Menu
▾
▴
1 Attachments
#323 Mapping rate anomaly for fragments slightly greater than maxins
I have aligned the DREAM mutation challenge real tumour/normal data sets (4TB of fastqs: 10 * hg19 WGS 100x coverage, split by library = 113 * ~30Gb fastq pairs) using bowtie2 with --very-sensitive-local and there appears to be an error in the bowtie2 mapping rate for fragments of length 500bp (the default -X value).
As can be seen from the attached image (generated from Picard insert size metrics for each library), these alignment artefacts are consistently present in all samples (with a non-zero distribution > 500).
This artefact is not present when aligned with novoalign indicating a bowtie2 error and causes serious issues with some downstream analysis tools. Can this error be fixed, or as an interim solution, change the default -X option to 1000 so this error is not encountered for most Illumina PE samples?
The anomaly peak appears to be determined by the read length. On a completely different data set, 2x35 peak at 515, 2x75 peak at 535, and 2x100 peak at 549.