FASTQC reports

[buthercup_ch]

Hello everyone,

I just received a set of sequences from an RNA-Seq experiment.
Illumina HiSeq 2000, pair-end sequencing was performed.

Quality of R1 sequences look good, but R2 not so... (please, see attached reports as an example). R2 shows a lot of biased.

My question is how to proceed with these data? Do I need to ask the Genomic platform to resequence, or I can just treat the R2 sequences in order to get rid of the bad quality nucleotides? I am thinking that I could just trim 5 nts at the 5' end and 25 at the 3'end and then continuing with mapping and DEG analysis.

What do you think?

Thanks in advance

[nucacidhunter]

R2 does not show any bias, your sequencing centre should explain why the R2 quality is so low. My guess is that they have over clustered or had chemistry/instrument issues.

[Persistent LABS]

Hi buthercup_ch,
The FastQC reports suggest that overall quality of reads are okay, and there are no known adapter/contaminants. Since re-sequencing will cost you extra, I would suggest to give a try to use this data for DEG analysis.
You can try to remove the bad quality bases using tools like cutadapt setting -q 30. Ask the sequencing center which adapter or primers they have used, and use that information to trim the reads. Then, re-run FastQC on this filtered data set, and see if you get better set of reads to proceed further.

[buthercup_ch]

Hello,

Thanks for your comments.

I agree trimming will improve the overall quality of the reads set.
But what about the Quality per tile diagram? It is the first time I face such a pattern, with those thick red areas all over the reads.
What might be the deleterious effect of such a pattern? Is it not so fatal effect?

Thanks again

[GenoMax]

Hello,

Thanks for your comments.

I agree trimming will improve the overall quality of the reads set.
But what about the Quality per tile diagram? It is the first time I face such a pattern, with those thick red areas all over the reads.
What might be the deleterious effect of such a pattern? Is it not so fatal effect?

Thanks again

It appears that some tiles consistently dropped out over the course of the run (red lines). There may have been bubbles stuck in those positions. That should not affect the rest of your data. The facility that produced this data should have QC'ed the flowcell, so we will assume the rest of the lanes were ok and there was no technical problem with the run.

Just trim the first 2 bp (mainly because you have N's in position 2). Most of the data is still Q15 and above. See what you get in terms of alignment before you start doing additional trimming.

[buthercup_ch]

Hello GenoMax,

The procedures in the Core Facility are in fact slightly dark for me.
Nevertheless, thanks for your comments and advised. At first, analysis was carried out trimming the first 3 nt and alignments looked rather good. We are now performing an additional trimming up to the first 5 nt, and will see if anything is improved.
I was hesitating in trimming any bp at the end of the reads, but finally decided not to do it, as the Q scores are, as you say, not soooo bad (still Q>15), and the overall interquantile is over Q=20.

[GenoMax]

3 nt at beginning should be plenty. Sounds like otherwise everything looks good.

[Persistent LABS]

Hello,

Thanks for your comments.

I agree trimming will improve the overall quality of the reads set.
But what about the Quality per tile diagram? It is the first time I face such a pattern, with those thick red areas all over the reads.
What might be the deleterious effect of such a pattern? Is it not so fatal effect?

Thanks again

Simon Andrews has talked about the possible types of errors introduced during sequencing that might cause the tile plot to appear abnormal.
Here is the link which might help you to understand your case: https://sequencing.qcfail.com/articl...-of-flowcells/