Suggested Pipeline Configuration 5.1.6
|Application|Function|Somatic|Germline|Notes | --- | --- | --- | --- | --- | --- | forthcoming | indel realignment|yes|yes| This is slow. 2-3x slower than the whole Pisces step. You can skip it if you don't need the precision. Gets rid of a lot of false positives if the upstream aligner tends to make a lot of judgment errors.|- |Stitcher|stitches reads|yes|yes|reconciles pairs of reads into single consensus reads,reduces FP due to noise, reduces FN due to strand bias, only works with paired end sequencing| |Pisces|variant calling|yes|yes|use -ThreadByChr for faster performance (results variable at the moment) but greater memory hit. You will need plenty of memory to do this. use qsub on the cluster, works fine with UMIs or for catching variants < 0.01%, so long as the data quality supports this and Pisces is properly configured.| |Scylla|long range phasing|yes|yes|Special germline settings: ploidy=diploid, crushvcf=true, dist=5;must load the same bam that pisces does. Ie, if Pisces acts on a stitched or re-aligned bam, then so should Scylla,Turn off upstream MNV calling by Pisces. | |Q Score Recalibration|Downgrade Q score for poor samples|yes|no|germline calling is not so sensitive to sample degradation anyway. Gets rid of a lot of false positives if the samples has a lot of FFPE or oxidative damage that otherwise would dominate the variant calls.|