistry) and s^6G is recoded to an adenine analog. Thus, fraction_design includes columns called TC and GA. A unique aspect of the TILAC fraction_design table is that one of the possible populations, TC and GA both TRUE, is denoted as not present (present = FALSE). This is because there is no RNA that was exposed to both s^4U and s^6G, thus a population of reads with both high T-to-C and G-to-A mutational content should not exist. To see an example of what a TILAC fraction_design table could look like, see the lazily loaded tilac_fraction_design object.

If TRUE, use U-content adjusted Poisson mixture modeling strategy. Often provides significant speed gain without sacrificing accuracy.

String denoting which new read mutation rate estimation strategy to use. Options include:

• standard: Estimate a single new read and old read mutation rate for each sample. This is done via a binomial mixture model aggregating over

• hierarchical: Estimate feature-specific new read mutation rate, regularizing the feature-specific estimate with a sample-wide prior. Currently only compatible with single mutation type mixture modeling.

Which feature columns should be used to filter out feature-less reads. The default value of "all" checks all feature columns for whether or not a read failed to get assigned to said feature.

Only two possible values for this make sense: `&` and `|`. If set to `&`, then all features in filter_cols must have a "null" value (i.e., a value included in remove_features) for the row to get filtered out. If set to `|`, then only a single feature in filter_cols needs to have one of these "null" values to get filtered out.

All of the feature names that could indicate failed assignment of a read to a given feature. the fastq2EZbakR pipeline uses a value of '__no_feature'.

If a set of reads maps ambiguously to multiple features, should data for such reads be copied for each feature in the ambiguous set? If this is TRUE, then multi_feature_cols also must be set. Examples where this should be set to TRUE includes when analyzing exonic bins (concept defined in original DEXSeq paper), exon-exon junctions, etc.

Character vector of columns that have the potential to include assignment to multiple features. Only these columns will have their features split if split_multi_features is TRUE.

String representing how ambiguous feature assignments are distinguished in the feature names. For example, the default value of "+" denotes that if a read maps to multiple features (call them featureA and featureB, for example), then the feature column will have a value of "featureA+featureB" for that read.

Mean for logit(pnew) prior.

Standard deviation for logit(pnew) prior.

Mean for logit(pold) prior.

Standard deviation for logit(pold) prior.

If strategy == hierarchical, only features with this many reads are used to infer the distribution of feature-specific labeled read mutation rates.

If strategy == hierarchical, this is the initial logit(pnew) prior standard deviation to regularize feature-specific labeled read mutation rate estimates.

The minimum logit(pnew) prior standard deviation when strategy is set to "hierarchcial". EZbakR will try to estimate this empirically as the standard deviation of initial feature-specific logit(pnew) estimates using high coverage features, minus the average uncertainty in the logit(pnew) estimates. As this difference can sometimes be negative, a value of pnew_prior_sd_min will be imputed in that case.

Similar to pnew_prior_sd_min, but now representing the maximum allowed logit(pnew) prior sd.

Background mutation rate estimates if you have them. Can either be a single number applied to all samples or a named vector of values, where the names should be sample names.

Fix background mutation rate estimate to mutation rates seen in -label samples. By default, a single background rate is used for all samples, inferred from the average mutation rate across all -label samples. The grouping_factors argument can be specified to use certain -label samples to infer background mutation rates for certain sets of +label samples.

If pold_from_nolabel is TRUE, then grouping_factors will specify the sample-detail columns in the metadf that should be used to group -label samples by. Average mutation rates in each group of -label samples will be used as the background mutation rate estimate in +label samples with the same values for the relevant metadf columns.

Maximum number of characters for naming out fractions output. EZbakR will try to name this as a "_" separated character vector of all of the features analyzed. If this name is greater than character_limit, then it will default to "fraction#", where "#" represents a simple numerical ID for the table.

If TRUE and a fractions estimate output already exists that would possess the same metadata (features analyzed, populations analyzed, and fraction_design), then it will get overwritten with the new output. Else, it will be saved as a separate output with the same name + "_#" where "#" is a numerical ID to distinguish the similar outputs.

An EZbakRData object

Character vector of the set of features you want to stratify reads by and estimate proportions of each RNA population. The default of "all" will use all feature columns in the obj's cB.

Mutational populations that were analyzed to generate the fractions table to use. For example, this would be "TC" for a standard s4U-based nucleotide recoding experiment.

"Design matrix" specifying which RNA populations exist in your samples. By default, this will be created automatically and will assume that all combinations of the mutrate_populations you have requested to analyze are present in your data. If this is not the case for your data, then you will have to create one manually. See docs for EstimateFractions (run ?EstimateFractions()) for more details.

If multiple fractions tables exist with the same metadata, then this is the numerical index by which they are distinguished.

If TRUE, then features and populations have to exactly match those for a given fractions table for that table to be used. Means that you can't specify a subset of features or populations by default, since this is TRUE by default.

Name of fraction estimate table to use. Should be stored in the obj$fractions list under this name. Can also rely on specifying features and/or populations and having EZget() find it.

Name of transcript isoform quantification table to use. Should be stored in the obj$readcounts list under this name. Use ImportIsoformQuant() to create this table. If quant_name is NULL, it will search for tables containing the string "isoform_quant" in their name, as that is the naming convention used by ImportIsoformQuant(). If more than one such table exists, an error will be thrown and you will have to specify the exact name in quant_name.

Table with columns transcript_id and all feature related columns that appear in the relevant fractions table. This is only relevant as a hack to to deal with the case where STAR includes in its transcriptome alignment transcripts on the opposite strand from where the RNA actually originated. This table will be used to filter out such transcript-feature combinations that should not exist.

If TRUE and a fractions estimate output already exists that would possess the same metadata (features analyzed, populations analyzed, and fraction_design), then it will get overwritten with the new output. Else, it will be saved as a separate output with the same name + "_#" where "#" is a numerical ID to distinguish the similar outputs.

Minimum TPM for a transcript to be kept in analysis.

Minimum expected_count for a transcript to be kept in analysis.

An EZbakRFractions object, which is an EZbakRData object on which EstimateFractions() has been run.

Kinetic parameter estimation strategy. Options include:

• standard: Estimate a single new read and old read mutation rate for each sample. This is done via a binomial mixture model aggregating over

• NSS: Use strategy similar to that presented in Narain et al., 2021 that assumes you have data which provides a reference for how much RNA was present at the start of each labeling. In this case, grouping_factor and reference_factor must also be set.

• shortfeed: Estimate kinetic parameters assuming no degradation of labeled RNA, most appropriate if the metabolic label feed time is much shorter than the average half-life of an RNA in your system.

• tilac: Estimate TILAC-ratio as described in Courvan et al., 2022.

• pulse-chase: Estimate kdeg for a pulse-chase experiment. By default kdeg will be estimated for each time point at which label was present. This includes any pulse-only samples, as well as all samples including a chase after the pulse. If you don't want to include the estimates from the pulse-only samples in the final output, set exclude_pulse_estimates to TRUE. Pulse-chases are suboptimal for a number of experimental reasons, so we urge users to avoid performing this kind of experiment whenever possible (favoring instead a pulse-label design). One of the challenges of analyzing pulse-chase data is that the fraction labeled after the pulse must be compared to that after each chase. Due to a number of technical reasons, it is possible for the estimated labeling after the chase to be *higher than that after the pulse. It is impossible to estimate kinetic parameters in this case. Thus, when this arises, a conservative kdeg estimate is imputed, equal to -log(1 - 1/(n+2))/tchase, which is the kdeg estimate you would get if you had no detected reads from labeled RNA and an uninformative prior on the fraction new (i.e., the estimated fraction new is (number of reads from labeled RNA + 1) / (number of reads + 2)).

Character vector of the set of features you want to stratify reads by and estimate proportions of each RNA population. The default of "all" will use all feature columns in the obj's cB.

Mutational populations that were analyzed to generate the fractions table to use. For example, this would be "TC" for a standard s4U-based nucleotide recoding experiment.

"Design matrix" specifying which RNA populations exist in your samples. By default, this will be created automatically and will assume that all combinations of the mutrate_populations you have requested to analyze are present in your data. If this is not the case for your data, then you will have to create one manually. See docs for EstimateFractions (run ?EstimateFractions()) for more details.

If multiple fractions tables exist with the same metadata, then this is the numerical index by which they are distinguished.

If TRUE, then features and populations have to exactly match those for a given fractions table for that table to be used. Means that you can't specify a subset of features or populations by default, since this is TRUE by default.

Which sample-detail columns in the metadf should be used to group samples by for calculating the average RPM (strategy = "NSS") or pulse fraction labeled (strategy = "pulse-chase") at a particular time point? Only relevant if strategy = "NSS" or strategy = "pulse-chase".

Which sample-detail column in the metafd should be used to determine which group of samples provide information about the starting levels of RNA for each sample? This should have values that match those in grouping_factor. #' Only relevant if strategy = "NSS".

Maximum number of characters for naming out fractions output. EZbakR will try to name this as a "_" separated character vector of all of the features analyzed. If this name is greater than character_limit, then it will default to "fraction#", where "#" represents a simple numerical ID for the table.

Table of effective lengths for each feature combination in your data. For example, if your analysis includes features named GF and XF, this should be a data frame with columns GF, XF, and length.

