STAR version

[Ryan Thompson]

I've recently upgraded to STAR version 2.5.2a, and it informed me that the STAR genome index that I had been using (generated by STAR version 2.3.0) could no longer be used and I had to re-generate the index using the new STAR version. So I ran the same command that had worked previously with version 2.3.0:

STAR --runMode genomeGenerate --genomeDir ./hg38-star --genomeFastaFiles ./analysisSet/hg38.analysisSet.chroms/*.fa --sjdbGTFfile ./knownGene.gff --sjdbGTFfeatureExon exon \ --sjdbGTFtagExonParentTranscript Parent --sjdbOverhang 100 --runThreadN 8 --genomeChrBinNbits 12

which failed with the following output & error:

Jun 16 20:46:59 ..... started STAR run
Jun 16 20:46:59 ... starting to generate Genome files
Jun 16 20:50:04 ... starting to sort Suffix Array. This may take a long time...
Jun 16 20:50:32 ... sorting Suffix Array chunks and saving them to disk...
Jun 16 21:16:12 ... loading chunks from disk, packing SA...
Jun 16 21:28:57 ... finished generating suffix array
Jun 16 21:28:57 ... generating Suffix Array index
Jun 16 21:34:10 ... completed Suffix Array index
Jun 16 21:34:10 ..... processing annotations GTF
terminate called after throwing an instance of 'std::out_of_range'
  what():  vector::_M_range_check: __n (which is 0) >= this->size() (which is 0)

I'm confused as to what might be causing this since the exact same command worked fine previously. Can anyone help me debug this?

[Alexander Dobin]

Hi Ryan,

Please send me a few first lines of the GFF file.

With the new version of STAR, it's best to convert the GFF file into GTF - for instance, with Cufflinks' gffread tool:

$ gffread -T In.gff3 -o Out.gtf

This is more robust, and you can also use the transcriptomic output and read-per-gene counting.

Cheers

Alex