PAC_deseq: A wrapper to DESeq2 that apply differential expression...

In Danis102/seqpac: Seqpac: A Framework for smallRNA analysis in R using Sequence-Based Counts

A wrapper to DESeq2 that apply differential expression analysis on a PAC object

Description

PAC_deseq DESeq2 analysis on PAC_object.

Usage

PAC_deseq(
  PAC,
  model,
  deseq_norm = FALSE,
  test = "Wald",
  fitType = "local",
  threads = 1,
  pheno_target = NULL
)

Arguments

PAC	PAC object containing a Pheno data.frame with samples as row names, an Anno data.frame with sequences as row names and a Counts table with raw counts. The Counts table must have the sample names as column names and sequences as row names.
model	Character vector describing the statistical model based on column names in Pheno.
deseq_norm	Logical whether to return deseq normalized values or raw counts (default=FALSE).
test	Character parsed directly to DESeq that controls what type of statistical test that should be used. Alternatives are either "Wald" (Wald significance test) or "LTR" (likelihood ratio test/chi-squared test). See DESeq for more details. (Default="Wald")
fitType	Character parsed directly to DESeq that controls what type of dispersion fit that should be used. Alternatives are either "parametric" (dispersion-mean relation), "local" (local regression of log dispersions), "mean" (mean of gene-wise dispersion). See DESeq for more details. (Default="local")
threads	Integer number of threads to run in parallel.
pheno_target	(optional) List with: 1st object being a character indicating the main target column in Pheno. 2nd object being a character vector of the target group(s) in the target Pheno column (1st object). Important: In PAC_deseq, pheno_target controls the main comparative factor category from which a summarized table and plots will be generated. If, for instance, a target column named "groups" in pheno(PAC) contains "control" and "treatment" categories, setting pheno_target=list("groups", c("treatment", "controls") ensures that "treatment" is presented 1st in the factor levels, making for example log2FC appear as "treatment vs control". As default, pheno_target=NULL will result in the factor levels being automatically sorted in ascending order, which in the example above would result in control vs treatment. If no pheno_target is given, the first feature in the model will also be the main comparison presented in the graphs and summary tables.

Details

Given a PAC object this function will apply a differential expression analysis using DESeq2.

Value

A list of objects:

1st object - Summarized result table merged with anno(PAC)

2nd object - Target graphs (p-val distribution and volcano)

3rd object - All output from DESeq

See Also

https://github.com/Danis102 for updates on the current package.

Other PAC analysis: PAC_covplot(), PAC_filter(), PAC_filtsep(), PAC_gtf(), PAC_jitter(), PAC_mapper(), PAC_nbias(), PAC_norm(), PAC_pca(), PAC_pie(), PAC_saturation(), PAC_sizedist(), PAC_stackbar(), PAC_summary(), PAC_trna(), as.PAC(), filtsep_bin(), map_rangetype(), tRNA_class()

Examples

# Note, these examples will generate some warnings since data is based on
# heavily down-sampled fastq files, where many sequences receives low counts
# in specific groups.

## Load test data
load(system.file("extdata", "drosophila_sRNA_pac_filt_anno.Rdata", 
                package = "seqpac", mustWork = TRUE))

## Simple model with embryonic stages using Wald test with local fit (default)
table(pheno(pac)$stage)
output_deseq <- suppressWarnings(PAC_deseq(pac, model= ~stage, threads=2))

## Batch corrected, graphs are generated for 'stage' (=first in the model)
output_deseq <- suppressWarnings(PAC_deseq(pac, model= ~stage + batch,
                                          threads=2))

## Using pheno_target 
output_deseq <- suppressWarnings(PAC_deseq(pac,model= ~stage + batch, 
                                          pheno_target=list("batch"),
                                          threads=2))

## With pheno_target we can change the direction for the comparison
# Stage5 vs Stage3 (reverse order):
output_deseq <- suppressWarnings(PAC_deseq(pac, model= ~stage + batch, 
                         pheno_target = list("stage", c("Stage5", "Stage3")),
                         threads=2))  

## In the output you find PAC merged results, target plots and output_deseq   
names(output_deseq)
tibble::as_tibble(output_deseq$result)

Danis102/seqpac documentation built on Aug. 26, 2023, 10:15 a.m.