Assessment of alternative pipelines for cross-platform microarray data integration using RNA-seq data

To reproduce the results you will need the following structure of directories:

article-microarrays
│   README.md
└───allsamples_exprs
└───combined
└───combined_exprs
└───exprs
└───general
└───pdata
└───plots
└───preprocessed
└───raws
|   └───GSE19615
|   └───GSE37614
|   └───GSE58644
|   └───GSE60785
|   └───GSE65194
└───rnaseq
└───RNA-seq
└───scores
└───scripts

Not all the scripts are meant to be launched in one go, read each script before executing. If you have troubles installing Brainarray CDFs manually, use install.brainarray.R script.

RNA-seq data

Use scripts in RNA-seq/ folder in order to download and assemble the data (change scripts in case you use some task manager etc), then use rnaseq_assemble.R to combine and average across samples. Alternatively, use already processed data in rnaseq/ folder.

Microarray data

Download raw Affymetrix and non-normalized Illumina data into corresponding subfolders in raw folder. Use preprocessing_affymetrix.R and preprocessing_illumina.R scripts to pre-process and normalize the data, instead use zipped datasets in preprocessed folder.

Use generate_common_genes_list.R to generate common genes list needed in most of the scripts.

Use combine.R script to generate combined microarray datasets, which are needed for futher analysis. This script also generates intermediate files, needed for arrays_comparison.R.

If you wish to test tranformation and normalization methods for Illumina raw bead-level data, use raw_illumina_preprocessing.R script and raw/047_20150203_Tchou_CAFs.zip file, which you should put into raw/GSE37614 folder. Alternatively, you can create plots using already generated files for this dataset in folder exprs/. See the script for details.

For array versus array comparison use arrays_comparison.R script, for studying many-to-one probesets groups variation use variation_analysis.R

Pipelines

Use brainarray.R and max_mean_random_scores.R to process datasets accroding to each pipeline and average across samples. Alternatively, use files in exprs/ folder instead.

Use pipelines_analysis.R to generate plots for datasets processed with different pipelines comparison, and combined_data_analysis.R to compared datasets, combined within each pipeline.

Feel free to create issues if you have troubles executing the scripts or installing libraries used in the scripts.