Small Indel Caller Of Choice

Question

[richardc.gsc]

Hi all, I'm wondering which tools people are using for Illumina whole genome indel calling. I'm specifically interested in finding indels that are shorter than a read length. While we use other tools for somatic (tumour-normal) calling, what I'm most interested in here are tools that work on just one sample at a time.

Currently we are using Samtools mpileup to generate the SNPs and INDELS all at once. What other tools should I consider?

[Ashutosh Pandey]

I have tried GATK, Pindel and Dindel. All of them are great.

[gsr9999]

I am trying to run scalpel 0.5.2, discovery-mode, on a simple bam file of ~300 MB size. The scalpel-discovery runs fine, but it ends up creating a variants.indel.vcf file with only headers and no content in it.

Here is the command that I used

$scalpel-discovery --single --bam bwa_aligned_sample.bam --bed MT:1-16569 --ref "/home/sga/refData/gatkBundle28/b37/human_g1k_v37.fasta"

I am wondering if @hanfang.cshl may have any insights?

Answer 1

[hanfang.cshl]

Speaking of INDEL callers, Scalpel was extensively benchmarked in this below paper, showing high accuracy (especially for large indels).

Accurate de novo and transmitted indel detection in exome-capture data using microassembly

Also, we highlighted the use of Scalpel to reduce errors in whole-genome and whole-exome studies in a follow-up paper.

Reducing INDEL calling errors in whole genome and exome sequencing data

Feel free to check out Scalpel - my go-to caller for indels :)

Scalpel is available here: http://scalpel.sourceforge.net/

Answer 2

[Charles Warden]

I typically use VarScan to call SNPs and small indels

I also noticed this algorithm was recently published, but I haven't really found small indels to be problem. Larger indels are tricker (for that, I've found CoNIFER to be the best option for exome-capture data, but I think there is still room for improvement).

http://biorxiv.org/content/early/2013/12/13/001370