bissnp speed stats

[Albert Vilella]

Dears,

This is just an email to report some performance metrics from BisSNP runs we are doing.

We have a human hg19 reference bam of about 200Gbytes in size, produced by bwameth alignment (https://github.com/avilella/bwa-meth), and

we are running bissnp on 32 cores (-nt 32) and 100Gb of RAM,

bissnp is running with the following options:

 export MALLOC_ARENA_MAX=1 ; java -XX:+UseG1GC -XX:ParallelGCThreads=2 -Xmx99g -jar /home/user/BisSNP-0.82.2.jar  -R /opt/hs37d5-bwameth/hs37d5.fa  -I /opt/SAMPLE.bwameth.MS05.MP00.CP00.mapped.sorted.combo.bam  -T BisulfiteGenotyper  --trim_5_end_bp 0  --trim_3_end_bp 0  -vfn1 /opt/SAMPLE.bwameth.MS05.MP00.CP00.mapped.sorted.combo.bissnp.met.vcf -vfn2 /opt/SAMPLE.bwameth.MS05.MP00.CP00.mapped.sorted.combo.bissnp.var.vcf   -mbq 12  -minConv 0  -toCoverage 1000  -mmq 60  -nt 32  -mm40 5 

the speeds we are currently seeing are 1Mb of variant calls every 60 seconds, which for a 3Gb reference would mean 3000 minutes of computation, or 50 wallclock hours.

We are happy with the performance of the system, we would also be happy to hear any comments from anyone else currently using BisSNP.

Cheers,

Albert Vilella.