[PICARD] Problems with SAM/BAM files without header

[apeltzer]

Hi there,

I have several problems when dealing with SAM/BAM files without an appropriate header inside. First of all, if there is a correct header the PicardTools SDK I use in my tool works without any flaws. If there is however no header, I had to change my tool to use the
VALIDATION_STRINGENCY=LENIENT setting, otherwise it would crash giving me the information that there is no header (so far so good and expected).

My problem now is: If I have a file without a header, but my reads have a corresponding header reference set (e.g. gi|040304024...) this is set to "*" by picard, without any necessity. Downstream my analysis, this is causing several issues in the pipeline. Can I force Picard to leave the already existing reference IDs set in each read as they are somehow?

Thanks,
Hope this is getting clear for anybody ;-)

[TiborNagy]

Header of SAM file is a very simple thing. Try to create manually.

[dpryan]

I'm not particularly familiar with the SAM JDK (I use the samtools C API mostly), but I can at least explain why this is happening. For both picard and samtools, alignments are stored internally in something like the BAM format, regardless of how they're written in a file. In a BAM file, the chromosome/contig name is absent, there's just a number (in the C API, this is read->core.tid, I don't know what the equivalent is in java). This number is just an index into an array stored in the header. At least in the C API, you can use an arbitrary header when reading a SAM/BAM file, since the reader function is given the header as an argument. Since java is an OO language, I wouldn't be surprised if there's no equivalently convenient way to do this in java. You might have to look at the code for SAMFileReader and see if you can manually change what it uses as a SAMFileHeader.