If strategy = "pulse-chase", would you like to exclude the pulse-only kdeg and ksyn estimates from the final output? This is a good idea if you used a very long pulse with the goal of labeling close to 100% of all RNA.

Data frame with two columns, "sample" and "scale_factor". sample should be the sample name, and scale_factor is the factor which read counts in that sample should be divided to get the normalized read counts. This should match the convention from DESeq2 and thus you can provide, for example, scale factors derived from DESeq2 for this. By default, EZbakR uses DESeq2's median of ratios normalization method, but specifying scale_factors is useful when you have spike-ins.

If TRUE and a fractions estimate output already exists that would possess the same metadata (features analyzed, populations analyzed, and fraction_design), then it will get overwritten with the new output. Else, it will be saved as a separate output with the same name + "_#" where "#" is a numerical ID to distinguish the similar outputs.

An EZbakRData or EZbakRArrowData object

Character vector of the set of mutational populations that you want to infer the fractions of. For example, say your cB file contains columns tracking T-to-C and G-to-A

logit-Normal mean for logit(pnew) prior.

logit-Normal sd for logit(pnew) prior.

logit-Normal mean for logit(pold) prior.

logit-Normal sd for logit(pold) prior.

Background mutation rate estimates if you have them. Can either be a single number applied to all samples or a named vector of values, where the names should be sample names.

Fix background mutation rate estimate to mutation rates seen in -label samples. By default, a single background rate is used for all samples, inferred from the average mutation rate across all -label samples. The grouping_factors argument can be specified to use certain -label samples to infer background mutation rates for certain sets of +label samples.

If pold_from_nolabel is TRUE, then grouping_factors will specify the sample-detail columns in the metadf that should be used to group -label samples by. Average mutation rates in each group of -label samples will be used as the background mutation rate estimate in +label samples with the same values for the relevant metadf columns.

ArrowDataset with the following fields:

• sample: Name given to particular sample from which data was collected.

• mutational counts: Integers corresponding to the number of a particular mutation seen in a sequencing read. The following column names are allowed:

  • TC: Number of Thymine-to-Cytosine mutations

  • TA: Number of Thymine-to-Adenine mutations

  • TG: Number of Thymine-to-Guanine mutations

  • CT: Number of Cytosine-to-Thymine mutations

  • CA: Number of Cytosine-to-Adenine mutations

  • CG: Number of Cytosine-to-Guanine mutations

  • CU: Number of Cytosine-to-Uridine mutations

  • AT: Number of Adenine-to-Thymine mutations

  • AC: Number of Adenine-to-Cytosine mutations

  • AG: Number of Adenine-to-Guanine mutations

  • AU: Number of Adenine-to-Uridine mutations

  • GT: Number of Guanine-to-Thymine mutations

  • GC: Number of Guanine-to-Cytosine mutations

  • GA: Number of Guanine-to-Adenine mutations

  • GU: Number of Guanine-to-Uridine mutations

  • TN: Number of Thymine-to-Adenine/Cytosine/Guanine mutations

  • CN: Number of Cytosine-to-Adenine/Thymine/Guanine/Uridine mutations

  • AN: Number of Adenine-to-Thymine/Cytosine/Guanine/Uridine mutations

  • GN: Number of Guanine-to-Adenine/Cytosine/Thymine/Uridine mutations

  • UN: Number of Uridine-to-Adenine/Cytosine/Guanine mutations

  • NT: Number of Adenine/Cytosine/Guanine-to-Thymine mutations

  • NC: Number of Adenine/Thymine/Guanine/Uridine-to-Cytosine mutations

  • NtoA: Number of Thymine/Cytosine/Guanine/Uridine-to-Adenine mutations. (Naming convention changed because NA taken)

  • NU: Number of Cytosine/Guanine/Adenine-to-Uridine mutations.

  • NN: Number of any kind of mutation

• base nucleotide count: Integers corresponding to the number of instances of a particular type of nucleotide whose mutations are tracked in a corresponding mutation count column. The following column names are allowed:

  • nT: Number of Thymines

  • nG: Number of Guanines

  • nA: Number of Adenines

  • nC: Number of Cytosines

  • nU: Number of Uridines

  • nN: number of any kind of nucleotide

• features: Any columns that cannot be interpreted as a mutation count or base nucleotide count (and that aren't named sample or n) will be interpreted as an ID for a genomic "feature" from which a read originated. Common examples of features and typical column names for said features include:

  • Genes; common column names: gene, gene_id, gene_name, GF

  • Genes-exonic; common column names: gene_exon, gene_id_exon, gene_name_exon, XF

  • Transcripts; common column names: transcripts, TF

  • Exonic bins; common column names: exonic_bins, EF, EB

  • Exons; common column names: exons, exon_ids

  In some cases, a read will often map to multiple features (e.g., exons). Many functions in bakR expect each of the feature IDs in these cases to be separated by +. For example, if a read overlaps with two exons, with IDs exon_1 and exon_2, then the corresponding entry in a column of exonic assignments would be "exon_1+exon_2". The default expectation can be overwritten though and is thus not strictly enforced.

• n: Number of reads with identical values for all other columns.

Data frame detailing various aspects of each of the samples included in the cBds. This includes:

• sample: The sample ID, which should correspond to a sample ID in the provided cBds.

• tl: Metabolic label time. There are several edge cases to be aware of:

  • If more than one metabolic label was used in the set of samples described by the metadf (e.g., s4U and s6G were used), then the tl column should be replaced by ⁠tl_<muttype>⁠, where ⁠<muttype>⁠ represents the corresponding mutation type count column in the cBds that the label whose incubation time will be listed in this column. For example, if feeding with s4U in some samples and s6G in others, then performing standard nucleotide recoding chemistry, you will include tl_TC and tl_GA columns corresponding to the s4U and s6G label times, respectively.

  • If a pulse-chase experimental design was used (!!this is strongly discouraged unless you have a legitimate reason to prefer this design to a pulse-label design!!), then you should have columns named tpulse and tchase, corresponding to the pulse and chase times respectively. The same ⁠<muttype>⁠ convention should be used in the case of multi-label pulse-chase designs.

• sample characteristics: The remaining columns can be named whatever you like and should include distinguishing features of groups of samples. Common columns might include:

  • treatment: The experimental treatment applied to a set of samples. This could represent things like genetic knockouts or knockdowns, drug treatments, etc.

  • batch: An ID for sets of samples that were collected and/or processed together. Useful for regressing out technical batch effects

Data frame with the following columns:

• sample: Name given to particular sample from which data was collected.

• mutational counts: Integers corresponding to the number of a particular mutation seen in a sequencing read. The following column names are allowed:

  • TC: Number of Thymine-to-Cytosine mutations

  • TA: Number of Thymine-to-Adenine mutations

  • TG: Number of Thymine-to-Guanine mutations

  • CT: Number of Cytosine-to-Thymine mutations

  • CA: Number of Cytosine-to-Adenine mutations

  • CG: Number of Cytosine-to-Guanine mutations

  • CU: Number of Cytosine-to-Uridine mutations

  • AT: Number of Adenine-to-Thymine mutations

  • AC: Number of Adenine-to-Cytosine mutations

  • AG: Number of Adenine-to-Guanine mutations

  • AU: Number of Adenine-to-Uridine mutations

  • GT: Number of Guanine-to-Thymine mutations

  • GC: Number of Guanine-to-Cytosine mutations

  • GA: Number of Guanine-to-Adenine mutations

  • GU: Number of Guanine-to-Uridine mutations

  • TN: Number of Thymine-to-Adenine/Cytosine/Guanine mutations

  • CN: Number of Cytosine-to-Adenine/Thymine/Guanine/Uridine mutations

  • AN: Number of Adenine-to-Thymine/Cytosine/Guanine/Uridine mutations

  • GN: Number of Guanine-to-Adenine/Cytosine/Thymine/Uridine mutations

  • UN: Number of Uridine-to-Adenine/Cytosine/Guanine mutations

  • NT: Number of Adenine/Cytosine/Guanine-to-Thymine mutations

  • NC: Number of Adenine/Thymine/Guanine/Uridine-to-Cytosine mutations

  • NtoA: Number of Thymine/Cytosine/Guanine/Uridine-to-Adenine mutations. (Naming convention changed because NA taken)

  • NU: Number of Cytosine/Guanine/Adenine-to-Uridine mutations.

  • NN: Number of any kind of mutation

• base nucleotide count: Integers corresponding to the number of instances of a particular type of nucleotide whose mutations are tracked in a corresponding mutation count column. The following column names are allowed:

  • nT: Number of Thymines

  • nG: Number of Guanines

  • nA: Number of Adenines

  • nC: Number of Cytosines

  • nU: Number of Uridines

  • nN: number of any kind of nucleotide

• features: Any columns that cannot be interpreted as a mutation count or base nucleotide count (and that aren't named sample or n) will be interpreted as an ID for a genomic "feature" from which a read originated. Common examples of features and typical column names for said features include:

  • Genes; common column names: gene, gene_id, gene_name, GF

  • Genes-exonic; common column names: gene_exon, gene_id_exon, gene_name_exon, XF

  • Transcripts; common column names: transcripts, TF

  • Exonic bins; common column names: exonic_bins, EF, EB

  • Exons; common column names: exons, exon_ids

  In some cases, a read will often map to multiple features (e.g., exons). Many functions in bakR expect each of the feature IDs in these cases to be separated by +. For example, if a read overlaps with two exons, with IDs exon_1 and exon_2, then the corresponding entry in a column of exonic assignments would be "exon_1+exon_2". The default expectation can be overwritten though and is thus not strictly enforced.

