FastQC: 3' bias (weird peaks) in Kmer content graphics

[Ann7]

Hi All,

My samples were run on Illumina HiSeq2000 using paired-ended stranded RNA-Seq protocol and the output reads are 100bp. The quality of my data are not bad in general. But I got some weird peaks at 3' end in FastQC kmer content graphics (plz see the attached graphs) and most of peaks are "GAAGA". Since this is part of the Illumina adapter (GATCGGAAGAGCTCGTATG), I tried Cutadapt (Cutadapt -a GATCGGA inputfile outputfile) to trim all the adapters at 3' end. This did get rid of the peaks in some of my samples, but the peaks still retain in a few samples, and another kmer "GAGGA" poped up. So I'm sure whether the 3' bias were caused by adapter remnant.

I also has a question about the 5' end bias. I was told they were caused by random hexmars. But shouldn't the hexmars only appear at 5' end of Read1? Why Read2 has the same bias at 5' end?

I will use trimmomatic to trim and filter the data next. But I'm not sure whether I should trim off the 3' end of Read2 or not, as the average overlapping between Read1 and Read2 is only 21bp (the average size of the library was 300bp, but all the fragments larger than 200bp were selected, which means the average inserts were of 179bp).

I would appreciate it very much if anyone has a better clue of what happened at both ends of the reads in my data set. Many thanks!

Attached graphs:
1: kmer content graph of Read1 (raw data)
2: kmer content graph of Read2 (raw data)
3: kmer content graph of Read1 (after Cutadapt)
4: kmer content graph of Read2 (after Cutadapt)
5: quality graph of Read2 (raw data)

The Cutadapt reports of Read1 and Read2:
52F-R1: Adapter 'GATCGGA', length 7, was trimmed 713126 times.
Processed reads: 5456168
Processed bases: 545616800 bp (545.6 Mbp)
Trimmed reads: 713126 (13.1%)
Trimmed bases: 9418910 bp (9.4 Mbp) (1.73% of total)
52F-R2: Adapter 'GATCGGA', length 7, was trimmed 544449 times.
Processed reads: 5456168
Processed bases: 545616800 bp (545.6 Mbp)
Trimmed reads: 544449 (10.0%)
Trimmed bases: 7082484 bp (7.1 Mbp) (1.30% of total)

[syfo]

I would try with the full adapter sequence in the Cutadapt command line ("GATCGGAAGAGCTCGTATG") in case some similarity score threshold was not reached.

[mastal]

When you are trimming adapters from the 3' end, you have to trim the same number of bases from the 3' end of both R1 and R2.

Basically, if you are reading into the adapter sequences, it means your insert is shorter than the read length, and the whole insert has been sequenced twice. The sequences of R1 and R2 will be reverse complements of each other, apart from the 3' end where you read into the adapters.