keep read address using tophat

[splaisan]

Maybe overlooking something but ...
when I use tophat with paired reads having a name as

@SRR479052.1 HWI-ST188:1:1101:1222:2140/1

I end up with the second part clipped and a arbitrary read number put instead in the resulting bam as SRR479052.5415964 which does not support marking optical duplicates with picard

READ_NAME_REGEX=[a-zA-Z0-9]+:[0-9]:([0-9]+):([0-9]+):([0-9]+).*.

Q: Can I preserve the full read address when using tophat with some magic argument? or should I parse both fastQ and bam to reconstitute this missing info?

Thanks for help
Stephane

[dpryan]

That's because the second part isn't part of the read name. There's an option in fastq-dump to put the original read name where it should be rather than just numbering things sequentially.

[splaisan]

Problem is I downloaded the fastq pre-made from the EBI repo and mapped them all :-( without figuring this out. I can fix this by patching the fatsQ but will still need to remap the whole shebang...

Thanks for the info anyway (for next time)

[splaisan]

picard markDuplicate compatible reads from SRA data

few days later, the issue is fixed by:

• NOT downloading the fastq files from SRA but instead the .sra formatted data using Aspera (I used the browser link)
• Use the sratoolkit command fastq-dump (thanks Devon) to convert .sra to .fastq and split reads in paired files. The trick was here to use the specific parameter -F|--origfmt to ensure 'Defline contains only original sequence name' and that the remaining text was discarder

The resulting command in my case was (after correcting typo!):

fastq-dump -F --split-3 --gzip *.sra -O fastq_read_folder

TIP: I used P|P|S|S to speed this dramatically for the 26 input files on my 24 thread machine.

My reads have now a header line as

@HWI-ST188:1:1101:1222:2140
NAGACGAAGGTTCTTCAGTTAAACAGTTTAGAGCCCCATAAGAGCAAACTGTAGTGTAAAGAGGAAAAGTAAGTACAATCTTTCCAGACACACAACTAATA
+HWI-ST188:1:1101:1222:2140
#1:BDDDDHHHHHIIIIIIIIIIIIIHIIIIIIIIIIIIIIIIIIIIIIIIHGIFHCGIEHIIIHIIIIDEHHCHEHEEEEEECCECCCBCCBBBBCCCCA

which after tophat mapping results for that particular read in

HWI-ST188:1:1101:1222:2140 99 chr10 59953037 50 101M = 59953061 125 NAGACGAAGGTTCTTCAGTTAAACAGTTTAGAGCCCCATAAGAGCAAACTGTAGTGTAAAGAGGAAAAGTAAGTACAATCTTTCCAGACACACAACTAATA #1:BDDDDHHHHHIIIIIIIIIIIIIHIIIIIIIIIIIIIIIIIIIIIIIIHGIFHCGIEHIIIHIIIIDEHHCHEHEEEEEECCECCCBCCBBBBCCCCA AS:i:-1 XN:i:0 XM:i:1 XO:i:0 XG:i:0 NM:i:1 MD:Z:0C100 YT:Z:UU NH:i:1

Running picard on such BAM data is now able to identify few 1000' optical repeats in the full sample.

CQFD

[GenoMax]

Don't see a "-F" in your fastq-dump command above. Typo?

[splaisan]

shame on me! corrected now (thanks)