• n: Number of reads with identical values for all other columns.

Data frame detailing various aspects of each of the samples included in the cB. This includes:

• sample: The sample ID, which should correspond to a sample ID in the provided cB.

• tl: Metabolic label time. There are several edge cases to be aware of:

  • If more than one metabolic label was used in the set of samples described by the metadf (e.g., s4U and s6G were used), then the tl column should be replaced by ⁠tl_<muttype>⁠, where ⁠<muttype>⁠ represents the corresponding mutation type count column in the cB that the label whose incubation time will be listed in this column. For example, if feeding with s4U in some samples and s6G in others, then performing standard nucleotide recoding chemistry, you will include tl_TC and tl_GA columns corresponding to the s4U and s6G label times, respectively.

  • If a pulse-chase experimental design was used (!!this is strongly discouraged unless you have a legitimate reason to prefer this design to a pulse-label design!!), then you should have columns named tpulse and tchase, corresponding to the pulse and chase times respectively. The same ⁠_<muttype>⁠ convention should be used in the case of multi-label pulse-chase designs.

• sample characteristics: The remaining columns can be named whatever you like and should include distinguishing features of groups of samples. Common columns might include:

  • treatment: The experimental treatment applied to a set of samples. This could represent things like genetic knockouts or knockdowns, drug treatments, etc.

  • batch: An ID for sets of samples that were collected and/or processed together. Useful for regressing out technical batch effects

Data frame with the following columns:

• sample: Name given to particular sample from which data was collected.

• estimates of population fractions: These columns refer to the estimate for the fraction of reads coming from a particular mutational population. For example, in a standard NR-seq experiment, you should have one column named fraction_highTC. This refers to the fraction of RNA inferred to have a high T-to-C mutation rate (e.g., the newly synthesized RNA in a pulse-labeling NR-seq experiment). If you estimated the fractions of more than 2 mutation types (e.g., T-to-C and G-to-A), then you need to explicitly list all fractions of interest estimated. For example, in a TILAC experiment, this would be fraction_highTC_lowGA, fraction_lowTC_highGA, and fraction_lowTC_lowGA.

• n: Number of reads assigned to a given feature in a given sample.

• features: Any columns that cannot be interpreted as an estimate of population fractions (and that aren't named sample or n) will be interpreted as an ID for a genomic "feature" from which a read originated. Common examples of features and typical column names for said features include:

  • Genes; common column names: gene, gene_id, gene_name, GF

  • Genes-exonic; common column names: gene_exon, gene_id_exon, gene_name_exon, XF

  • Transcripts; common column names: transcripts, TF

  • Exonic bins; common column names: exonic_bins, EF, EB

  • Exons; common column names: exons, exon_ids

  In some cases, a read will often map to multiple features (e.g., exons). Many functions in bakR expect each of the feature IDs in these cases to be separated by +. For example, if a read overlaps with two exons, with IDs exon_1 and exon_2, then the corresponding entry in a column of exonic assignments would be "exon_1+exon_2". The default expectation can be overwritten though and is thus not strictly enforced.

• n: Number of reads with identical values for all other columns.

Data frame detailing various aspects of each of the samples included in the fractions data frame. This includes:

• sample: The sample ID, which should correspond to a sample ID in the provided fractions data frame.

• tl: Metabolic label time. There are several edge cases to be aware of:

  • If more than one metabolic label was used in the set of samples described by the metadf (e.g., s4U and s6G were used), then the tl column should be replaced by ⁠tl_<muttype>⁠, where ⁠<muttype>⁠ represents the corresponding mutation type referenced in the fractions that the label whose incubation time will be listed in this column. For example, if feeding with s4U in some samples and s6G in others, then performing standard nucleotide recoding chemistry, you will include tl_TC and tl_GA columns corresponding to the s4U and s6G label times, respectively.

  • If a pulse-chase experimental design was used (!!this is strongly discouraged unless you have a legitimate reason to prefer this design to a pulse-label design!!), then you should have columns named tpulse and tchase, corresponding to the pulse and chase times respectively. The same ⁠_<muttype>⁠ convention should be used in the case of multi-label pulse-chase designs.

• sample characteristics: The remaining columns can be named whatever you like and should include distinguishing features of groups of samples. Common columns might include:

  • treatment: The experimental treatment applied to a set of samples. This could represent things like genetic knockouts or knockdowns, drug treatments, etc.

  • batch: An ID for sets of samples that were collected and/or processed together. Useful for regressing out technical batch effects

Optional; name to give to fractions table.

Maximum number of characters for naming out fractions output. EZbakR will try to name this as a "_" separated character vector of all of the features analyzed. If this name is greater than character_limit, then it will default to "fraction#", where "#" represents a simple numerical ID for the table.

Data frame with the following columns:

• sample: Name given to particular sample from which data was collected.

• features: Any columns that cannot be interpreted as a mutation count or base nucleotide count (and that aren't named sample or n) will be interpreted as an ID for a genomic "feature" from which a read originated. Common examples of features and typical column names for said features include:

  • Genes; common column names: gene, gene_id, gene_name, GF

  • Genes-exonic; common column names: gene_exon, gene_id_exon, gene_name_exon, XF

  • Transcripts; common column names: transcripts, TF

  • Exonic bins; common column names: exonic_bins, EF, EB

  • Exons; common column names: exons, exon_ids

  In some cases, a read will often map to multiple features (e.g., exons). Many functions in bakR expect each of the feature IDs in these cases to be separated by +. For example, if a read overlaps with two exons, with IDs exon_1 and exon_2, then the corresponding entry in a column of exonic assignments would be "exon_1+exon_2". The default expectation can be overwritten though and is thus not strictly enforced.

• n: Number of reads with identical values for all other columns.

• kinetic parameter estimates: These can be named whatever you would like as long as they do not start with the string "se_". This should be reserved for kinetic parameter uncertainties, if provided.

• kinetic parameter uncertainties: Uncertainty in your kinetic parameter estimates. These should be named "se_" followed by the kinetic parameter as its name appears in the relevant column name of the kinetics table.

Data frame detailing various aspects of each of the samples included in the kinetics data frame. This includes:

• sample: The sample ID, which should correspond to a sample ID in the provided kinetics data frame.

• tl: Metabolic label time. There are several edge cases to be aware of:

  • If more than one metabolic label was used in the set of samples described by the metadf (e.g., s4U and s6G were used), then the tl column should be replaced by ⁠tl_<muttype>⁠, where ⁠<muttype>⁠ represents the corresponding mutation type referenced in the fractions that the label whose incubation time will be listed in this column. For example, if feeding with s4U in some samples and s6G in others, then performing standard nucleotide recoding chemistry, you will include tl_TC and tl_GA columns corresponding to the s4U and s6G label times, respectively.

  • If a pulse-chase experimental design was used (!!this is strongly discouraged unless you have a legitimate reason to prefer this design to a pulse-label design!!), then you should have columns named tpulse and tchase, corresponding to the pulse and chase times respectively. The same ⁠_<muttype>⁠ convention should be used in the case of multi-label pulse-chase designs.

• sample characteristics: The remaining columns can be named whatever you like and should include distinguishing features of groups of samples. Common columns might include:

  • treatment: The experimental treatment applied to a set of samples. This could represent things like genetic knockouts or knockdowns, drug treatments, etc.

  • batch: An ID for sets of samples that were collected and/or processed together. Useful for regressing out technical batch effects

• assay: This optional column should include a string that describes the type of experiment that was done so as to influence how EZbakR analyzes and interprets the data from those samples. Possible values for assay currently include:

  • standard: Refers to the "standard" nucleotide recoding RNA-seq methods (e.g., TimeLapse-seq, SLAM-seq, TUC-seq, etc.), in which cells are fed with a single metabolic label, RNA is extracted and sequenced, and mutations of a particular type are counted

  • STL: Refers to Start-TimeLapse-seq, a method combining Start-seq (developed by Karen Adelman's lab) with TimeLapse-seq. Used to infer the kinetics of transcription initiation and promoter-proximal pause site departure.

  • TT: Refers to Transient-Transcriptome NR-seq, a method combining TT-seq (developed by Patrick Cramer's lab) with NR-seq. TT-seq involves biochemically enriching for labeled RNA. By combining this method with nucleotide recoding chemistry (as was first done by the Simon lab with TT-TimeLapse-seq and has since been done with SLAM chemistry, often referred to as TTchem-seq), it is possible to bioinformatically filter out reads coming from unlabeled RNA background.

  • TILAC: Refers to TILAC, a method developed by the Simon lab to achieve spike-in free normalization of RNA-seq data through the use of a dual labeling approach inspired by the proteomic method SILAC.

  • subcellular: Refers to techniques such as subcellular TimeLapse-seq (developed by Stirling Churchman's lab) which combine subcellular fractionation with NR-seq to infer additional kinetic parameters.

