[sschavan]

Hello Brian,

Thank you for the detailed explanation, I appreciate your response.

I was wondering if I can set BBSplit to allow (a) maximum one mismatch and (b) 0 mismatch. If I could set it in this way probably I could use it to separate Mouse and Human reads, at a high stringency level, to start with.

About the error in filterbyname.sh, the result output (OnlyHumReads_BBv4.sam) was an incomplete file with only two header lines, presumably its truncated output. I am re-running it with -da flag, I will update if that worked for me.

Thanks,
SSC

[Brian Bushnell]

BBSplit supports all the flags supported by BBMap, including these:

idfilter=0              Independant of minid; sets exact minimum identity 
                        allowed for alignments to be printed.  Range 0 to 1.
subfilter=-1            Ban alignments with more than this many substitutions.
insfilter=-1            Ban alignments with more than this many insertions.
delfilter=-1            Ban alignments with more than this many deletions.
indelfilter=-1          Ban alignments with more than this many indels.
editfilter=-1           Ban alignments with more than this many edits.
inslenfilter=-1         Ban alignments with an insertion longer than this.
dellenfilter=-1         Ban alignments with a deletion longer than this.

Those are postfiltering operations - they take place after alignment, after ambiguity was decided, etc. and just suppress alignments failing the filters.

You can alternately map with the "semiperfectmode" flag, which will only allow alignments in which there are no mismatches or indels (this is different from "perfectmode" in that semiperfectmode allows a read to go off the end of a contig; perfectmode requires all bases to match the reference). Or, you could use the flag "minid=0.99" which will require 99% identity. These latter options are faster than mapping at default sensitivity, then imposing a specific filter.

[sschavan]

Thank you Brian, this information certainly helps. I will try it out.

SSC