FastQC: odd kmer content

[zshuhua]

Hi,
I'm checking the data quality using fastqc for the Illumina Hi-Seq2000 reads and got odd kmer patterns (see attachment). Although an old thread suggested that 5' pattern is common, I want to know how to eliminate the odd patterns for GGGGG, AAAAA and TCTTC. Could you please help to interpret and get rid of them?
Thanks in advance!

[Jeremy]

TRIMMOMATIC, or any other read cleaning program, should clean those up. I would use options: SLIDINGWINDOW:4:15 HEADCROP:15 MINLEN:36
Then run FastQC again and you should see a much nicer report, if you are just mapping to a reference genome then cleaning reads is less of an issue so you could leave out the HEADCROP (which cuts bases from the start). But if you are assembling de-novo then I would cut that non-random sequence at the start.

[zshuhua]

Thanks for your help!
I intend to calculate genes/transcripts expression levels. Do I have to cut the 5' 10 bp?
I'm wondering what causes the patterns of for GGGGG, AAAAA and TCTTC, because I saw those patterns in other tens of RNA-seq data files.

[Jeremy]

Like I said, you only need to cut if doing a de-novo assembly because the non-randomness of the start of the reads could cause problems in assembly, if just aligning to a genome then the read should align anyway. The non-randomness of the start is usually caused by the 'random' primers used to make cDNA. The stretch of Gs or As are either bad quality reads or legitimate stretches of repeats that happened to get sequenced, either way the SLIDINGWINDOW option seems to remove them.