  • sc: Refers to single-cell RNA-seq implementations of NR-seq.

Features tracked in kinetics data frame. Needs to be specified explicitly as it cannot be automatically inferred.

Optional; name to give to fractions table.

If name is chosen automatically, limit on the number of characters in said name. If default name yields a string longer than this, then kinetics table will be named kinetics1

Currently must be an EZbakRData object on which AverageAndRegularize has been run. In the future, will also support (in case where all species are assayed in every sample) providing output of just EstimateFractions() as input, acting as a generalization of EstimateKinetics() in that case.

An NxN adjacency matrix, where N represents the number of species being modeled. One of these species must be called "0" and represent the "no RNA" species. This is the species from which some species are synthesized (e.g., 0 -> P, means premature RNA is synthesized from no RNA), and the species to which some species are degraded (e.g., M -> 0 means mature RNA is converted to "no RNA" via degradation). The rows and columns of this matrix must be the names of all modled species, and rownames(graph) == colnames(graph). Entry i,j of the matrix is either 0 if species i cannot be converted into species j under your model, and an integer from 1:npars (where npars = total number of parameters to be estimated) if it can.

For example, the model 0 -> P -> M -> 0 would have the graph: ⁠matrix(c(0, 1, 0, 0, 0, 2, 3, 0, 0), nrow = 3, ncol = 3, byrow = TRUE⁠.

Which feature columns distinguish between the different measured species? Note, the measured species need not have the same name, and may not be directly equivalent to, the modeled species. The relationship between the modeled species in graph and sub_features needs to be specified in modeled_to_measured if the names are not equivalent though.

Which features are the overarching feature assignments by which sub_features should be grouped? This will usually be the feature columns corresponding to full-gene assignments, as well as any higher order assignments (e.g., chromosome). A sub_feature can be included in grouping_features if it never has the value of unassigned_name ("__no_feature" by default). Only one sub_feature should ever fulfill this criterion though.

Data frame mapping samples to factors by which to multiply read counts so as ensure proper normalization between different RNA populations. Only relevant if you are modeling relationships between distinct RNA populations, for example RNA from nuclear and cytoplasmic fractions. Will eventually be inferred automatically.

If different samples involve assaying different species, this must be the name of the metadf column that distinguishes the different classes of samples. For example, if analyzing a subcellular fractionation dataset, you likely included a column in your metadf called "compartment". This would then be your sample_feature, assuming you ran AverageAndRegularize(), with a mean_formula that included compartment as a term.

If sub_features is not identical to the non "0" species names in graph, then you must specify the relationship between sub_features, sample_feature (if specified), and the species in graph. This is done through a list of formulas, whose left side is a sub_feature and whose right side is a function of species in graph. If sample_feature is not specified, then modeled_to_measured should be a nameless list of formulae. If sample_feature is specified, then modeled_to_measured must be a named list of formulas where the names correspond to unique values of sample_feature. In this latter case, the elements, should be the mapping of measured features (sub_features) to modeled species (names in graph that aren't "0").

For example, if your model is 0 -> P -> M -> 0, where P stands for premature RNA and M stands for mature RNA, and you have a column called GF that corresponds to assignment anywhere in the gene, and XF that corresponds to assignment of exclusively exon mapping reads, then your modeled_to_measured should be ⁠list(GF ~ P + M, XF ~ M).⁠.

As another example, if your model is 0 -> N -> C -> 0, where N stands for "nuclear RNA" and C stands for "cytoplasmic RNA", and your sample_feature takes on values of "nuclear", "cytoplasmic", and "total", and you have a single sub_feature called XF, then your modeled_to_measured shouild be list(nuclear = GF ~ N, cytoplasmic = GF ~ C, total = GF ~ C + N). This is interpreted as meaning in groups of samples for which sample_feature == "nuclear", reads assigned to a GF are from the N species (nuclear RNA). When sample_feature == "cytoplasmic", reads assigned to GF correspond to the C species (cytoplasmic RNA). When sample_feature == "total", reads assigned to GF correspond to a combination of N and C (nuclear and cytoplasmic RNA).

Vector of names you would like to give to the estimated parameters. ith element should correspond to name of parameter given the ID i in graph. By default, this is just ki, where i is this numerical index.

What value will a sub_feature column have if a read was not assigned to said feature? "__no_feature" by default.

What type of table would you like to use? Currently only supports "averages", but will support "fractions" in the near future.

Mean of log-Normal prior for kinetic parameters. Should be vector where ith value is mean for ith parameter, i = index in graph

Std. dev. of log-Normal prior for kinetic parameter. Should be vector where ith value is mean for ith parameter, i = index in graph.

Names of parameters in averages table that you would like to keep. Other parameters will be discarded. Don't include the prefixes "mean_", "sd_", or "coverage_"; these will be imputed automatically. In other words, this should be the base parameter names.

Names of parameters in averages table that you would like to drop. Other parameters will be kept. Don't include the prefixes "mean_", "sd_", or "coverage_"; these will be imputed automatically. In other words, this should be the base parameter names.

Name of relevant label time column that will be found in the parameter names. Defaults to the standard "tl".

Character vector of the set of features you want to stratify reads by and estimate proportions of each RNA population. The default of "all" will use all feature columns in the obj's cB.

Mutational populations that were analyzed to generate the fractions table to use. For example, this would be "TC" for a standard s4U-based nucleotide recoding experiment.

"Design matrix" specifying which RNA populations exist in your samples. By default, this will be created automatically and will assume that all combinations of the mutrate_populations you have requested to analyze are present in your data. If this is not the case for your data, then you will have to create one manually. See docs for EstimateFractions (run ?EstimateFractions()) for more details.

Parameter to average across replicates of a given condition. Has to be ⁠logit_fraction_high<muttype>⁠, where ⁠<muttype>⁠ is the type of mutation modeled in EstimateFractions() (e.g, TC) in this case.

If multiple fractions tables exist with the same metadata, then this is the numerical index by which they are distinguished.

If TRUE, then features and populations have to exactly match those for a given fractions table for that table to be used. Means that you can't specify a subset of features or populations by default, since this is TRUE by default.

Table of effective lengths for each feature combination in your data. For example, if your analysis includes features named GF and XF, this should be a data frame with columns GF, XF, and length.

If TRUE, normalized read counts will be used to inform kinetic parameter estimates. If FALSE, only fraction news will be used, which will leave some parameters (e.g., synthesis rate) unidentifiable, though has the advantage of avoiding the potential biases induced by normalization problems.

For extracting the fractions table needed for normalization of multi-sample data. If multiple fractions tables exist with the same metadata, then this is the numerical index by which they are distinguished.

For extracting the fractions table needed for normalization of multi-sample data. If TRUE, then features and populations have to exactly match those for a given fractions table for that table to be used. Means that you can't specify a subset of features or populations by default, since this is TRUE by default.

List mapping individual RNA species in graph to different sample_feature values (sf). This is relevant if you are modeling both pre- and mature RNA dynamics in subcellular fractionation data. EZbakR can usually automatically infer this, but if not, then you can manually specify this mapping. Should be a list with one element per unique sample_feature type. Each element should be a vector of the RNA species in graph that belong to that sample_feature type. For example, if you have whole cell, cytoplasmic, and nuclear fraction data, this should have one element called "cytoplasmic" and one element called "nuclear". The whole cell sample_feature is a sum of cytoplasmic and nuclear and thus does not apply. The "cytoplasmic" element of this list should be the set of RNA species that are present in the cytoplasm, e.g. cytoplasmic pre-RNA (maybe referred to in graph as CP) and cytoplasmic mature RNA (maybe referred to in graph as CM).

If TRUE and a fractions estimate output already exists that would possess the same metadata (features analyzed, populations analyzed, and fraction_design), then it will get overwritten with the new output. Else, it will be saved as a separate output with the same name + "_#" where "#" is a numerical ID to distinguish the similar outputs.

EZbakRData object

The class of EZbakR outputs would you like to search through. Equivalent to the name of the list in the EZbakRData object that contains the tables of interest.

Features that must be present in the table of interest. If exactMatch is TRUE, then these features must also be the only features present in the table.

Only relevant if type == "fractions". Mutational populations that must have been analyzed to generate the table of interest.

Only relevant if type == "fractions". Fraction design table used to generate the table of interest.

If the relevant table is the result of isoform deconvolution

Only relevant if type == "kinetics". Short for "kinetics strategy"; the strategy used to infer kinetic parameters.

Only relevant if type == "averages" or "comparisons". Which parameter was being averaged or compared?

String denoting what type of read count information you are looking for. Current options are "TMM_normalized", "transcript", and "matrix". TO-DO: Not sure this is being used in any way currently...

design_factor specified in relevant run of CompareParameters(). Therefore, only relevant if type == "comparisons".

design_factors included in final EZDynamics() output. Therefore, only relevant if type == "dynamics".

Sample group scale factors used in EZDynamics(). Therefore, only relevant if type == "dynamics"

Strategy used for comparative analyses. Can be:

• contrast: If two parameters were compared via specifying the reference and experimental levels in CompareParameters() (for type == "averages").

• single_param: If a single parameter was passed to CompareParameters() via its param_name option.

