STARsolo SmartSeq segfault (2.7.7a)

[vasiliosz]

Hi Alex,

I've used STAR for SmartSeq2 library alignments (and then separate read counting). Now trying out the STARsolo option. However, I'm getting segfault errors with any cells I try (always paired-end/two files per cell). From the log and output SAM-file I can't see that it even starts mapping, meaning a reference issue?

Same error for 2.7.6a and 2.7.7a -- both installed via bioconda. Trimmed or untrimmed reads doesn't matter. Re-built the genome index a couple of times too.

Any ideas where to start?

$ STAR --version
2.7.7a

$ STAR \
--genomeDir /wrk/resources/genomes/hg38-iGenome-ERCC/hg38-iGenome-ERCC_STAR_FULL_ANNOTATED_v2.7.7 \
--readFilesIn fq1.fastq fq2.fastq \
--soloType SmartSeq \
--soloStrand Unstranded \
--soloFeatures Gene \
--soloUMIdedup Exact \
--outFileNamePrefix out/

Jan 05 17:02:52 ..... started STAR run
Jan 05 17:02:52 ..... loading genome
Jan 05 17:03:02 ..... started mapping
Segmentation fault

Log-file and example cell (fq1 + fq2) attached. Reference is iGenomes hg38 build + ERCC reads with corresponding GTF file. No errors given in index building.

Log-file just ends with:

Finished loading the genome: Tue Jan  5 17:15:35 2021

Processing splice junctions database sjdbN=256213,   pGe.sjdbOverhang=69 
alignIntronMax=alignMatesGapMax=0, the max intron size will be approximately determined by (2^winBinNbits)*winAnchorDistNbins=589824
Loaded transcript database, nTr=87490
Loaded exon database, nEx=962563
Thread #0 end of input stream, nextChar=-1

STARsolo_SS2.zip

Our standard alignment settings for STAR, and the equivalent log section:

STAR \
--genomeDir $refdir \
--readFilesIn fq1.fastq fq2.fastq \
--outFileNamePrefix out-std/ \
--outFilterType BySJout \
--outFilterMultimapNmax 20 \
--alignSJoverhangMin 8 \
--alignSJDBoverhangMin 1 \
--outFilterMismatchNmax 999 \
--outFilterMismatchNoverLmax 0.04 \
--alignIntronMin 20 \
--alignIntronMax 1000000 \
--alignMatesGapMax 1000000 \
--outSAMstrandField intronMotif \
--outSAMtype BAM SortedByCoordinate \
--outSAMattributes NH HI NM MD \
--genomeLoad LoadAndRemove \
--limitBAMsortRAM 12000000000

[...]
Finished loading the genome: Tue Jan  5 17:20:19 2021

Processing splice junctions database sjdbN=256213,   pGe.sjdbOverhang=69 
To accommodate alignIntronMax=1000000 redefined winBinNbits=18
To accommodate alignIntronMax=1000000 and alignMatesGapMax=1000000, redefined winFlankNbins=4 and winAnchorDistNbins=8
Opening the file: out-std/_STARtmp//FilterBySJoutFiles.mate1.thread0 ... ok
Opening the file: out-std/_STARtmp//FilterBySJoutFiles.mate2.thread0 ... ok
Thread #0 end of input stream, nextChar=-1
BAM sorting: 365078 mapped reads
[...]

[alexdobin]

Hi Vasilios,

--soloType SmartSeq requires --readFilesManifest /path/to/manifest.tsv with multiple files listed (corresponding to different cells listed there). I need to throw an error if --readFilesManifest is not provided.

Cheers
Alex

[vasiliosz]

Thanks Alex,

Adding >=2 files through a manifest indeed works as expected. However, it is useful to run with a single cell (ie a single line in the manifest) - to simplify n+1 analysis and data structure as new data comes in. If I understand you correctly, this is not the intended use of the solo mode?

[alexdobin]

Hi Vasilios,

it's actually a bug, I will fix it in the next release.

Cheers
Alex

[alexdobin]

Hi Vasilios,

this issue was fixed in https://github.com/alexdobin/STAR/releases/tag/2.7.8a.

Cheers
Alex

[vasiliosz]

Thanks, can confirm this works for individual cells now using --readFilesManifest.
But still segfaults if provided directly as --readFilesIn $fq1 $fq2

[alexdobin]

Hi Vasilios,

thanks for checking!
If you do not use --readFilesManifest, you need to provide read group ID in --readFilesIn - these will be the "cell barcodes" for SmarSeq, using --outSAMattrRGline ID:sample1.
I will add a check for this so that it exits with error if RG IDs are not supplied.

Cheers
Alex