Segmentation fault using genome index built using Ensemble but not UCSC files

[Koen K]

Hi Alexander,

In the past I've used STAR using reference genome files from UCSC and it worked perfectly.

Now however I want to focus on splice isoforms, and the package I want to use for this (DEXSeq) uses the genome files from Ensembl. Therefore I want to realign the reads using the Ensembl files in order to avoid any problems further down the road. However now I'm having trouble aligning the reads with STAR, even though the genome indexing goes fine:

Sep 30 12:44:53 ..... started STAR run

Sep 30 12:44:53 ... starting to generate Genome files

Sep 30 12:45:32 ... starting to sort Suffix Array. This may take a long time...

Sep 30 12:45:41 ... sorting Suffix Array chunks and saving them to disk...

Sep 30 13:06:10 ... loading chunks from disk, packing SA...

Sep 30 13:12:24 ... finished generating suffix array

Sep 30 13:12:24 ... generating Suffix Array index

Sep 30 13:14:30 ... completed Suffix Array index

Sep 30 13:14:30 ..... processing annotations GTF

Sep 30 13:14:42 ..... inserting junctions into the genome indices

Sep 30 13:16:33 ... writing Genome to disk ...

Sep 30 13:17:00 ... writing Suffix Array to disk ...

Sep 30 13:20:59 ... writing SAindex to disk

Sep 30 13:21:16 ..... finished successfully

But for alignment:

Sep 30 14:03:21 ..... started STAR run

Sep 30 14:03:22 ..... loading genome

Sep 30 14:07:29 ..... started mapping

Segmentation fault

I don't think it is a memory issue because I'm still able to run succesful alignements using the UCSC-genearted genome index; this also makes me think something's wrong with the Ensembl index (or the files I used to make it). Therefore I've used two different sources (iGenome and GenCode) to download the .gtf and .fa files needed for the genome index, but no luck.

Also I ended up trimming some of the reads using Trimmomatic, just to be sure I also ran the untrimmed read files, but still the same result.

Any ideas what might be going wrong? Attached is a Log.out file from one of the alignment runs.

Many thanks!

Koen

[Alexander Dobin]

Hi Koen,

this is weird, there is nothing suspicious in the Log file.

Could you send me a small subset of reads that still seg-faults?

Cheers

Alex