Suboptimal gap opening

[Gene Selkov]

I've been running into situations with pindel where it seemed to call short indels at wrong loci but in the general vicinity of the true variant. At a glance, they seemed to be ambiguous representations and I ignored them. Now I can see that they are actually incorrect. Why does it pick a deletion accompanied by four mismatches, while a perfectly-matching deletion of the same size is sitting next door? Does this algorithm pick the first solution that satisfies all criteria? How can I persuade it to do the right thing without the loss of sensitivity?

Example (the first alignment is by pindel, the second is a perfect match):

AATTTTTCCTataATAAAACAAATAATGTTAAAATGTTA
AATTTTTCCT---ATAATAAAACAAATGTTAAAATGTTA
AATTTTTCCTATAATAAAACA---AATGTTAAAATGTTA

Pindel version 0.2.5b9, 20160729, with default parameters

[Kai Ye]

Can you provide a test bam for me to look into this? I've been running into situations with pindel where it seemed to call short indels at wrong loci but in the general vicinity of the true variant. At a glance, they seemed to be ambiguous representations and I ignored them. Now I can see that they are actually incorrect. Why does it pick a deletion accompanied by four mismatches, while a perfectly-matching deletion of the same size is sitting next door? Does this algorithm pick the first solution that satisfies all criteria? How can I persuade it to do the right thing without the loss of sensitivity? Example (the first alignment is by pindel, the second is a perfect match): AATTTTTCCTataATAAAACAAATAATGTTAAAATGTTA AATTTTTCCT---ATAATAAAACAAATGTTAAAATGTTA AATTTTTCCTATAATAAAACA---AATGTTAAAATGTTA Pindel version 0.2.5b9, 20160729, with default parameters — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#66>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AB9s--WhMIVGILB5pbYDOfiuCzALQ_Cvks5sBvMagaJpZM4NzJBS>.

[Gene Selkov]

Mini-bams are attached. The case in point is 2:198264900, mapped to Homo_sapiens.GRCh37.75.dna.primary_assembly.fa Thank you very much! —Gene

…

On Jun 7, 2017, at 7:39 PM, Kai Ye ***@***.***> wrote: Can you provide a test bam for me to look into this? I've been running into situations with pindel where it seemed to call short indels at wrong loci but in the general vicinity of the true variant. At a glance, they seemed to be ambiguous representations and I ignored them. Now I can see that they are actually incorrect. Why does it pick a deletion accompanied by four mismatches, while a perfectly-matching deletion of the same size is sitting next door? Does this algorithm pick the first solution that satisfies all criteria? How can I persuade it to do the right thing without the loss of sensitivity? Example (the first alignment is by pindel, the second is a perfect match): AATTTTTCCTataATAAAACAAATAATGTTAAAATGTTA AATTTTTCCT---ATAATAAAACAAATGTTAAAATGTTA AATTTTTCCTATAATAAAACA---AATGTTAAAATGTTA Pindel version 0.2.5b9, 20160729, with default parameters — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#66>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AB9s--WhMIVGILB5pbYDOfiuCzALQ_Cvks5sBvMagaJpZM4NzJBS>. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#66 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAca6HGzU-4zJa7umHhoWyyEcTEdgiMUks5sB0KkgaJpZM4NzJBS>.