CNV and INDEL calling

[Nilaksha]

Hello,

I have been able to successfully develop a pipeline for SNP calling from exome sequence data and keen to look for indels and CNVs as well. I have heard that dindel is good for indel calling so I'm giving it a try. What are the options available for CNV calling? Any alternative for Indel calling is also appreciated

Thanks in advance.

[Brian Bushnell]

The most important factor when calling indels (particularly long ones) is the aligner, which places the reads and generates the cigar strings. And the best aligner for calling indels is BBMap, which marks them correctly on cigar strings in the first place, so you can simply call them from the pileup rather than laboriously realigning or doing local de-novo reassembly.

P.S. You should never, ever call SNPs and indels separately from each other; they must be called at the same time, in the same way, by the same tool, from the same data. Otherwise, locations that contain indels could alternatively be mapped to as though they contain SNPs rather than indels, or vice versa, and you can get contradictory reports of, say, 2 SNPs and 2 insertions at the same location, or one tool call heterozygous SNP at a location and another call heterozygous DEL at the same location, even though more than 50% of the reads indicate REF at that location. The tool must be aware of all things going on at a given location simultaneously to even have a chance of generating trustworthy output. Ideally, the same tool would be aware of CNVs, too, and yield calls of variable zygosity, but that's asking a bit much.

[Nilaksha]

The most important factor when calling indels (particularly long ones) is the aligner, which places the reads and generates the cigar strings. And the best aligner for calling indels is BBMap, which marks them correctly on cigar strings in the first place, so you can simply call them from the pileup rather than laboriously realigning or doing local de-novo reassembly.

P.S. You should never, ever call SNPs and indels separately from each other; they must be called at the same time, in the same way, by the same tool, from the same data. Otherwise, locations that contain indels could alternatively be mapped to as though they contain SNPs rather than indels, or vice versa, and you can get contradictory reports of, say, 2 SNPs and 2 insertions at the same location, or one tool call heterozygous SNP at a location and another call heterozygous DEL at the same location, even though more than 50% of the reads indicate REF at that location. The tool must be aware of all things going on at a given location simultaneously to even have a chance of generating trustworthy output. Ideally, the same tool would be aware of CNVs, too, and yield calls of variable zygosity, but that's asking a bit much.

Thanks for the valuable info. Much appreciated.