Skip to content

Latest commit

 

History

History
51 lines (43 loc) · 2.38 KB

ercc.md

File metadata and controls

51 lines (43 loc) · 2.38 KB
Raw

ERCC transcript-enabled pipeline

October 27, 2022

Treeshop can now be run to process fastq files which contain ERCC transcripts. This path will run only the MD5sum, expression, and QC steps. It creates differently-named outputs so it can be run "on top" of an ERCC-unaware output directory.

How to run

Follow the Treeshop guidelines to create one or more virtual machines with fab up, fab configure. However, instead of running fab reference, run fab reference_ercc. This will download the ERCC-aware reference files.

Then, when you run fab process, include the flag ercc=True to indicate that the ERCC-aware path should be taken. For example:

fab process:manifest=manifest.tsv,base=/data/treehouse/allFiles,ercc=True

All samples listed in your manifest will be run under the ERCC-aware path; you cannot mix ERCC and non-ERCC in a single fab process.

Output files and directories

The ERCC-aware process's outputs are named differently from the standard pipeline so that they can coexist.

Under primary/derived/SAMPLE, the files:

sortedByCoord.md.ERCC.bam
sortedByCoord.md.ERCC.bam.bai

Under downstream/SAMPLE/secondary, the dirs:

md5sum-ERCC-3.7.0-ccba511
ucsc_cgl-rnaseq-cgl-pipeline-ERCC-3.3.4-785eee9
ucsctreehouse-bam-mend-qc-ERCC-v2.0.2-1c3c627

And the files in ucsctreehouse-bam-mend-qc-ERCC-v2.0.2-1c3c627 are:

bam_mend_qc.json
bam_mend_qc.tsv
methods.json
readDist.txt

(Other directories have the same file names as their non-ERCC counterparts.)

Known Problems

  • The Kallisto/abundance.tsv file has numbers 1,2,3 etc for its targets instead of the intended gene or transcript IDs.
  • The RSEM/Hugo/rsem_genes.hugo.results file is bad in some way that I have not investigated: it has 60,449 rows but should have 92 extra due to the ERCC transcripts. (Compare RSEM/rsem_genes.results, which has 60,591 rows instead of the standard 60,499). Use with caution.
  • When the fusions step was run, it crashed on every ERCC sample attempted and did not generate fusions.

How to use existing virtual machines

If you have already spun up one or more virtual machines using fab for a non-ERCC run, you can reuse those machines (once idle) for an ERCC run. To do this, run fab reference_ercc once to download the ERCC reference files onto the machines. Then, continue with fab process:ercc=True,manifest=...etc as above.