This repository has code for reproducing a part of polyester_code. In particular figures in section 3.1.
The goal was to reproduce these figures, try another seed, and try to determine a measure that we can calculate by gene to see if there is information. Particularly, if there is more information than what we would expect by chance.
Folders NA06985, NA12144, NA12776, NA18858, NA20542, NA20772, and NA20815 are taken as available from polyester_code. Their output is reproduced (described below) for 20 genes instead of 10. The same is true for countmat.rda. sequences.rda was not used.
select_genes has the files for selecting the 20 genes to use. It selects the 10 that were used in polyester_code and another 10 genes with FPKM > 20 in the Geuvadis data. See select_genes.Rmd for the details.
geuvadis has the files for re-aligning the selected 7 samples from Geuvadis. subset_pop.Rmd subsets the info to just the 7 samples. Then run_tophat.sh is used for aligning the data in 4 different TopHat modes:
- G: using –G with the full annotation of hg19
- noG: without using –G
- GaT: using –G with the full annotation of hg19 and also using –T (so it doesn’t look beyond the known transcripts)
- incG: using –G with an incomplete annotation of hg19. It’s the full annotation minus the information for the 20 genes selected.
simulate_reads has the files for simulating the reads with polyester. See simulate_reads.Rmd for the details. Once the reads are simulated, they are aligned using run-paired-tophat.sh.
compare_countmats.Rmd compares the count matrices from the 4 TopHat modes.
coverage_plots has the files for making the coverage plots using derfinderPlot. coverage_plots.Rmd creates plots by gene (20), gene by sample (20 * 7), by exon (250), by exon by sample (250 * 7), and some other plots. The HTML files are not version controlled because they are about 350 MB in size.
To reproduce the actual figure instead of using derfinderPlot, see reproduce_fig.Rmd.
This step explores some measures such as correlation,