perform_anova_gen: A Function Performs Differential Expression Analysis Based On...

In fchuffar/epimedtools: Useful Tools Allowing Multi-Omics Analysis

A Function Performs Differential Expression Analysis Based On ANOVA

Description

This function peforms an ANOVA for each probe/gene (lines) across confditions (colums) of a data matrix. It returns a dataframe that includes many metrics as a results (betas, p-values, ...). Genes are indexed by 'g' in 'G', conditions are indexed by 'h' in 'H' (histology, sex...) and samples are indexed by 'i' in 'I'. For each gene 'g':

• we compute a linear model of the level of expression value according to histological groups: 'log2(exprs_g, i) = beta_g, h + epsilon_g, i'.

• we perform ANOVA test on this model and harvest corresponding p-value

• we extract of each group: 'beta_g, h = fracsum_g in G mean(log2(exprs_g,i))|G| - mean(log2(exprs_g,i))'

• we compute corresponding logratio scores: 'logratio_g, h = (1 + frac1/|H| - 1) beta_g, h'

• we compute corresponding foldchange scores using 'gtools::foldchange2logratio' function

These values are retruned as a dataframe which fields are:

• probename: the name of the corresponding probe

• ad_pval: the result of the Anderson-Darling normality test (hypothesis for ANOVA)

• intercept: the mean of means of each group

• beta_hhh_FFF: the previously describe beta value for the factor 'FFF' of the experimental grouping field 'hhh'

• logratio_hhh_FFF : the previously describe logratio value for the factor 'FFF' of the experimental grouping field 'hhh'

• foldchange_hhh_FFF: the previously describe foldchange value for the factor 'FFF' of the experimental grouping field 'hhh'

• pval_hhh: the p-value associated to the ANOVA test

• adj_pval_hhh: the adjusted p-value associated to the ANOVA test using Benjamini-Hochberg procedure to the threshold of 0.05.

Usage

perform_anova_gen(design, model, data, key = NULL,
  MONITORED_APPLY = FALSE, AD_TEST = TRUE)

Arguments

design	A dataframe that describes the design of the experiment.
model	A model describe the test to apply.
data	A matrix of exrpression values for probe/gene (lines) and confditions (colums).
key	A character string that will suffix the column names of the resulting data frame
MONITORED_APPLY	A boolean set to TRUE if we want to monitor the loop on lines (usefull for debugging).
AD_TEST	A boolean set to TRUE if we want to perform normality test.

fchuffar/epimedtools documentation built on Feb. 3, 2024, 2:21 a.m.