FastQ drop (illumina)

[strob]

Hello all,

when comparing the fastq files obtained from illumina/solexa sequencing, we see a huge drop in values. Normally, you would only see a small dip near the 3 prime end, while our current sequences show a dramatic drop when going to the 3 prime end and even their 5 prime end doesn't start at value ~40 (see att).

After contacting people at illumina, they claim that the quality of the reads is still good, but that due to their new pipeline, you can have a drop in the fastq values.

Some questions/remarks.

- Did other people already experience this drop in fastq value?
- How can we now standardize the quality of the reads when these values change over time?
- What about assembly tools that try to incorporate these values in order to perform a better assembly?
- How do you use/interpret these fastQ files?
- Should we recalculate these values as described in following paper: http://www.ncbi.nlm.nih.gov/pubmed/1...ubmed_RVDocSum


Many thanks

Steven

[Torst]

Illumina switched from "Solexa quality values" to "Phred quality values" since GAPipeline 1.3. They also changed the ASCII mapping. So your ASCII to "Q value" mapping could be misleading. See http://en.wikipedia.org/wiki/FASTQ_format

[Bacteria Genomes]

I am also getting a drop in my quality scores as the cycles go but mine get really bad at the end. I attached graphs of the quality scores distributions. I got two slightly different patterns for the two different bacteria genera that I sequenced. I don't know why there would be a difference with the different genera unless the DNA preps weren't the same.

I kind of don't know what to do with this information about the quality scores now though. At what point does the quality score mean the data is bad and shouldn't be used? Does anyone have a suggestion for how to trim the bad bases off?

Note: I am aware that something went wrong around cycle 39, and that is not what my question is referring to.