• dynamics: If output of EZDynamics() was passed to CompareParameters()

• function: If function of multiple parameters was passed to CompareParameter() via its param_function option.

Table of feature lengths used for length normalization.

Experimental condition specified in relevant run of CompareParameters(). Therefore, only relevant if type == "comparisons".

Reference condition specified in relevant run of CompareParameters(). Therefore, only relevant if type == "comparisons".

Parameter name specified in relevant run of CompareParameters(). Therefore, only relevant if type == "comparisons"

Function of parameters specified in relevant run of CompareParameters(). Therefore, only relevant if type == "comparisons".

An R formula object specifying how the parameter of interest depends on the sample characteristics specified in obj's metadf. Therefore, only relevant if type == "averages".

What metadf columns should data be grouped by when estimating standard deviations across replicates? Therefore, only relevant if type == "averages".

Character vector of names of parameters in linear model fit. Therefore, only relevant if type == "averages".

Numerical ID for duplicate objects with same metadata.

Only relevant if type == "dynamics". Feature columns that distinguished between the different measured species when running EZDynamics().

Only relevant if type == "dynamics. Features that were the overarching feature assignments by which sub_features were grouped when running EZDynamics().

Only relevant if type == "dynamics". Name of the metadf column that distinguished the different classes of samples when running EZDynamics().

Only relevant if type == "dynamics". Specifies the relationship between sub_features, sample_feature (if specified), and the species in graph.

Only relevant if type == "dynamics". NxN adjacency matrix, where N represents the number of species modeled when running EZDynamics().

Whether median difference was subtracted from estimated kinetic parameter difference. Thus, only relevant if type == "comparisons".

Only relevant if type == "fractions". Boolean that is TRUE if fractions table is result of performing multi-feature deconvolution with DeconvolveFractions().

If TRUE, then only the names of tables that passed your search criteria will be returned. Else, the single table passing your search criteria will be returned. If there is more than one table that passes your search criteria and returnNameOnly == FALSE, an error will be thrown.

If TRUE, then for features and populations, which can be vectors, ensure that provided vectors of features and populations exactly match the relevant metadata vectors.

If TRUE, then even if there is only a single table for the type of interest, still run all checks against queries.

An object of class EZbakRCompare, which is an EZbakRData object on which you have run CompareParameters

Name of parameter whose comparison you want to plot.

Name of factor from metadf whose parameter estimates at different factor values you would like to compare.

Name of reference condition factor level value.

Name of condition factor level value to compare to reference.

If you want to assess the significance of a single parameter, rather than the comparison of two parameters, specify that one parameter's name here.

NOT YET IMPLEMENTED. Will allow you to specify more complicated functions of parameters when hypotheses you need to test are combinations of parameters rather than individual parameters or simple differences in two parameters.

Character vector of feature names for which comparisons were made.

Defunct parameter that has been replaced with design_factor. If provided gets passed to design_factor if design_factor is not already specified.

If multiple kinetics or fractions tables exist with the same metadata, then this is the numerical index by which they are distinguished.

If TRUE, then features and populations have to exactly match those for a given fractions table for that table to be used. Means that you can't specify a subset of features or populations by default, since this is TRUE by default.

If TRUE, assume that log(parameter) difference is passed in and that you want to plot log2(parameter) difference.

False discovery cutoff by which to color points.

Minimum absolute difference cutoff by which to color points.

Size of points, passed to geom_point() size parameter. If not specified, a point size is automatically chosen.

Features you want to highlight in the plot (black circle will be drawn around them). This can either be a data frame with one column per feature type in the comparison table you are visualizing, or a vector of feature names if the relevant comparison table will only have one feature type noted.

Shape of points overlayed on highlighted features. Defaults to an open circle

Sets how much larger should the points overlayed on the highlighted features be than the original plot points.

Stroke width of the points overlayed on the highlighted features.

Fill color of the points overlayed on the highlighted features. Default is for them to be fill-less (highlight_fill == NA).

Color of the points overlayed on the highlighted points.

An object of class EZbakRCompare, which is an EZbakRData object on which you have run CompareParameters

Specifies what data to use for the PCA. Options are "fraction_labeled" (default; means using fraction high T-to-C or other mutation type estimate from EZbakR) or "reads" (means using log10(read counts + 1)).

Character vector of feature names for which comparisons were made.

If TRUE, then features and populations have to exactly match those for a given fractions table for that table to be used. Means that you can't specify a subset of features or populations by default, since this is TRUE by default.

Integer between (inclusive) 1 and 9. Features with sample-to-sample variance greater than the nth decile (n = variance_decile) will go into PCA.

From prcomp(): a logical value indicating whether the variables should be shifted to be zero centered. Alternately, a vector of length equal the number of columns of x can be supplied. The value is passed to scale.

From prcomp(): a logical value indicating whether the variables should be scaled to have unit variance before the analysis takes place. Alternatively, a vector of length equal the number of columns of x can be supplied. The value is passed to scale.

Size of points in PCA plot

Columns in the EZbakR metadf that will be used to color points in the PCA plot. Points will be colored by the interaction between all of these columns (i.e., samples with unique combinations of values of these columns will get unique colors). Default is to use all columns (except "sample"), specified as "all".

EZbakRData or EZbakRFractions object.

Parameters passed to the class-specific method. If you have provided an EZbakRFractions object, then these can be (all play the same role as in EstimateKinetics(), that is they get passed to EZget() to find the fractions table you are interested in. See ?EstimateKinetics() for details.):

• features

• populations

• fraction_design

If you have provided an EZbakRData object, then these can be (all same the same purpose as in EstimateFractions, so see ?EstimateFractions() for details):

• mutrate_populations

• features

• filter_cols

• filter_condition

• remove_features

An EZbakRData object

Same as in EstimateFractions(). See ?EstimateFractions() for details.

Same as in EstimateFractions(). See ?EstimateFractions() for details.

Same as in EstimateFractions(). See ?EstimateFractions() for details.

Same as in EstimateFractions(). See ?EstimateFractions() for details.

Same as in EstimateFractions(). See ?EstimateFractions() for details.

Additional arguments. Currently goes unused.

An EZbakRData object

Same as in EstimateFractions(). See ?EstimateFractions() for details.

Same as in EstimateFractions(). See ?EstimateFractions() for details.

Same as in EstimateFractions(). See ?EstimateFractions() for details.

Same as in EstimateFractions(). See ?EstimateFractions() for details.

Same as in EstimateFractions(). See ?EstimateFractions() for details.

Additional arguments. Currently goes unused.

EZbakRFractions object, which is an EZbakRData object on which EstimateFractions has been run.

Set of features analyzed in the fractions table you are interested QCing. This gets passed to EZget() to help find this table.

Set of mutation types analyzed in the fractions table you are interested in QCing. This gets passed to EZget() to help find this table.

The fraction "design matrix" specified to get the fractions table you are interested in QCing. This gets passed to EZget() to help find this table.

Additional arguments. Currently goes unused.

Number of "features" (e.g., genes) for which to simulated data.

Currently, EZSimulate can simulate in two modes: "standard" and "dynamics". The former is the default and involves simulating multiple conditions of standard NR-seq data. "dynamics" calls SimulateDynamics() under the hood to simulate a dynamical systems model of your choice. Most of the additional parameters do not apply if mode == "dynamics", except for those from dynamics_preset and on.

Number of distinct treatments to simulate. This parameter is only relevant if metadf is not provided.

Number of replicates of each treatment to simulate. This parameter is only relevant if metadf is not provided

Number of -s4U replicates of each treatment to simulate. This parameter is only relevant if metadf is not provided.

A data frame with the following columns:

• sample: Names given to samples to simulate.

• <details>: Any number of columns with any names (not taken by other metadf columns) storing factors by which the samples can be stratified. These can be referenced in mean_formula, described below.

These parameters (described more below) can also be included in metadf to specify sample-specific simulation parameter:

• seqdepth

• label_time

• pnew

• pold

• readlength

• Ucont

A formula object that specifies the linear model used to relate the factors in the <details> columns of metadf to average log(kdegs) and log(ksyns) in each sample.

A data frame with one row for each column of the design matrix obtained from model.matrix(mean_formula, metadf) that describes how to simulate the linear model parameters. The columns of this data frame are:

• param: Name of linear model parameter as it appears in the column names of the design matrix from model.matrix(mean_formula, metadf).

• reference: Boolean; TRUE if you want to treat that parameter as a "reference". This means that all other parameter values that aren't global parameters are set equal to this unless otherwise determined (see ⁠pdiff_*⁠ parameters for how it is determined if a parameter will differ from the reference).

• global: Boolean; TRUE if you want to treat that parameter as a global parameter. This means that a single value is used for all features.

• logkdeg_mean: If parameter is the reference, then its value for the log(kdeg) linear model will be drawn from a normal distribution with this mean. If it is a global parameter, then this value will be used. If it is neither of these, then its value in the log(kdeg) linear model will either be the reference (if there is no difference between this condition's value and the reference) or the reference's value + a normally distributed random variable centered on this value.

• logkdeg_sd: sd used for draws from normal distribution as described for logkdeg_mean.

• logksyn_mean: Same as logkdeg_mean but for log(ksyn) linear model.

• logksyn_sd: Same as logkdeg_sd but for log(kdeg) linear model.

