feigeliudan01 / biostat_reply Goto Github PK

This code is provided to reproduce the analysis for table 1 for BIOSTS-15205. For our differential expression analysis, we utilize the metadata and expression data from GSE61901

• utils (we utilized R 3.2.3's built-in version)
• lme4 (we utilized version 1.1-10)
• limma (we utilized version 3.26.5)
• sva (we utilized version 3.18.0)

Run the script using

$ Rscript table1_code.R

or simply source('table1_code.R') within R to download the data from GEO and run and output the analysis

The following methods utilize quantile Normalization, no averaging of technical replicates, and utilize CHIP as blocking variable for technical replicates

• The ranova function utilizes a repeated-measures anova to take technical replicates into account as part of the differential expression analysis
• The lmem function utilizes a linear mixed-effects model to take technical replicates into account as part of the differential expression analysis
• The limma_dupcorr function utilizes LIMMA's duplicate correlation function to take technical replicates into account as part of the differential expression analysis

The following methods utilize quantile normalization, average technical replicates and apply ComBat including treatments as covariates

• The limma_noblocking function utilizes LIMMA for differential expression analysis
• The twowayanova function utilizes an ANOVA for differential expression analysis

The following methods utilize quantile normalization, average technical replicates and CHIP as blocking variable

• The limma_blocking function utilizes LIMMA for differential expression analysis. We note that we are not utilizing this function and instead are citing Nygaard et al's result directly for Table 1. Nygaard et al's analysis utilizes the labels from GSE40566 and the expression data from GSE61901.
• The twowayanova_blocking function utilizes ANOVA for differential expression analysis adjusting for CHIP