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

[apredeus]

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]

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

[apredeus]

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?