Pindel & pileup?

[dingxiaofan1]

Hi, right now I am trying to discover some indels from a next generation sequencing dataset by using both of the Pindel and SAMtools. While, potential indels identified from SAMtools seem to be not too much and a bit noisy when comparing with that from Pindel. However, Pindel only uses reads of which one end is not mapped. Thus Pindel won't find indels from reads both mapped. So any suggestions? Which one is better?

[dingxiaofan1]

no reply

[nilshomer]

no reply

Please be patient and wait at least a few days before bumping.

[epigen]

Pindel actually also takes reads where one end has a low mapping quality. We see that a lot with BWA mapped reads, where we mostly get 1 indel.
BWA already makes gapped alignments so you can locate indels from the alignments, e.g. with samtools pileup. It is highly recommendable to do local realignment before.
In contrast, Pindel finds indels de novo. There is also the Indel Genotyper and other tools. All seem to be rather complementary than "better". Indel detection is still early in evolution.

[KaiYe]

new Pindel version

There are two types of input reads Pindel will take:
1. one-end mapped, and mainly relate to structural variants
2. reads aligned with gap. Pindel will realign the reads to discover the events de novo.

A new version of Pindel is open for test. You can download the package at http://www.ebi.ac.uk/~kye/pindel/test/.

New features are
1. allow mismatches during string matching so that reads with sequence errors and SNPs can be used. indels and SVs near SNPs (covered by the same reads) can be detected reliably.
2. new variant types are implemented: tandem duplications and inversions as well as breakpoints of large insertions.
3. non-template sequence can be detected for deletions, tandem duplications and inversions.
4. In the test version, the maximum event size is set to about 8k. BreakDancer result can be incorporated to search for larger events.
5. scripts to convert bam/sam to Pindel input are provided.

Please first try the demo data. Any suggestions are highly appreciated.

Kai