Variable per tile sequence quality across amplicon-seq fastqs

[dross11]

I sequenced numerous amplicon-seq libraries today and found the sequence per tile quality to be highly variable across the fastq files. I initially thought I had overclustered or debris was present upon/within the flow cell but a third of the fastq files look good. I have clustered the same library type previously at 1200K and never had this issue. My run metrics were pretty good:

~1000M cluster density, 85% PF, 91%>Q30

Any help would be greatly appreciated.

Can I even use this data for downstream analysis?

Collection of FastQC images below:

http://imgur.com/a/6P9v7

[Brian Bushnell]

I think you can use that data, and I have just the tool for it... FilterByTile! I wrote it mainly for improving the overall quality of normal runs with specific problem areas on the flowcell (out of focus, bubbles, etc), rather than for salvaging runs with massive position-dependent quality problems, but it can be used either way. It will simply eliminate all reads from the areas with quality problems, so the only drawback is the you end up with lower overall coverage (but it looks like you will still retain >80% of the data). In your case you probably want to use more aggressive parameters than the defaults.

As for why this happened... I never use Illumina's software, so I'm not very familiar with it, but my recollection is that the flow in these images is top to bottom rather than left to right... correct? Which makes your images odd and I can't explain them. There's no way a flow blockage could cause low quality in the last 5 tiles of a row, good quality in the last 5 tiles of the next row, and low quality again in the last 5 tiles of the row after that.

I'd certainly have an Illumina technician give the machine a thorough review before using it again. If they don't find any problems, or problems persist afterward, I'd insist on refunds for all of the reagents used.

[dross11]

Hi Brian,

The images I posted are actually after I used FilterByTile. When I compared the sequence per tile quality images with and without using FilterByTile I found that it improved tiles in which a whole 'column' of tiles were of poor quality while quality images that are seemingly poor in random areas did not improve at (in some cases it actually made it worse). I have used your script in the past with amazing results but in this instance it doesn't seem to be very helpful.

I will contact Illumina today.

Thank You