Bam files generated with STAR cause a segmentation fault core dump error when used with another tool

Question

[bkffadia]

I am mapping RNA-Seq data using STAR, using multi-sample two-pass mapping. I first mapped all samples with one-pass then concatenated their SJOut files and filtered junctions. I launched the second mapping by using this SJOut file.

I used this command to generate genome : `

/home/STAR-2.7.10b/bin/Linux_x86_64/STAR \ 
--runThreadN 10 \ 
--runMode genomeGenerate \ 
--genomeDir /home/genomeDir3 \ 
--genomeFastaFiles /home/Homo_sapiens.GRCh38.dna.primary_assembly.fa \ 
--sjdbGTFfile /home/Homo_sapiens.GRCh38.110.gtf \ 
--sjdbOverhang 100 \ 
--sjdbGTFfeatureExon exon \ 
--genomeSAindexNbases 14 \ 
--limitGenomeGenerateRAM 51200000000

`

then used this command to map each sample : `

/home/lebechekhadidja/STAR-2.7.10b/bin/Linux_x86_64/STAR \ 
--genomeDir /home/genomeDir3 \ 
--readFilesIn "/home/Brain2/$file_name".fastq.gz \ 
--sjdbGTFfile /home/Homo_sapiens.GRCh38.110.gtf \ 
--sjdbFileChrStartEnd "/home/Human.Brain.All.SJ.out.RedunChrReadsNonCanonFilter" \ 
--readFilesCommand zcat \ 
--runThreadN 10 \ 
--genomeLoad NoSharedMemory \ 
--readFilesCommand zcat \ 
--outSAMtype BAM SortedByCoordinate \ 
--twopassMode None \ 
--quantMode TranscriptomeSAM GeneCounts \ 
--quantTranscriptomeBan Singleend \ 
--outSAMattrIHstart 1 \ 
--outSAMattributes NH HI AS nM NM MD jM jI MC ch XS \ 
--outSAMprimaryFlag OneBestScore \ 
--outSAMmapqUnique 60 \ 
--outSAMunmapped Within \ 
--outBAMsortingThreadN 10 \ 
--outBAMsortingBinsN 50 \ 
--winAnchorMultimapNmax 50 \ 
--limitBAMsortRAM 51200000000

` when I use Multimapper resolution tool to resolve multimapped reads I get this erro message :

Parsing segment boundaries from annotation file: /home/Homo_sapiens.GRCh38.110.gtf Segmentation fault (core dumped)

The thing is that when I mapped samples using STAR in galaxy platform, using same parameters. I didn't encounter the same error. so I tried to use command line in galaxy + STAR version in galaxy (this is why I used STAR-2.7.10b) but I keep getting the same error.

these are command lines for genome generation step and mapping step in galaxy :

when I ran MMR debgging with gdb I got this error : `

[New Thread 0x7ffff7a59700 (LWP 3020)] 
[New Thread 0x7ffff7258700 (LWP 3021)] 
[New Thread 0x7ffff6a57700 (LWP 3022)] 
[New Thread 0x7ffff6256700 (LWP 3023)] 
[New Thread 0x7ffff5a55700 (LWP 3024)] 
[New Thread 0x7ffff5254700 (LWP 3025)] 
[New Thread 0x7ffff4a53700 (LWP 3026)] 
[New Thread 0x7ffff4252700 (LWP 3027)] 
[New Thread 0x7ffff3a51700 (LWP 3028)] 
[New Thread 0x7ffff3250700 (LWP 3029)] 
[New Thread 0x7ffff2a4f700 (LWP 3030)] 
[New Thread 0x7ffff224e700 (LWP 3031)] 
[New Thread 0x7ffff1a4d700 (LWP 3032)] 
[Detaching after vfork from child process 3033] 
------------ tool progres messages -------------------- 
Thread 5 "mmr" received signal SIGSEGV, Segmentation fault. [Switching to Thread 0x7ffff6256700 (LWP 3023)] 0x0000555555560569 in Alignment::fill_coverage_vector (this=0x555570139400, cov_keep=std::vector of length 0, capacity 0) at /usr/include/c++/9/bits/stl_bvector.h:237 237         { return reference(_M_p, 1UL << _M_offset); }
`

I tried various STAR versions, experimented with various options and still the problem persists.

[bkffadia]

[GenoMax]

received signal SIGSEGV

That indicates an invalid memory location error. Are you assigning enough RAM to the Multimapper tool?

Have you checked the BAM to make sure it is not corrupt. In general not using STAR to sort the BAM is the better option. That operation can be done efficiently with samtools after the alignment.

[bkffadia]

hello GenoMax, thank you for your help.

I have 128Go RAM. when using MMR tool with both files (the one from galaxy and the one I mapped on vm) they have the same size, same content, and I use the same command line for both, the only difference is the input file. So this excludes the idea of insufficient memory because MMR works with the one from galaxy and outputs segfault with the one I mapped on the VM.

I thought that sorting file with STAR may be the cause, so I set --outSAMtype to SAM and launched the tool with the sam file and still got the same error.

I changed star version, used the latest, the same in galaxy and installed it using sudo apt-get rna-star and the error perisits.

I checked the bam file that I mapped on VM and compared it with the same from galaxy and found exactly same content except that the one I mapped on VM doesn't contain unmapped reads, which I think won't be the problem.

these are tools versions : Galaxy : STAR : 2.7.10b - Samtools : 2.0.4 VM : STAR : 2.7.11a , 2.7.10b - Samtools : 1.9