Illumina quality scores

[ewilbanks]

Hi everyone,

I've got a couple questions about dealing with quality info from Illumina data. What's the general opinion on the following things I've heard:

1) After a "B" indicating a bad call, all downstream nucleotides can't be relied upon and should be thrown out. (http://brianknaus.com/software/srtoolbox/shortread.html)
2) The Illumina pipeline filtering throws out lots of useful data. You should work directly from the _qseq and not the filtered _sequence.txt files.

Thoughts?? Opinions?? Is quality filtering like what's described in 1) overkill if I'm using MAQ as a quality aware aligner downstream?

I'm eventually aligning (MAQ) all these to a reference genome for ChIP-seq on microbial samples. Our coverage is super deep so I haven't just been throwing away repeat reads, since those could be expected based on our depth of sequence coverage. I'm looking for good ways to clean up our data!

Thanks!!!!
Lizzy

[maubp]

About the Read Segment Quality Control Indicator (B thing), are you seeing B in the middle of the quality string, or just a reads which end with a run of BBBBB? See also:

http://seqanswers.com/forums/showpost.php?p=17491&postcount=3

http://news.open-bio.org/news/2010/04/illumina-q2-trim-fastq/

[drio]

@2: Most aligners come with the qseq to fastq converter encouraging passing data from qseq. If you have cpu cycles, it'd be nice to map from and after qseq and see what results you are getting. Once you have that you can take a look to various things. What's the difference on total number of alignments reads and MAPq distributions. Also, you can check the MAPq of the reads that were filtered out by the software. Let us know what you find out!

[ewilbanks]

Thanks!!

@maubp: those links were really helpful and cleared up my concerns! I had read an isoltated report and heard tell of the scary "B" but really didn't know what to make of it. Good thing I looked into it, because I was about to trim my Sanger-format FASTQ file that way, which would have been silly... Going back to re-trim from the Illumina format file!

@drio: good point! I'll try that and let you know how it goes. I guess I was just concerned about figuring out the best way to throw out the "bad quality" reads. I think armed with my new knowledge, I'll try some quality trimming and then run the test you suggest. Will post once the data is chewed