• pdiff_ks: Proportion of features whose value of this parameter in the log(ksyn) linear model will differ from the reference's. Should be a number between 0 and 1, inclusive. For example, if pdiff_ks is 0.1, then for 10% of features, this parameter will equal the reference parameter + a normally distributed random variable with mean logksyn_mean and sd logksyn_sd. For the other 90% of features, this parameter will equal the reference.

• pdiff_kd: Same as pdiff_ks but for log(kdeg) linear model.

• pdiff_both: Proportion of features whose value for this parameter in BOTH the log(kdeg) and log(ksyn) linear models will differ from the reference. Value must be between 0 and min(c(pdiff_kd, pdiff_ks)) in that row.

If param_details is not specified by the user, the first column of the design matrix is assumed to represent the reference parameter, all parameters are assumed to be non-global, logkdeg_mean and logksyn_mean are set to the equivalently named parameter values described below for the reference and logkdeg_diff_avg and logksyn_diff_avg for all other parameters, logkdeg_sd and logksyn_sd are set to the equivalently named parameter values described below for the reference and logkdeg_diff_sd and logksyn_diff_sd for all other parameters, and pdiff_kd, pdiff_ks, and pdiff_both are all set to the equivalently named parameter values.

Total number of reads in each sample.

Length of s^4^U feed to simulate.

Probability that a T is mutated to a C if a read is new.

Probability that a T is mutated to a C if a read is old.

Length of simulated reads. In this simple simulation, all reads are simulated as being exactly this length.

Probability that a nucleotide in a simulated read from a given feature is a U is drawn from a beta distribution with shape1 = Ucont_alpha.

Probability that a nucleotide in a simulated read from a given feature is a U is drawn from a beta distribution with shape2 = Ucont_beta.

Name given to the i-th feature is paste0(feature_prefix, i). Shows up in the feature column of the output simulated data table.

Negative binomial dispersion parameter "slope" with respect to read counts. See DESeq2 paper for dispersion model used.

Negative binomial dispersion parameter "intercept" with respect to read counts. See DESeq2 paper for dispersion model used.

Slope for log10(read count) vs. log(kdeg) replicate variability trend

Intercept for log10(read count) vs. log(kdeg) replicate variability trend

Slope for log10(read count) vs. log(ksyn) replicate variability trend

Intercept for log10(read count) vs. log(ksyn) replicate variability trend

Mean of normal distribution from which reference log(kdeg) linear model parameter is drawn from for each feature if param_details is not provided.

Standard deviation of normal distribution from which reference log(kdeg) linear model parameter is drawn from for each feature if param_details is not provided.

Mean of normal distribution from which reference log(ksyn) linear model parameter is drawn from for each feature if param_details is not provided.

Standard deviation of normal distribution from which reference log(ksyn) linear model parameter is drawn from for each feature if param_details is not provided.

Mean of normal distribution from which non-reference log(kdeg) linear model parameters are drawn from for each feature if param_details is not provided.

Mean of normal distribution from which reference log(ksyn) linear model parameter are drawn from for each feature if param_details is not provided.

Standard deviation of normal distribution from which reference log(kdeg) linear model parameter are drawn from for each feature if param_details is not provided.

Standard deviation of normal distribution from which reference log(ksyn) linear model parameter are drawn from for each feature if param_details is not provided.

Proportion of features for which non-reference log(kdeg) linear model parameters differ from the reference.

Proportion of features for which non-reference log(ksyn) linear model parameters differ from the reference.

Proportion of features for which BOTH non-reference log(kdeg) and log(ksyn) linear model parameters differ from the reference.

Dropout rate; think of this as the probability that a s4U containing molecule is lost during library preparation and sequencing. If pdo is 0 (default) then there is not dropout.

Which preset model to use for simulation of dynamics. Therefore, only relevant if mode == dynamics. Options are:

nuc2cyto

Simplest model of nuclear and cytoplasmic RNA dynamics: 0 -> N -> C -> 0

preRNA

Simplest model of pre-RNA and mature RNA dynamics: 0 -> P -> M -> 0

preRNAwithPdeg

Same as preRNA, but now pre-RNA can also degrade.

nuc2cytowithNdeg

Same as nuc2cyto, but now nuclear RNA can also degrade.

subtlseq

Subcellular TimeLapse-seq model, similar to that described in Ietswaart et al., 2024. Simplest model discussed there, lacking nuclear degradation: 0 -> CH -> NP -> CY -> PL -> 0, and CY can also degrade.

nuc2cytowithpreRNA

Combination of nuc2cyto and preRNA where preRNA is first synthesized, then either processed or exported to the cytoplasm. Processing can also occur in the cytoplasm, and mature nuclear RNA can be exported to the cytoplasm. Only mature RNA degrades.

String to give to reads not assigned to a given feature.

Negative binomial size parameter to use for simulating read counts

Logit(fn) replicate variability.

Data frame describing effects of treatment on each parameter. Should have five columns: "parameter_index", "treatment_index", "mean", "sd", and "fraction_affected". Each row corresponds to the effect the ith (i = treatment_index) treatment has on the jth (j = parameter_index) kinetic parameter. Effect sizes, on a log-scale, are drawn from a Normal distribution with mean and standard deviation set by the mean and sd columns, respectively. The number of non-zero effects is set by "fraction_affected", and is equal to ceiling(nfeatures * fraction_affected). treatment_index of 1 will be ignored and can either be included or not.

If ntreatments > 1, and treatment_effects is not provided, this will be the value of mean for all treatments and parameters imputed in treatment_effects.

If ntreatments > 1, and treatment_effects is not provided, this will be the value of sd for all treatments and parameters imputed in treatment_effects.

If ntreatments > 1, and treatment_effects is not provided, this will be the value of fraction_affected for all treatments and parameters imputed in treatment_effects.

Vector of log-Normal logmeans from which the distribution of feature-specific parameters will be drawn from. Length of vector should be the same as max(entries in graph), i.e., the number of parameters in your specified model. If not provided, will by default be c(1, seq(from = -0.3, to = -2.5, length.out = max(graph) - 1 )). 1 for the ksyn parameter (which is always denoted 1 in the preset graph) is arbitrary. Remaining parameters will make it so indices order parameters from fastest to slowest process.

Vector of log-Normal logsds from which the distribution of feature-specific parameters will be drawn from. If not provided, will be 0.4 for all parameters.

An object of class EZbakRCompare, which is an EZbakRData object on which you have run CompareParameters

Name of parameter whose comparison you want to plot.

Name of factor from metadf whose parameter estimates at different factor values you would like to compare.

Name of reference condition factor level value.

Name of condition factor level value to compare to reference.

If you want to assess the significance of a single parameter, rather than the comparison of two parameters, specify that one parameter's name here.

NOT YET IMPLEMENTED. Will allow you to specify more complicated functions of parameters when hypotheses you need to test are combinations of parameters rather than individual parameters or simple differences in two parameters.

Character vector of feature names for which comparisons were made.

Defunct parameter that has been replaced with design_factor. If provided gets passed to design_factor if design_factor is not already specified.

Whether or not the median was subtracted from the estimated parameter differences.

If multiple kinetics or fractions tables exist with the same metadata, then this is the numerical index by which they are distinguished.

If TRUE, then features has to exactly match those for a given comparisons table for that table to be used. Means that you can't specify a subset of features by default, since this is TRUE by default.

If TRUE, assume that log(parameter) difference is passed in and that you want to plot log2(parameter) difference. TO-DO: probably best to change this to a more general scale parameter by which the parameter is multiplied. Default would be log2(exp(1)) to convert log() to log2().

False discovery cutoff by which to color points.

Minimum absolute difference cutoff by which to color points.

Size of points, passed to geom_point() size parameter. If not specified, a point size is automatically chosen.

Features you want to highlight in the plot (black circle will be drawn around them). This can either be a data frame with one column per feature type in the comparison table you are visualizing, or a vector of feature names if the relevant comparison table will only have one feature type noted.

Shape of points overlayed on highlighted features. Defaults to an open circle

Sets how much larger should the points overlayed on the highlighted features be than the original plot points.

Stroke width of the points overlayed on the highlighted features.

Fill color of the points overlayed on the highlighted features. Default is for them to be fill-less (highlight_fill == NA).

Stroke color of the points overlayed on the highlighted points.

An EZbakRData or EZbakRFractions object.

Features in relevant table

Name of fractions table to use

Table of effective lengths for each feature combination in your data. For example, if your analysis includes features named GF and XF, this should be a data frame with columns GF, XF, and length.

Dataframe with two columns, one being "sample" (sample names) and the other being "scale_factor" (value to divide read counts by to normalize them)

An EZbakRData object.

A named vector of paths to all transcript quantification files that you would like to import. This will be passed as the first argument of tximport::tximport() (also named files). The names of this vector should be the same as the sample names as they appear in the metadf of the EZbakRData object.

String denoting the type of software used to generate the abundances. Will get passed to the type argument of tximport::tximport(). As described in the documentation for tximport 'Options are "salmon", "sailfish", "alevin", "piscem", "kallisto", "rsem", "stringtie", or "none". This argument is used to autofill the arguments below (geneIdCol, etc.) "none" means that the user will specify these columns. Be aware that specifying type other than "none" will ignore the arguments below (geneIdCol, etc.)'. Referenced 'arguments below' can be specified as part of ....

