NextSeq having problem sequencing amplicons with stretch of C

[Patrick_Li]

We seems to have problem sequencing several different amplicons with a stretch of C on the NextSeq using v2 chemistry 300-cycle kits. This happen to 2 different experiments on 2 different NextSeqs. Every single one of these amplicons sequenced fine on the MiSeq. We contacted Illumina tech support and they said it could be RTA2 issue, a 2-channel chemistry issue, or enzyme slipping issue. They did specifically said that “RTA2 which does not perform empirical phasing correction on a per cycle basis, but rather on an algorithm based on the phasing rate from the first 25 cycles.”. Their bioinformatics group is looking into the issue, but I’d like to see if other NextSeq users have this problem too and perhaps have some insights.

Here are some background info:

1) The first run has 476 unique amplicons and the second run has 42 different amplicons. The overall run was good (~90% cluster PF, 86.2% >= Q30, > 60G yield) despite slightly higher cluster density and vast majority of the amplicons sequenced just fine. I’ve attached 2 screen shots of the SAV.

2) All amplicons were sequences at least twice (2 different samples) and the problem was very reproducible between different samples with the same amplicon.

3) We did some position weight matrix analysis and it seems to pull out a CCCCCCCACCCC motif. However, a good number of amplicons that have a good matching score to the motif sequenced just fine. We look for other near homopolymer (A, T and G) in our set and they don’t seems to cause too much problem. The C stretch seems necessary but not sufficient.

4) Our amplicons were designed to be short so that they (most if not all) were covered by both reads. This problematic sequence seems to only affect one of the 2 reads. i.e. CCCCCCCACCCC is bad, but GGGGTGGGGGGG is fine. Therefore, we don’t think it’s because G is dark. (I double-checked and I believed that I did not flip the strand.)

5) Looking at the sequence quality, we observe a substantial drop off that persist through the rest of the read. Maybe this is a RTA basecalling algorithm problem rather than a polymerase dye chemistry problem? This looks similar to an issue we had a few years ago with our MiSeq that was resolved with a software update.

6) These 3 amplicons have problem with read 1: NNNNCTTTGAAGGCACAGCTATTTGAGAACAATAAAAAAAGAAAGCATTTTGTTCTCTTAACTGCTGTTCAGGTGGTGGAGGCCCCCCCACCGCCCCCATCCTTTCCACAGGGAGGGGAGTGGCAACGTTGTGTTTTATGGTGGCCAAAACCCTTCTCCTGCACC

NNNNGATTGTGCTGGCCAACGGAGGACCTGGTGGTGGACCCGTCCAGTCTAGCCCTGCACCCCTACAACCCCCTCCTCCCCACCCCACCTCTGGGCTGAGCAGGGAGCTCAGACCTCTGTGCACTCCAAGTATCAGCCCATTG

NNNNTAAAACACAAAGATCAGGCCCACACAGATGCCCACACTGCCCGCCCCCCCCTCCACCCCACTTCAGGATCAAGATGAAATGGAGCCATAACATTCAAAAACAGAGACACTGGCTTTCAGAATAAAGGACGGCTTGAC

7) These 5 amplicons have problem with read 2:
NNNNATGTCATTGTCTCTGGGGAAGGGTGGGCGGGGACCCAGCTCTGCTGGAGAACAGCCGGTGATAGCCAGAAGCCCTCAGGGTTTCTAGAGGGATGGAAAGAAAAGAGACTAGGTGAAGCAAAGC

NNNNGTTTCACGGGATTAGCTGACACAGCATGTAATCACCTTTCTGCTGCTCCCTAAGGCGGGGGGTGGGGGGTCTATTACTGTTGCTTCTTAAGTAAGCTTAGATGGAGCCTGGCTTCGCAGGCCCAGAATCC

NNNNCAGAACTCTAGTCTCAGCCTGATCCCATGGAGAACTCAAGAAGTCATGAGTTGCATCACAGAGGCTGTTGCACCCAGAGGTGGGGGTGGGGGGTGTCTGTTAGATCACCACATCAGTTAGTCATTGGTCTCAGCTTGCT

NNNNGAGGGGACATGAATCAGGAGAGAAGGAGGAAGGAAGAGTGAGCCGGGAGGGGAGAGGACAGGAGAGGCGGAGGCGGGGTGGGGGGCTCGGGCTGGGCCGTCTGGAGTCCCAGCTCCTCTCACTGTCATTAAGGAT

NNNNGTGGAAACCCGGGGCCAGCGGCTGGCAGCCCGGGATCCAAAATAACCTGAGGTGGGGGGGGAGGGGTCGCCCACACCCCGTACGGGGTCGGGAGCGTCCCCAGGGGAGCCCGTCTAGCCGCACCCTCCAGTTG

Any thoughts or comments?

[Codemonkey]

“RTA2 which does not perform empirical phasing correction on a per cycle basis, but rather on an algorithm based on the phasing rate from the first 25 cycles.”

I can assure you that this is categorically incorrect.

Spikes in the phix mismatch rate by cycle will show any basecaller failures. No spikes would indicate a sequence specific error/chemistry issue on this particular motif.