Decreased number of isoform when comparing multiple samples

[brookslabucsc]

Decreased number of isoform when comparing multiple samples

[belgravia]

Hi,

Thanks for the details. There's a bug here somewhere...

1. Did you get any error messages?
2. Can you run with the option --keep_intermediate so you can see at which step the failure is occurring? You may also want to specify the directory of intermediate files (--temp_dir) to somewhere more convenient for you to look at. It would be nice to know if there are empty/badly formatted files early on in the pipeline that are responsible for the failure.

-Alison

[martynakgajos]

1. I didn't get any failure message.
2. In general, I do get results, however many isoforms detected for a given sample are absent in the results if I analyze together multiple samples.

[belgravia]

Ok, it's possible what's happening is that with additional data, you are getting a greater amount of firstpass isoforms (you can verify this by running flair-collapse with --keep_intermediate and you can specify a --temp_dir if you like), which would mean more ambiguity when assigning reads to isoforms, which causes fewer reads to be aligning with mapping quality>0, which in turn results in fewer isoforms that pass the coverage threshold of 3.

You could try running collapse with --quality 0 to allow the primary alignments of reads that align equally well to multiple isoforms to be counted. This could potentially add more noise to the dataset, but I expect it should prevent the problem of more reads -> fewer final isoforms.

-Alison