Whether or now you are providing isoform level quantification files. Alternative (txIn = FALSE) is gene-level quantification. In ImportIsoformQuant, txIn gets passed to BOTH the txIn and txOut parameters in tximport().

Additional arguments to be passed to tximport::tximport(). Especially relevant if you set quant_tool to "none".

Arrow dataset tracking the sample ID, mutational and nucleotide content, and feature assignment of sequencing reads.

Data frame tracking features of each of the samples included in cBds.

Data frame tracking the sample ID, mutational and nucleotide content, and feature assignment of sequencing reads.

Data frame tracking features of each of the samples included in cB.

Data frame containing information about the fraction of reads from each mutational population of interest.

Data frame reporting aspects of each of the samples included

Optional; name to give to fractions table.

Maximum number of characters for naming out fractions output. EZbakR will try to name this as a "_" separated character vector of all of the features analyzed. If this name is greater than character_limit, then it will default to "fraction#", where "#" represents a simple numerical ID for the table. in fractions

Data frame containing information about the kinetic parameters of interest for each set of features tracked.

Features tracked in kinetics data frame. Needs to be specified explicitly as it cannot be automatically inferred.

Data frame describing each of the samples included

Optional; name to give to fractions table.

Maximum number of characters for naming out fractions output. EZbakR will try to name this as a "_" separated character vector of all of the features analyzed. If this name is greater than character_limit, then it will default to "fraction#", where "#" represents a simple numerical ID for the table. in kinetics

An EZbakRFractions object, which is an EZbakRData object on which you have run EstimateFractions().

If TRUE, samples from different label times will be normalized by finding a max inferred degradation rate constant (kdeg) sample and using that as a reference. Degradation kinetics with this max will be assumed to infer reference fraction news at different label times

Which sample-detail columns in the metadf should be used to group -s4U samples by for calculating the average -s4U RPM? The default value of NULL will cause no sample-detail columns to be used.

Character vector of the set of features you want to stratify reads by and estimate proportions of each RNA population. The default of NULL will expect there to be only one fractions table in the EZbakRFractions object.

Mutational populations that were analyzed to generate the fractions table to use. For example, this would be "TC" for a standard s4U-based nucleotide recoding experiment.

"Design matrix" specifying which RNA populations exist in your samples. By default, this will be created automatically and will assume that all combinations of the mutrate_populations you have requested to analyze are present in your data. If this is not the case for your data, then you will have to create one manually. See docs for EstimateFractions (run ?EstimateFractions()) for more details.

If multiple fractions tables exist with the same metadata, then this is the numerical index by which they are distinguished.

If TRUE, then features must exactly match the features metadata for a given fractions table for it to be used. Means that you cannot specify a subset of features by default. Set this to FALSE if you would like to specify a feature subset.

Minimum number of reads for a feature to be used to fit dropout model.

An EZbakRArrowData object.

Maximum number of characters to print on each line

Ignored

An EZbakRData object.

Maximum number of characters to print on each line

Ignored

Number of reads to simulate

Fraction of reads that are new in simulation. Whether a read will be new will be determined by a draw from a Bernoulli(fn) distribution.

T-to-C mutation rate in new reads

T-to-C mutation rate in old reads

Length of simulated reads

Fraction of nucleotides in simulated reads that are Ts (U in RNA)

Number of "features" to simulate data for. A "feature" in this case may contain a number of "sub-features". For example, you may want to simulate pre-RNA and mature RNA for a set of "genes", in which case the number of features is the number of genes.

An adjacency matrix describing the reaction diagram graph relating the various RNA species to one another.

Data frame with two required columns (sample and tl). sample represents names given to each simulated sample. tl represents the label time for that sample. Additional columns can specify other features of the sample, like what subcellular compartment the sample is taken from. NOTE: Not sure I am actually using these optional columns in any useful capacity anymore.

Vector of log-Normal logmeans from which the distribution of feature-specific parameters will be drawn from. Length of vector should be the same as max(entries in graph), i.e., the number of parameters in your specified model.

Vector of log-Normal logsds from which the distribution of feature-specific parameters will be drawn from.

Number of distinct experimental treatments to simulate. By default, only a single "treatment" (you might refer to this as wild-type, or control) is simulated. Increase this if you would like to explore performing comparative dynamical systems modeling.

Data frame describing effects of treatment on each parameter. Should have five columns: "parameter_index", "treatment_index", "mean", "sd", and "fraction_affected". Each row corresponds to the effect the ith (i = treatment_index) treatment has on the jth (j = parameter_index) kinetic parameter. Effect sizes, on a log-scale, are drawn from a Normal distribution with mean and standard deviation set by the mean and sd columns, respectively. The number of non-zero effects is set by "fraction_affected", and is equal to ceiling(nfeatures * fraction_affected). treatment_index of 1 will be ignored and can either be included or not.

A list of named lists. The names of each sub-list should be the same as the sample names as they are found in metadf. Each sub-list should be a list of formula relating feature names that will show up as columns of the simulated cB to species modeled in your graph. This only needs to be specified if you want to simulate the scenario where some of the measured species are a sum of modeled species.

String to give to reads not assigned to a given feature.

Total number of reads in each sample.

Negative binomial size parameter to use for simulating read counts

Logit(fn) replicate variability.

If ntreatments > 1, and treatment_effects is not provided, this will be the value of mean for all treatments and parameters imputed in treatment_effects.

If ntreatments > 1, and treatment_effects is not provided, this will be the value of sd for all treatments and parameters imputed in treatment_effects.

If ntreatments > 1, and treatment_effects is not provided, this will be the value of fraction_affected for all treatments and parameters imputed in treatment_effects.

Parameters passed to SimulateOneRep().

Number of "features" to simulate data for. Each feature will have a simulated number of transcript isoforms

(Optional), can provide a vector of the number of isoforms you would like to simulate for each of the nfeatures features. Vector can either be length 1, in which case that many isoforms will be simulated for all features, or length equal to nfeatures.

Total number of sequencing reads to simulate

Length of s^4^U feed to simulate.

Character vector to assign to sample column of output simulated data table (the cB table).

Name given to the i-th feature is paste0(feature_prefix, i). Shows up in the feature column of the output simulated data table.

Probability that a T is mutated to a C if a read is new.

Probability that a T is mutated to a C if a read is old.

Fraction of reads that uniquely "map" to a single isoform.

Length of simulated reads. In this simple simulation, all reads are simulated as being exactly this length.

Percentage of nucleotides simulated to be U's.

Average number of isoforms for each feature. Feature-specific isoform counts are drawn from a Poisson distribution with this average. NOTE: to insure that all features have multiple isoforms, the simulated number of isoforms drawn from a Poisson distribution is incremented by 2. Thus, the actual average number of isoforms from each feature is avg_numiso + 2.

Percentage of genes for which all isoform abundance differences are synthesis driven. If not synthesis driven, then isoform abundance differences will be driven by differences in isoform kdegs.

meanlog of a log-normal distribution from which kdegs are simulated

sdlog of a log-normal distribution from which kdegs are simulated

meanlog of a log-normal distribution from which ksyns are simulated

sdlog of a log-normal distribution from which ksyns are simulated

Number of "features" (e.g., genes) to simulate data for

A data frame with the following columns:

• sample: Names given to samples to simulate.

• <details>: Any number of columns with any names (not taken by other metadf columns) storing factors by which the samples can be stratified. These can be referenced in mean_formula, described below.

These parameters (described more below) can also be included in metadf to specify sample-specific simulation parameter:

• seqdepth

• label_time

• pnew

• pold

• readlength

• Ucont

A formula object that specifies the linear model used to relate the factors in the <details> columns of metadf to average log(kdegs) and log(ksyns) in each sample.

A data frame with one row for each column of the design matrix obtained from model.matrix(mean_formula, metadf) that describes how to simulate the linear model parameters. The columns of this data frame are:

• param: Name of linear model parameter as it appears in the column names of the design matrix from model.matrix(mean_formula, metadf).

• reference: Boolean; TRUE if you want to treat that parameter as a "reference". This means that all other parameter values that aren't global parameters are set equal to this unless otherwise determined (see ⁠pdiff_*⁠ parameters for how it is determined if a parameter will differ from the reference).

• global: Boolean; TRUE if you want to treat that parameter as a global parameter. This means that a single value is used for all features.

• logkdeg_mean: If parameter is the reference, then its value for the log(kdeg) linear model will be drawn from a normal distribution with this mean. If it is a global parameter, then this value will be used. If it is neither of these, then its value in the log(kdeg) linear model will either be the reference (if there is no difference between this condition's value and the reference) or the reference's value + a normally distributed random variable centered on this value.

• logkdeg_sd: sd used for draws from normal distribution as described for logkdeg_mean.

• logksyn_mean: Same as logkdeg_mean but for log(ksyn) linear model.

• logksyn_sd: Same as logkdeg_sd but for log(kdeg) linear model.

• pdiff_ks: Proportion of features whose value of this parameter in the log(ksyn) linear model will differ from the reference's. Should be a number between 0 and 1, inclusive. For example, if pdiff_ks is 0.1, then for 10% of features, this parameter will equal the reference parameter + a normally distributed random variable with mean logksyn_mean and sd logksyn_sd. For the other 90% of features, this parameter will equal the reference.

