piquelab / gxe_pipeline Goto Github PK

When running QuASAR, the QuASAR_pipeline.R script assumes that your controls are named "CTRL-" in the "Control" column, as it greps for "CTRL". If the controls are not named with "CTRL", the grep returns integer(0), which, when passed to the data.frame with the -operator to drop rows, results in an empty data frame.

Not only is this a bad assumption (a control name could simply be "Media"), but it departs from the normal standard of using the "Control.ID" column and grepping for "CO", which is a much better convention.

The first time QuASAR_prep.R is run for a project, the QuASAR_results/data/ directory has not been created. As a result, trying to save to that location throws an error.

If a group of fastq links are removed (for example, to remove a bad sequencing run), removal of the L1 replicate (e.g., DP6-HT58_L1_R1.fastq.gz) will effectively block that barcode from the alignment rules of the Makefile.

For example, if a barcode has replicates L1 through L6, where L1-L3 correspond to one sequencing run and L4-L6 correspond to a second, removal of the first sequencing run (L1-L3) will result in the removal of the *-L1-*.fastq.gz file, which will block the re-alignment of those files, as the Makefile rule requires the *-L1-*.fastq.gz file.

Currently the only solution is to rename links to ensure there is an L1, or to remove/rename the original fastqs themselves and rerun the util script.

Our HPCC updated their jobs submission system, and as a result all the scripts that submit jobs need to be updated.