FLASH use question

[yaximik]

I tried to use FLASH to join MiSeq paired end reads. The average library length was about 550 nt, so out of curiosity I pushed 2x251 reads. It did not go well and added a lot of low quality junk to reads. Anyway, when I run FLASH on the output, it joined only about 1.8% of reads, and most of reads were left unextended, which is likely not surprizing. But when I scratched the pumpkin a bit I got conflicting questions about FLASH use.

1. Do I need to clean reads to remove Ns and low quality parts/reads first?

2. If so, then I am going to violate one of FLASH requirements that paired reads must be in the same order in two files, as some reads will be obviously removed from both files in unlikely synchronous manner.

So how to address this problem?

[kammoji]

FLASH quality values

Hi yaximik,

Just to let you know, I am currently suspicious about the quality values associated with the FLASH overlapping reads. I overlapped some Illumina MiSeq reads and the qualities of overlaps were immensely low. I found a lot of "G" quality value in my overlaps which would indicate a phred score of 7 (p[err]≃0.19).

I am now wondering which scale is FLASH using when reporting qualities? Some older scale perhaps?

[JackieBadger]

this may help: http://seqanswers.com/forums/showthr...ighlight=FLASH

Basically you need to get you PE reads trimmed down for good alignment.
Also, if there are drastically different product sizes (ie you used different primers) you should de-multiplex these first.

There is a link to a blog which details a solution in the link provided.

[kmcarr]

Hi yaximik,

Just to let you know, I am currently suspicious about the quality values associated with the FLASH overlapping reads. I overlapped some Illumina MiSeq reads and the qualities of overlaps were immensely low. I found a lot of "G" quality value in my overlaps which would indicate a phred score of 7 (p[err]≃0.19).

I am now wondering which scale is FLASH using when reporting qualities? Some older scale perhaps?

No, FLASH is using the newer scale (well it is also 'older' in the sense that it is the scale originally used when the Sanger Center defined the FASTQ format). Long, long ago (in NGS time scale) Illumina used Phred+64 to encode its quality scores, but then a few years ago they switched to the more conventional Phred+33 scale. Using ASCII(Phred+33) scale 'G'==38.