run_delboy: run_delboy

In alextkalinka/delboy: Differential-representation analysis by Elastic-net Logistic regression with BinOmial-thinning validitY tests

Description

Performs a differential-representation analysis using an elastic-net logistic regression approach for normalized count data that is split into two groups.

Usage

run_delboy(
  data,
  group_1,
  group_2,
  filter_cutoff,
  gene_column,
  batches = NULL,
  max.iter = 10,
  bcorr_data_validation = NULL,
  alpha = 0.5
)

Arguments

data	A data frame containing normalized count data for two different groups and their replicates. Can be a path to a file. Ideally the counts will already have been normalized to Transcripts Per Million (TPM), using, for example, the bias-aware quantification methods employed by salmon (Patro et al. 2017).
group_1	A character string naming the columns that belong to group 1.
group_2	A character string naming the columns that belong to group 2.
filter_cutoff	A numerical value indicating the cutoff below which (summed across all replicates) a gene will be removed from the data. For example, to keep only genes with more than 1 TPM on average across both groups, set the cutoff to 10 if there are 10 replicates in total.
gene_column	A character string naming the column containing gene names.
batches	A named character vector identifying the batch structure with names identifying sample columns in the data input. The length must equal length(group_1) + length(group_2). If NULL, there are no batches, or batches have already been corrected. Defaults to NULL. Batch correction will be conducted using sva::ComBat using non-parametric priors.
max.iter	An integer value indicating the maximum number of validation samples (default = 10). NULL indicates all sample combinations should be taken.
bcorr_data_validation	NULL if no batch (signal) corrected data is already available for validation. Otherwise, a data frame of treatment-corrected data should be supplied (to speed up validation, if already available). Defaults to NULL. Batch correction will be conducted using sva::ComBat using non-parametric priors.
alpha	The elastic-net regression penalty, between 0 and 1 (default = 0.5). If NULL, alpha is chosen automatically.

Value

An object of class delboy. Access this object using hits, plot.delboy, get_performance_stats, and get_deseq2_results.

Author(s)

Alex T. Kalinka, [e-mail]

References

• Kalinka, A. T. 2020. Improving the sensitivity of differential-expression analyses for under-powered RNA-seq experiments. bioRxiv 10.1101/2020.10.15.340737.

• Patro, R. et al. 2017. Salmon provides fast and bias-aware quantification of transcript expression. Nature Methods 14: 417-419.

See Also

hits, plot.delboy, get_performance_stats, get_deseq2_results

alextkalinka/delboy documentation built on Feb. 2, 2022, 4:19 p.m.