First 20bp of MiSeq reads are weird

[reubennowell]

Hello,

The first 20 bases of my MiSeq reads show abnormal %A, T, G and C, as evidenced by the 'per base sequence content' tab of the FastQC report (see the attached PNG). The per base GC content is similarly weird, but the quality of these bases is good.

The issue can be easily rectified by removing the first 20 bp of each read, but can anyone enlighten me as to what is causing this? I have used both CutAdapt and TagDust on these reads to get rid of adapter sequences. I thought maybe it was the Illumina barcodes, except the barcode sequence is usually contained within the fastq header, thus:

@MVM-RI-I124161:11:000000000-A3985:1:1101:18249:1757 1:N:0:TAAGGCGANAGATCGC

And searching for this sequence (i.e. TAAGGCGANAGATCGC as above) doesn't reveal it to be at the start of the read.

What is it?? And what's the best way of dealing with it? Simply chop the first 20bp off my reads or is it something that requires a bit more QC?

Thanks!

[Bukowski]

Is that RNA-Seq by any chance?

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2896536/

[mastal]

First 20bp of MiSeq reads are weird

We are seeing this in most of our Illumina data too, and we aren't doing RNA-Seq.

I think the reasons have been discussed in previous threads, if I can find one of the discussions I'll post the link.

[reubennowell]

Thanks guys,

Nope, not RNA-Seq - this is bacterial genomic DNA. Mastal, can you remember what you did to account for it? Trim the first X bases off the 5' end? In my datasets, it seems to disappear completely after base #20.

[mastal]

First 20bp of MiSeq reads are weird

I just remove those 20 or so bases from the start of the reads.

[GenoMax]

This is not an abnormal observation. Generally you should not need to trim this data.

Is this a known genome or unknown? If known you can take a few reads and map them. There should be no problems getting this data to map.

http://seqanswers.com/forums/showthread.php?t=11843

http://seqanswers.com/forums/showthread.php?t=17219

[JackieBadger]

Ive seen something similar @~9bases in one or two amplicon libraries.
I didn't delve too deep into what the cause was, I just knew that it had to go.

[nickloman]

If this is from libraries made with Nextera it may represent biases in incorporation sites favoured by the transposase. We see this phenomenon frequently and don't find it necessary to trim it.