Miseq Paired-End Cluster Identification

[retroant]

Hello All,

I have been scouring the internet to try to find the answer to my question, but to no avail. We are attempting to sequence multiple barcoded 16s paired-end amplicons on the Miseq and to get around the low-diversity issue we plan to stagger the sequencing start position for our runs by 0-5bp (incorporated into our amplicon oligos).

The question that I have concerns the cluster identification. If the read 1 sequence has the first 4bp at a ratio of 1:1:1:1 for A:T:G:C and the identification of clusters is good, are the values obtained here used for read 2 as well? Or does the same process occur on read 2 with cluster identification... I would assume not since the location on the flow-cell determines the clusters, and doing this twice may give two completely different cluster location "pictures".

Any insight here would be greatly appreciated! Thanks!

[HMorrison]

Hello All,

I have been scouring the internet to try to find the answer to my question, but to no avail. We are attempting to sequence multiple barcoded 16s paired-end amplicons on the Miseq and to get around the low-diversity issue we plan to stagger the sequencing start position for our runs by 0-5bp (incorporated into our amplicon oligos).

The question that I have concerns the cluster identification. If the read 1 sequence has the first 4bp at a ratio of 1:1:1:1 for A:T:G:C and the identification of clusters is good, are the values obtained here used for read 2 as well? Or does the same process occur on read 2 with cluster identification... I would assume not since the location on the flow-cell determines the clusters, and doing this twice may give two completely different cluster location "pictures".

Any insight here would be greatly appreciated! Thanks!

Warning--more than just the first four bases of read one are critical, but not for cluster finding. We tried the game of putting four random bases at the start of read 1 based on what we knew about cluster finding. Clusters are found, and are used in read 2, but the calculation of phasing/prephasing depends on bases 1-12 of both reads. Unless you are okay with routinely spiking in 30% OR MORE (current recommendation from Illumina is 60% on Miseq), you wil have to design primers so that the first 12 bases read from both ends are high complexity. Illumina is working on a software fix. What we observed was a good read 1 (with 9 bases of randomness by design at start) and a failed read 2--okay to ~125 bases of 250 and then a precipitous drop off. We had ~20% PhiX included.

[GenoMax]

retroant: See this thread. Check posts #47 onwards. http://seqanswers.com/forums/showthr...t=20455&page=3

Using a hard-coded matrix (from a good phiX run) seems to help mitigate the problem to some extent. If you are using an upgraded MiSeq (running MCS v.2.0.5) then you can work with your FAS to see if this workaround helps. The quality values are still degraded but you may be willing to accept those specially if you have a good overlap between the two reads from the 2 x 250 bp run.

[Vinz]

Using a hard-coded matrix (from a good phiX run) seems to help mitigate the problem to some extent. If you are using an upgraded MiSeq (running MCS v.2.0.5) then you can work with your FAS to see if this workaround helps. The quality values are still degraded but you may be willing to accept those specially if you have a good overlap between the two reads from the 2 x 250 bp run.

Hi GenoMax, could you explain what you mean by degraded quality values? Did you analyze whether the Q values with hardcoded matrix/phasing still are correct?

[GenoMax]

Hi GenoMax, could you explain what you mean by degraded quality values? Did you analyze whether the Q values with hardcoded matrix/phasing still are correct?

The quality values still dropped rapidly after ~130 cycles on both reads (2 x 250 bp) while using the hard-coded matrix/phasing (average Q30 was about 61%). No phiX was included in samples since this was a custom primer run.

We have not been able to do a direct comparison of running a set of samples with and without the hard-coded matrix as yet. We should have that data by next week.

We were not concerned about the quality for run I refer to since we knew that MiSeq has a problem with samples having reduced nucleotide diversity. We did get very good overlap with the two reads in the middle as expected with no errors (disclaimer: based on a few read pairs I checked, this data was for someone else).