Conversion of old pileup file to more useful format

Question

[mpjuers]

I've been handed a collection of pileup files, which as far as I can tell, have been created using the deprecated pileup command in samtools via Galaxy's 'generate pileup' tool. These are not the 10-column files with the SNPs called, which I need. Is there a way of converting them back into a more useful format, such as VCF? My ultimate goal is to calculate Fst, using snpStats, SNP Pipeline, or some other toolkit.

I have tried sam2vcf.pl, however it returns a cryptic error message, perhaps owing to the old filetype:

    $  sam2vcf.pl -r reference.fasta < Galaxy7-\[P2_pileup\].tabular > test.vcf
        FIXME: what is this [N]?
        at /usr/bin/sam2vcf.pl line 41, <STDIN> line 1408.
            main::error("FIXME: what is this [N]?\x{a}") called at /usr/bin/sam2vcf.pl line 89
            main::iupac_to_gtype("G", "N") called at /usr/bin/sam2vcf.pl line 214
            main::do_pileup_to_vcf(HASH(0x6dc398)) called at /usr/bin/sam2vcf.pl line 32

The person who did the processing is apparently unavailable and I don't have access to the original raw data. Am I stuck until someone can rerun all the processing?

[Brian Bushnell]

I have a hard time considering VCF to be superior to anything... it's the worst text-based format I've encountered in bioinformatics.

[mpjuers]

Superior in that it appears to work for my purposes. The files I have seem to be unusable for calling SNPs. I'm more than a bit green at this, however.

Answer 1

[Fabio Marroni]

I suggest to try varscan which takes pileup as input and do SNP calling:

http://varscan.sourceforge.net/using-varscan.html

Another option (of which I am no expert, I just googled it) is to convert the pileup to vcf using the pileup2vcf script in galaxy

https://searchcode.com/codesearch/view/63202730/

[mpjuers]

I must have missed those. Thanks.

[mpjuers]

After a fruitless struggle with those tools, it appears that the person who handed this dataset off to me also used `filter pileup` and appended the extra 5 columns (number of base differences) to the original 6-column output. I am continuing to suspect there isn't much I can do with what has been given to me in its overprocessed state. I ran VarScan on a sample 10-column consensus file I had and it worked fine.