[emboss-dev] Bug reports and patches: BAM quality, SAM negative ISIZE

Peter Rice pmr at ebi.ac.uk
Mon Aug 2 11:42:00 EDT 2010


On 02/08/10 14:55, Peter C. wrote:

> In the funny BAM to Sanger FASTQ conversion, EMBOSS has used
> "]" which is ASCII 93, giving PHRED 93-33 = 60. i.e. 33 more than it
> should be. I suspected that the EMBOSS code for reading BAM files
> was wrongly applying a 33 offset to the quality scores. In BAM files
> the scores are simply encoded directly as uint8_t without any offset.

Thanks for spotting that. We will make a patch with that fix in.


> Looking at the SAM file, I guessed EMBOSS doesn't like a negative
> ISIZE field in the next record, EAS54_61:4:143:69:578,  .........
>
> Looking at the source code, currently EMBOSS is wrongly assuming
> an unsigned integer will be used. This is not true, the spec allows for
> a negative ISIZE. I replaced this code in ajax/core/ajseqread.c

Thanks for the fix. We will add that to the patch.

> A related question is why did this error condition not give any
> error message to stdout or stderr?

This appears to be a general issue with reading unknown and known 
formats. We will fix it so that error messages are turned on for this 
failure condition.

Many thanks for the bug reports - and the fixes!!

Peter R.


More information about the emboss-dev mailing list