FASTQC: Quality Scores

[rlowe]

Hi,

I have some Illumina GAII paired end reads (bisulfite converted) taken from a previous study.

It seems that based on the FASTQC report that there are two sequence lengths 50BP and 76BP contained in my read.

I was just wondering how the quality scores are calculated as they drop off quite considerably after 50BP. E.g. is this a real drop off in qualities on the 76BP reads (and hence should be trimmed) or is it because it is taking the 50BP reads as having 0 quality above 50BP and hence producing an incorrect average quality score?

I attempted looking at the code but I don't know Java well enough to understand whats going on.

Many thanks for any help,
Rob

[fatakias]

Hi,
score computation and normalization for length of reads is surely implemented in FastQC in the version I have used with reads generated for AB's SOLiD sequencer.
sarosh

[arvid]

Only scores from reads long enough to span a base are including in the calculations for that base, no fake zeroes are included. Quality typically drops towards the end of the reads, so your results are probably normal.