Gene Expression Normalization Paper

Custom code for reproducing the results of the research paper by Roca, Gomes, Amorim & Scott-Fordsmand: "Variation-preserving normalization unveils blind spots in gene expression profiling".

Instructions

To reproduce the results of the paper, please

1. Install/Update R and Bioconductor. Results were obtained with R version 3.3.1 and Bioconductor version 3.3.

2. Install/Update required libraries: plotrix and RColorBrewer (R), and affy, drosgenome1.db, drosophila2.db, genefilter, and limma (Bioconductor).

3. Download the files of this repository into a local directory (or clone/fork this repository).

4. Download the raw microarray data of the E. crypticus dataset, and copy the *.txt files into the rawdata subdirectory. The dataset is available at NCBI/GEO, with GEO accessions GSE69746, GSE69792, GSE69793 and GSE69794.

5. Download the raw microarray data of the Golden Spike dataset, and copy the *.CEL files into the rawdata subdirectory. The dataset is available at the online version of the paper by Choe et al., Genome Biology 6, R16, 2005, in the additional data files 6 and 7.

6. Download the raw microarray data of the Platinum Spike dataset, and copy the *.cel files into the rawdata subdirectory. The dataset is available at NCBI/GEO, with accession GSE21344.

7. Execute the five R scripts analyze_gene_expression*.r from the local directory with the repository. For example,

    $ cd ~/gene-expr-norm-paper
    $ nice R CMD BATCH --quiet --vanilla analyze_gene_expression.r
    $ nice R CMD BATCH --quiet --vanilla analyze_gene_expression_golden_spike.r
    $ nice R CMD BATCH --quiet --vanilla analyze_gene_expression_golden_spike_all_probe.r
    $ nice R CMD BATCH --quiet --vanilla analyze_gene_expression_platinum_spike.r
    $ nice R CMD BATCH --quiet --vanilla analyze_gene_expression_platinum_spike_all_probe.r
   

Results

Execution of the scripts analyze_gene_expression*.r generates the following subdirectories:

• bc_variation between-condition variation (figures 2, S1)
• boxplot boxplots of expression levels (figure 1)
• data intermediate results
• diff_expr differential gene expression for the E. crypticus and synthetic datasets (figures 3-7, S2)
• diff_expr_golden_spike ROC curves for the Golden Spike dataset (figures 8, S4)
• diff_expr_golden_spike_all_probe ROC curves for the Golden Spike dataset without restricting probes (figure S3)
• diff_expr_platinum_spike ROC curves for the Platinum Spike dataset (figures 9, S6)
• diff_expr_platinum_spike_all_probe ROC curves for the Platinum Spike dataset without restricting probes (figure S5)
• pvalue identification of no-variation genes (movies S4-S6)
• stdvec convergence of standard-vector normalization (movies S1-S3)
• table tables of experimental conditions for the E. crypticus dataset (tables S1, S2)