cufflink doesn't like sam from bwasw

[lukas1848]

Hi all,

I am doing an ab initio assembly of my transcriptome and I was 1x10^6 single end 454 reads. I aligned them with bwasw to my reference genome and now I wanted to feed the resulting (& sorted) sam file to cufflinks. I tried the same workflow on a small subset of my reads and there it worked like a charm, but now with the complete set, I get an error saying that some CIGAR ops have a length of zero.

Here's the complete output I get from cufflinks:

Warning: Your version of Cufflinks is not up-to-date. It is recommended that you upgrade to Cufflinks v1.3.0 to benefit from the most recent features and bug fixes (http://cufflinks.cbcb.umd.edu).
[bam_header_read] EOF marker is absent. The input is probably truncated.
[bam_header_read] invalid BAM binary header (this is not a BAM file).
File aln.bwasw.sam.sorted doesn't appear to be a valid BAM file, trying SAM...
[16:34:45] Inspecting reads and determining fragment length distribution.
SAM error on line 81315: CIGAR op has zero length
> Processed 61739 loci. [*************************] 100%
> Map Properties:
> Total Map Mass: 1355708.00
> Read Type: 0bp single-end
> Fragment Length Distribution: Truncated Gaussian (default)
> Default Mean: 200
> Default Std Dev: 80
Segmentation fault

The only thing I could come up with to explain this is that bwasw includes unmapped reads in the sam file which cufflinks can't deal with.
Yet I didn't find any quote confirming that cufflinks can't deal with unmapped reads in a sam file.

So, do I need to remove all unmapped reads from the alignment or am I completely on the wrong track?

cheers,
Lukas

[lukas1848]

The error seems to be caused by false CIGAR strings produced by bwa. Does anyone know whether there's a bugfix for that available?