haplotypecaller and recalibration relations when workin on rough reference genome

pawel_osipowskipawel_osipowski Warsaw, PolandPosts: 8Member


Would it matter a lot to do recalibration before usage of hc? If yes, it's worth to do realignment before recalibration to reduce no of snv's, I guess. I'm searching for variants in one of not too much known genomes. What I've got is 30x coverage and haploids I work on. I did realignment and recalibration but I'm thinking if it helps me in anything while I don't know any obvious snp sites. And if it's not better to use haplotypecaller stright away after mapping and then do the recalibration with a bunch of best quality variants?

Please, help in that matter.


Best Answers


  • Geraldine_VdAuweraGeraldine_VdAuwera Posts: 9,971Administrator, Dev admin

    Hi Paul,

    For realignment, you don't need to have a set of known variants. The program is able to identify regions that need to be realigned and do so without known variants. For recalibration, known variants are required, so if you don't have a set available you can generate one yourself from your data. You'll need to do a first round of calling (without recalibration), choose the highest-confidence variants, then try recalibrating with those. You may need to do several cycles to refine your set of known variants, for best results.Good luck!

    Geraldine Van der Auwera, PhD

  • pawel_osipowskipawel_osipowski Warsaw, PolandPosts: 8Member

    Hi Geraldine,

    Thank you for your feedback. I like your software more and more! But I've got a feeling that you didn't understand me correctly. I assume my written english is far from clear. I'll try to put it in other words. Is it sensible to use haplotype caller before recalibration opposed to standard pipeline. Off course I could 'traditionally' use realigner on my raw data, than call my SNVs, do the recab (with highest-confidence data) and perform realignment. But I'm kind of curious if instead of calling variants traditionally and picking up the highest-confidence SNVs I could call them by hyplotype caller stright away on raw data (and use it twice per pipeline this way) as it's better in calling highest-confidence SNVs than anything else?

  • pawel_osipowskipawel_osipowski Warsaw, PolandPosts: 8Member

    Yes, I'm thinking of skipping realignment and recalibration. And using haplotype caller to get highest-confidence data to do recalibration. Forget about "use it twice per pipeline this way" :)

  • pawel_osipowskipawel_osipowski Warsaw, PolandPosts: 8Member

    I've had second scenario on my mind - thanks a lot! Sorry for imprecise writing from the beginning. We could deal with that quicker. I'm taking it into account.

  • pawel_osipowskipawel_osipowski Warsaw, PolandPosts: 8Member

    Dear Geraldine,

    Could you just explain me what do I need the second step (realign using high-conf variants) for?


Sign In or Register to comment.