• pdiff_kd: Same as pdiff_ks but for log(kdeg) linear model.

• pdiff_both: Proportion of features whose value for this parameter in BOTH the log(kdeg) and log(ksyn) linear models will differ from the reference. Value must be between 0 and min(c(pdiff_kd, pdiff_ks)) in that row.

If param_details is not specified by the user, the first column of the design matrix is assumed to represent the reference parameter, all parameters are assumed to be non-global, logkdeg_mean and logksyn_mean are set to the equivalently named parameter values described below for the reference and logkdeg_diff_avg and logksyn_diff_avg for all other parameters, logkdeg_sd and logksyn_sd are set to the equivalently named parameter values described below for the reference and logkdeg_diff_sd and logksyn_diff_sd for all other parameters, and pdiff_kd, pdiff_ks, and pdiff_both are all set to the equivalently named parameter values.

Only relevant if read_vect is not provided; in that case, this is the total number of reads to simulate.

Length of s^4^U feed to simulate.

Probability that a T is mutated to a C if a read is new.

Probability that a T is mutated to a C if a read is old.

Length of simulated reads. In this simple simulation, all reads are simulated as being exactly this length.

Probability that a nucleotide in a simulated read from a given feature is a U is drawn from a beta distribution with shape1 = Ucont_alpha.

Probability that a nucleotide in a simulated read from a given feature is a U is drawn from a beta distribution with shape2 = Ucont_beta.

Name given to the i-th feature is paste0(feature_prefix, i). Shows up in the feature column of the output simulated data table.

Negative binomial dispersion parameter "slope" with respect to read counts. See DESeq2 paper for dispersion model used.

Negative binomial dispersion parameter "intercept" with respect to read counts. See DESeq2 paper for dispersion model used.

Slope for log10(read count) vs. log(kdeg) replicate variability trend

Intercept for log10(read count) vs. log(kdeg) replicate variability trend

Slope for log10(read count) vs. log(ksyn) replicate variability trend

Intercept for log10(read count) vs. log(ksyn) replicate variability trend

Mean of normal distribution from which reference log(kdeg) linear model parameter is drawn from for each feature if param_details is not provided.

Standard deviation of normal distribution from which reference log(kdeg) linear model parameter is drawn from for each feature if param_details is not provided.

Mean of normal distribution from which reference log(ksyn) linear model parameter is drawn from for each feature if param_details is not provided.

Standard deviation of normal distribution from which reference log(ksyn) linear model parameter is drawn from for each feature if param_details is not provided.

Mean of normal distribution from which non-reference log(kdeg) linear model parameters are drawn from for each feature if param_details is not provided.

Mean of normal distribution from which reference log(ksyn) linear model parameter are drawn from for each feature if param_details is not provided.

Standard deviation of normal distribution from which reference log(kdeg) linear model parameter are drawn from for each feature if param_details is not provided.

Standard deviation of normal distribution from which reference log(ksyn) linear model parameter are drawn from for each feature if param_details is not provided.

Proportion of features for which non-reference log(kdeg) linear model parameters differ from the reference.

Proportion of features for which non-reference log(ksyn) linear model parameters differ from the reference.

Proportion of features for which BOTH non-reference log(kdeg) and log(ksyn) linear model parameters differ from the reference. ksyns are simulated

Dropout rate; think of this as the probability that a s4U containing molecule is lost during library preparation and sequencing. If pdo is 0 (default) then there is not dropout.

Number of "features" (e.g., genes) to simulate data for

Vector of mutation populations you want to simulate.

Fraction design matrix, specifying which potential mutational populations should actually exist. See ?EstimateFractions for more details.

Matrix of fractions of each mutational population to simulate. If not provided, this will be simulated. One row for each feature, one column for each mutational population, rows should sum to 1.

Vector of length = nfeatures; specifies the number of reads to be simulated for each feature. If this is not provided, the number of reads simulated is equal to round(seqdepth * (ksyn_i/kdeg_i)/sum(ksyn/kdeg)). In other words, the normalized steady-state abundance of a feature is multiplied by the total number of reads to be simulated and rounded to the nearest integer.

Character vector to assign to sample column of output simulated data table (the cB table).

Name given to the i-th feature is paste0(feature_prefix, i). Shows up in the feature column of the output simulated data table.

Vector of length = nfeatures; specifies the degradation rate constant to use for each feature's simulation. If this is not provided and fn_vect is, then kdeg_vect = -log(1 - fn_vect)/label_time. If both kdeg_vect and fn_vect are not provided, each feature's kdeg_vect value is drawn from a log-normal distrubition with meanlog = logkdeg_mean and sdlog = logkdeg_sd. kdeg_vect is actually only simulated in the case where read_vect is also not provided, as it will be used to simulate read counts as described above.

Vector of length = nfeatures; specifies the synthesis rate constant to use for each feature's simulation. If this is not provided, and read_vect is also not provided, then each feature's ksyn_vect value is drawn from a log-normal distribution with meanlog = logksyn_mean and sdlog = logksyn_sd. ksyn's do not need to be simulated if read_vect is provided, as they only influence read counts.

If necessary, meanlog of a log-normal distribution from which kdegs are simulated

If necessary, sdlog of a log-normal distribution from which kdegs are simulated

If necessary, meanlog of a log-normal distribution from which ksyns are simulated

If necessary, sdlog of a log-normal distribution from which ksyns are simulated

Vector of probabilities of mutation rates in labeled reads of each type denoted in populations. Should be a named vector, with names being the corresponding population.

Vector of probabilities of mutation rates in unlabeled reads of each type denoted in populations. Should be a named vector, with names being the corresponding population.

Only relevant if read_vect is not provided; in that case, this is the total number of reads to simulate.

Length of simulated reads. In this simple simulation, all reads are simulated as being exactly this length.

Minimum possible value of alpha element of Dirichlet random variable

Maximum possible value of alpha element of Dirichlet random variable

Probability that a nucleotide in a simulated read is a U.

Probability that a nucleotide in a simulated read is an A.

Probability that a nucleotide in a simulated read is a G.

Probability that a nucleotide in a simulated read is a C.

Number of "features" (e.g., genes) to simulate data for

Vector of length = nfeatures; specifies the number of reads to be simulated for each feature. If this is not provided, the number of reads simulated is equal to round(seqdepth * (ksyn_i/kdeg_i)/sum(ksyn/kdeg)). In other words, the normalized steady-state abundance of a feature is multiplied by the total number of reads to be simulated and rounded to the nearest integer.

Length of s^4^U feed to simulate.

Character vector to assign to sample column of output simulated data table (the cB table).

Name given to the i-th feature is paste0(feature_prefix, i). Shows up in the feature column of the output simulated data table.

Vector of length = nfeatures; specifies the fraction new to use for each feature's simulation. If this is not provided and kdeg_vect is, then fn_vect = 1 - exp(-kdeg_vect*label_time). If both fn_vect and kdeg_vect are not provided, then kdegs are simulated from a joint distribution as described below and converted to a fn_vect as when kdeg_vect is user-provided.

Vector of length = nfeatures; specifies the degradation rate constant to use for each feature's simulation. If this is not provided and fn_vect is, then kdeg_vect = -log(1 - fn_vect)/label_time. If both kdeg_vect and fn_vect are not provided, each feature's kdeg_vect value is drawn from a log-normal distrubition with meanlog = logkdeg_mean and sdlog = logkdeg_sd. kdeg_vect is actually only simulated in the case where read_vect is also not provided, as it will be used to simulate read counts as described above.

Vector of length = nfeatures; specifies the synthesis rate constant to use for each feature's simulation. If this is not provided, and read_vect is also not provided, then each feature's ksyn_vect value is drawn from a log-normal distribution with meanlog = logksyn_mean and sdlog = logksyn_sd. ksyn's do not need to be simulated if read_vect is provided, as they only influence read counts.

Probability that a T is mutated to a C if a read is new.

Probability that a T is mutated to a C if a read is old.

If necessary, meanlog of a log-normal distribution from which kdegs are simulated

If necessary, sdlog of a log-normal distribution from which kdegs are simulated

If necessary, meanlog of a log-normal distribution from which ksyns are simulated

If necessary, sdlog of a log-normal distribution from which ksyns are simulated

Only relevant if read_vect is not provided; in that case, this is the total number of reads to simulate.

Length of simulated reads. In this simple simulation, all reads are simulated as being exactly this length.

Probability that a nucleotide in a simulated read from a given feature is a U is drawn from a beta distribution with shape1 = Ucont_alpha.

Probability that a nucleotide in a simulated read from a given feature is a U is drawn from a beta distribution with shape2 = Ucont_beta.

Boolean; if TRUE, simulate a different pnew for each feature

Boolean; only relevant if feature_pnew is TRUE. If so, then setting pnew_kdeg_corr to TRUE will ensure that higher kdeg transcripts have a higher pnew.

If feature_pnew is TRUE, then the logit(pnew) for each feature will be drawn from a normal distribution with this mean.

If feature_pnew is TRUE, then the logit(pnew) for each feature will be drawn from a normal distribution with this standard deviation.

An object of class EZbakRArrowData

An object of class EZbakRData