SOS, Question about FastQC report!

[lovenlong]

Hi all,
I'm newcomer to NGS field, and I would be grateful if someone could take a quick look at the FASTQC result of my NGS data (rice genome, 100 bp Paired-end, ~ 30x coverage) generated by Hiseq2000.
For the most part of the FASTQC result seem normal, except unnormal distribution for the per sequence GC content, the per base GC content, the sequence Duplication Levels and the Kmer Content.
Could this be due to PCR bias? Or be contaminated?
If conduct quality filtering with Fastx_toolkit, how do I set parameters?

Thanks a lot,

[jgibbons1]

For the KMER and GC plots, things start to normalize around 10 bp you may want to trim the 1st 10 bp.

As a rule of thumb I usually do the following QC and THEN run my data through FastQC:
(1) Remove duplicate reads with Fastx_toolkit
(2) Remove low quality bases from read set (I usually use trim_galore with a Q30 (--quality 30) filter and retain a minimum of 75% read length (so for your data --length 75))
(3) An output option of trim_galore is to pipe directly into FastQC, so I do this, and see if my plots are still "off"
(4) If they are, I trim the entire read set by X bp (depending on the plots)

This should leave you with a nice quality data set regardless of your downstream analysis.

[lovenlong]

For the KMER and GC plots, things start to normalize around 10 bp you may want to trim the 1st 10 bp.

As a rule of thumb I usually do the following QC and THEN run my data through FastQC:
(1) Remove duplicate reads with Fastx_toolkit
(2) Remove low quality bases from read set (I usually use trim_galore with a Q30 (--quality 30) filter and retain a minimum of 75% read length (so for your data --length 75))
(3) An output option of trim_galore is to pipe directly into FastQC, so I do this, and see if my plots are still "off"
(4) If they are, I trim the entire read set by X bp (depending on the plots)

This should leave you with a nice quality data set regardless of your downstream analysis.

Thanks for your suggestions.
I've made reads filtering and trimming as you suggested, and the data looks nice now.