I am developing/refining a pipeline for exome data. According to reported articles there are mainly two callers I found out to be more acceptable among researchers: One is SAMTools' variant caller and GATK's UnifiedGenotyper. However, GATK's HaplotypeCaller is believed to be more accurate, specifically while calling in-dels. There are quite a number of instances where people are relying on some other callers. Now it is quite obvious that due to different working principles of algorithms, mostly common variants and a few unique variants usually come up from the same sample. What if one or more non-synonymous variant or frame-shift deletion fall among those unique variants called by either of the caller. Sanger sequencing is one way to confirm those unique variant for sure, but as a bioinformatician, what can be the most favorable approach to deal with such situations and which variant caller do you prefer?
Exactly. It depends on what you want to do with the data. If you only need a set of highly confident SNPs you can use the intersection of Samtools and GATK's results. But I've heard cases where the overlap was too small. So it will really depend on the organism, type of data, depth and quality of the data.
@ Rohan Chaubal and Inês Pires:
First of all thanks for your inputs. Yes, I understood your point. Basically I am working with a hard core genetics group where our primary aim is to find genes with mutation. In this regard choice of variant caller is very much crucial in our case indeed. That is the reason why I raised this question to get a feedback from experts. Thanks once again.
As per my personal experience, If you are looking for a SNVs/small InDels then bowtie alignment & samtools (when data quality is extremely good) would be the very good choice. If you are looking for SVs then bwa (aln/mem) and pindel pipeline produces very reliable results. CNVnator can be used for CNV analysis is recommended.
If you looking for some complete end2end pipeline then HugeSeq (http://hugeseq.hugolam.com/) would be blindly prefered.
Hi, I am a newbie in this field...still learning the terms ans tools.
Our group works on TB bacteria genome sequnced by Pacific Biosciences and for SNP calling, they use
Alignment: BWA
SNP/Indel Calls: VarScan
Google Brain Team recently released DeepVariant. We implemented a reproducible version that was submitted to NF-Core (https://lifebit.page.link/UW58).
We also made it available in Lifebit (https://lifebit.page.link/os1m) if you want to try it with example parameters.
In practice, DeepVariant first builds images based on the BAM file, then it uses a DeepLearning image recognition approach to obtain the variants and eventually it converts the output of the prediction in the standard VCF format.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools