It looks to me like you might be having two problems, though the degree to which each contributes to a failed run may be hard to isolate
The 16s libraries seem to be causing an issue since they're rather low complexity. The first 25 cycles or so want a decent signal in each channel-- the first five cycles for cluster registration and the remaining twenty for calculating the PF rate. MiSeqs assume a more balanced % base composition for the purposes of calculating dye crosstalk, too, so if you have extremely weak signal (which you do in A and G) the system will miscount and/or fail more clusters than it otherwise should.
You should be able to address this problem by bringing up the amount of PhiX that you load on the flow cell-- even if 70% of the aligned sequence on this run is PhiX, there's not enough signal in the A and G channels, I think. That 70% is 70% of 13% X 330K, which is a cluster density of something like 30K PhiX/mm^2. Maybe other users can provide information on their experiences, but I think you need more clusters than that to get good data. You said you're using a spike in of 20-25% Phix, so I assume that is in reference to the 4pM library which means you might be using a lot less PhiX (1-2pM?) than Illumina recommends for runs with low complexity libraries. For a low complexity library, you might want to be more in the 5pM range for PhiX.
Also, it appears as though you are having some kind of issue with library hybridization to the flow cell or sequencing primer hybridization, which is why your cluster density on this last run is so low in comparison to the other run you shared (the most recent run cluster count is three times lower). The fact that you reloaded these libraries at a concentration of 8pM and saw little change in cluster density is proof of this. If you saw a higher raw cluster count but a similarly low PF rate, then I think we could conclude that the PhiX is the primary problem, but if the raw cluster count is similarly depressed when loading more library, I think you have several things to look at, two of which are nucacidhunter's suggestions: the PCR primers/adapters and the sequencing primers.
The next suggestions may be more relevant to the sequencing core you use: if you've ruled out the oligos and the libraries passed QC at the core, and assuming the lab quantifies the libraries via qPCR, it might be worth checking out the NaOH that's used to denature libraries. NaOH does go bad, especially in the presence of CO2 from the air. Libraries should be denatured with a fresh NaOH dilution (I'm in the habit of checking the pH of my dilutions with pH strips to ensure it's above 12). Also, stock solutions should be double checked to make sure the math works out-if the core lab recently ordered new NaOH stocks and got 10N instead of 1N, the calculations in the Illumina denature/dilute guide are going to be off by a factor of 10. Carryover of excess NaOH results in incomplete neutralization with the hyb buffer, which can then interfere with flow cell hybridization.
I apologize for the wall of text, but hopefully these suggestions will be of some help to you.
The 16s libraries seem to be causing an issue since they're rather low complexity. The first 25 cycles or so want a decent signal in each channel-- the first five cycles for cluster registration and the remaining twenty for calculating the PF rate. MiSeqs assume a more balanced % base composition for the purposes of calculating dye crosstalk, too, so if you have extremely weak signal (which you do in A and G) the system will miscount and/or fail more clusters than it otherwise should.
You should be able to address this problem by bringing up the amount of PhiX that you load on the flow cell-- even if 70% of the aligned sequence on this run is PhiX, there's not enough signal in the A and G channels, I think. That 70% is 70% of 13% X 330K, which is a cluster density of something like 30K PhiX/mm^2. Maybe other users can provide information on their experiences, but I think you need more clusters than that to get good data. You said you're using a spike in of 20-25% Phix, so I assume that is in reference to the 4pM library which means you might be using a lot less PhiX (1-2pM?) than Illumina recommends for runs with low complexity libraries. For a low complexity library, you might want to be more in the 5pM range for PhiX.
Also, it appears as though you are having some kind of issue with library hybridization to the flow cell or sequencing primer hybridization, which is why your cluster density on this last run is so low in comparison to the other run you shared (the most recent run cluster count is three times lower). The fact that you reloaded these libraries at a concentration of 8pM and saw little change in cluster density is proof of this. If you saw a higher raw cluster count but a similarly low PF rate, then I think we could conclude that the PhiX is the primary problem, but if the raw cluster count is similarly depressed when loading more library, I think you have several things to look at, two of which are nucacidhunter's suggestions: the PCR primers/adapters and the sequencing primers.
The next suggestions may be more relevant to the sequencing core you use: if you've ruled out the oligos and the libraries passed QC at the core, and assuming the lab quantifies the libraries via qPCR, it might be worth checking out the NaOH that's used to denature libraries. NaOH does go bad, especially in the presence of CO2 from the air. Libraries should be denatured with a fresh NaOH dilution (I'm in the habit of checking the pH of my dilutions with pH strips to ensure it's above 12). Also, stock solutions should be double checked to make sure the math works out-if the core lab recently ordered new NaOH stocks and got 10N instead of 1N, the calculations in the Illumina denature/dilute guide are going to be off by a factor of 10. Carryover of excess NaOH results in incomplete neutralization with the hyb buffer, which can then interfere with flow cell hybridization.
I apologize for the wall of text, but hopefully these suggestions will be of some help to you.
Comment