Variant calling from fasta files.

[dimo]

Greetings and salutations.

I am trying to analyze some old data which consists of several hundred fasta sequences from individual tumour samples which I believe came from some sort of Sanger sequencing. Is there tool that will align these to a genome (BLAST) and call nucleotide and amino acid variants (similar to a NGS analysis)?

Thanks in advance.

[vivek_]

You could likely use BWA-Mem to do the alignment and call variants using the GATK's UnifiedGenotyper module.

[dimo]

You could likely use BWA-Mem to do the alignment and call variants using the GATK's UnifiedGenotyper module.

Interesting, any idea how GATK would handle a lack of quality scores? I would also basically have only 1 read per sample as well so no idea if it would even flag.

[vivek_]

Interesting, any idea how GATK would handle a lack of quality scores? I would also basically have only 1 read per sample as well so no idea if it would even flag.

If it complains about quality scores, you can just generate a file with some dummy scores. I'm not sure how it would handle one read per sample though. You can also try samtools for variant calling and modify the min. reads flags.

[Brian Bushnell]

You can use BBMap's reformat.sh to convert fasta to fastq, and assign whatever fake quality score you want (the default is Q30, though you can adjust that with the flag "qfake=X"). I like 30 because it ends up with quality scores that are all "?", which is appropriate.

Also, I feel like 30 is a good insurance against bad trimming - I would never, ever recommend trimming to Q30 or above (meaning bases with a claimed quality below 30 are discarded), since it is virtually always a bad idea if you have accurate quality scores, but some pipelines do that automatically. So, a default of Q30 seems reasonable.

[colindaven]

I've done this once using freebayes. It will happily call variants in VCF format from a BAM file with a coverage of 1. Took me a while to find this solution.