Normalize for experimental/bioinformatic dropout of labeled RNA.

Uses the strategy described here, and similar to that originally presented in Berg et al. 2024, to normalize for dropout. Normalizing for dropout means identifying a reference sample with low dropout, and estimating dropout in each sample relative to that sample.

Usage

NormalizeForDropout(
  obj,
  normalize_across_tls = FALSE,
  grouping_factors = NULL,
  features = NULL,
  populations = NULL,
  fraction_design = NULL,
  repeatID = NULL,
  exactMatch = TRUE,
  read_cutoff = 25
)

Arguments

obj

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

normalize_across_tls

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

grouping_factors

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.

features

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.

populations

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.

fraction_design

"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.

repeatID

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

exactMatch

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.

read_cutoff

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

Value

An EZbakRData object with the specified "fractions" table replaced with a dropout corrected table.

Details

NormalizeForDropout() has a number of unique advantages relative to CorrectDropout():

• NormalizeForDropout() doesn't require -label control data.

• NormalizeForDropout() compares an internally normalized quantity (fraction new) across samples, which has some advantages over the absolute dropout estimates derived from comparisons of normalized read counts in CorrectDropout().

• NormalizeForDropout() may be used to normalize half-life estimates across very different biological contexts (e.g., different cell types).

There are also some caveats to be aware of when using NormalizeForDropout():

• Be careful using NormalizeForDropout() when you have multiple different label times. Dropout normalization requires each sample be compared to a reference sample with the same label time. Thus, normalization will be performed separately for groups of samples with different label times. If the extent of dropout in the references with different label times is different, there will still be unaccounted for dropout biases between some of the samples.

• NormalizeForDropout() effectively assumes that there are no true global differences in turnover kinetics of RNA. If such differences actually exist (e.g., half-lives in one context are on average truly lower than those in another), NormalizeForDropout() risks normalizing away these real differences. This is similar to how statistical normalization strategies implemented in differential expression analysis software like DESeq2 assumes that there are no global differences in RNA levels.

By default, all samples with same label time are normalized with respect to a reference sample chosen from among them. If you want to further separate the groups of samples that are normalized together, specify the columns of your metadf by which you want to additionally group factors in the grouping_factors parameter. This behavior can be changed by setting normalize_across_tls to TRUE, which will

Examples

# Simulate data to analyze
simdata <- EZSimulate(30)

# Create EZbakR input
ezbdo <- EZbakRData(simdata$cB, simdata$metadf)

# Estimate Fractions
ezbdo <- EstimateFractions(ezbdo)
#> Estimating mutation rates
#> Summarizing data for feature(s) of interest
#> Averaging out the nucleotide counts for improved efficiency
#> Estimating fractions
#> Processing output

# Normalize for dropout
ezbdo <- NormalizeForDropout(ezbdo)