Paired-end 10x (2x151 bp) segfaults with mouse but not with human index?

apredeus commented 1 month ago

Hi Alex,

I was wondering if you could help me debug a rather unusual issue I've encountered trying to map a 5' paired-end mouse dataset. Basically, I'm using the same STAR version (2.7.10a_alpha_220818 compiled with --msse4.2 option) to process many thousands of samples already, and usually have very few issues I couldn't fix or which weren't related to cluster issues. I've also processed many paired-end 5' datasets and they work OK too; they are most often human though.

This data is mouse (exact re-creation of Cell Ranger mouse 2020-A, as described here, and no matter what I tried, it segfaults with a non-specific error. What makes it so weird is:

I first erroneously mapped it using human reference (which maps about 10% of reads), and it doesn't segfault!
I tried using gzipped and inflated reads
I tried using static 2.7.11a binary
I tried re-making the reference index with exact same version of STAR as used for mapping.

here's the full log, although it's very non-specific: Log.out.txt

I've attached a test set of 150k reads Could you please take a look if it works for you, or if you could see anything off about this data?

Thank you so much in advance!

-- Alex

apredeus commented 1 month ago

Oh yeah, and there's definitely enough RAM (128 Gb). The command segfaults a second after it says "started mapping.." in the log.

apredeus commented 1 month ago

After much experimentation, I've narrowed it down to having "Velocyto" option in the output. I've removed it and now everything runs.

Any ideas as to why this might be happening?

alexdobin / STAR

Paired-end 10x (2x151 bp) segfaults with mouse but not with human index? #2139