Poor v.3 run

[westerman]

We have done our first v.3 run and the data, while more voluminous (330M beads at 50bp mate-pair vs. 180M beads at 25bp mate-pair), seems to be of worse quality in that we are (a) not achieving as good of a coverage to the reference and (b) the number of reads with errors versus the reference is much higher.

We were fortunate in that the customer who just completed the v.2 mate pair run wanted to redo it using v.3. So I have a good baseline for how our non-model-organism eukaryotic DNA should match up to the reference. For v.2 the coverage ranged from 69% to 87% (depending on the chromosome) while the v.3 data achieved only 56% to 73%. Obviously with about 4 times the data (~2x the beads and 2x the read length) we were expected better coverage with v.3 or at least equal coverage but certainly not worse!

Also the number of beads with errors is much higher than with v.2.

This is our very first v.3 run and, of course, we expect teething problems. I am ready to toss this problem off as being a startup bug. Still, if anyone has some insight into what might be causing the problem then please get hold of me. ([email protected]). Thanks.

[nilshomer]

We have done our first v.3 run and the data, while more voluminous (330M beads at 50bp mate-pair vs. 180M beads at 25bp mate-pair), seems to be of worse quality in that we are (a) not achieving as good of a coverage to the reference and (b) the number of reads with errors versus the reference is much higher.

We were fortunate in that the customer who just completed the v.2 mate pair run wanted to redo it using v.3. So I have a good baseline for how our non-model-organism eukaryotic DNA should match up to the reference. For v.2 the coverage ranged from 69% to 87% (depending on the chromosome) while the v.3 data achieved only 56% to 73%. Obviously with about 4 times the data (~2x the beads and 2x the read length) we were expected better coverage with v.3 or at least equal coverage but certainly not worse!

Also the number of beads with errors is much higher than with v.2.

This is our very first v.3 run and, of course, we expect teething problems. I am ready to toss this problem off as being a startup bug. Still, if anyone has some insight into what might be causing the problem then please get hold of me. ([email protected]). Thanks.

Have you checked for primer/adapter sequence. This could be more of a problem on the v3 machines though it would not explain the worse performance.

[westerman]

Have you checked for primer/adapter sequence. This could be more of a problem on the v3 machines though it would not explain the worse performance.

Yes we did check for adapter read-through. There is some but it appears to be less than 1% of the beads and thus we do not see that causing our problems.

[cmm8cmm8]

Did you perform a WFA on the beads? How did the titration metrics look? Also, how many beads were you getting from each emulsion after 3' end modification? Finally, how much genomic DNA did you start with for the library prep for V3 and for your V2 runs?

[Northeaster]

Hi Westerman,
I am new to v3. Did v3 mate-pair mean the 2x50bp mate-pair and the v2 mate-pair mean EcoP 2x25? Did these two libraries amplified with similar cycles? What percentage of your v3 reads are mappable? I am practicing with E. coli now. It looked good so far.

[westerman]

Northeaster: Yes, v.3 was 2x50bp and v.2 was 2x25bp. I didn't do the amplification but I was told that v.3 was amplified "slightly less". Percentage unique matching for v.3 was 33%. For v.2 it was 37%. But these numbers really will not be applicable to your case since I am working with an eukaryotic organism to a distantly related reference.

cmm8cmm8: I am still trying to find out from the lab people the answers to your questions.

[snetmcom]

You really do need to look at the quality of the beads you put on the slide and the deposition. If the bead counting was fine with no panel failures, then you may want to look at the N2S plots. if cycle 1 has greater than 20% N2S , then you can assume the lab made low quality beads. the WFA would tell you all of this before you ran a full slide.

[westerman]

The WFA run looked fine.

It has been a while since I have posted to this thread. We have been talking with ABI/Lifetech about our run. I believe that the final conclusion is that the run itself was good but that the number of starting points is much less than it should have been. Thus we have a very high coverage on a small sub-set of the genome. How this happened is still up to debate but there seems to be a limited number possibilities: (1) the new version 3 protocol does not work well on our type of genomes or (2) our PCR reactions selected for a limited amount of the initial sequence (2) the lab techs made a mistake. #1 seems unlikely. #2 we are trying to test for. #3 is being denied by the poor lab techs who seem to get all of our abuse in the first place.

[snetmcom]

there is nothing in the v3 protocol that really changes how the system runs. You might want to look at your analysis as well. 50bp has allowed many more unique positions in our runs.

[westerman]

snetmcom:

I'll disagree that there is nothing in the v.3 protocol that has changed from v.2. One really big difference is using nick translation instead of restriction enzymes for paired ends. The latter is exact while the former is a timed reaction which could be subject to a lot of vagrancies. There are other differing, albeit minor, lab steps that have changed.

That said, I do believe that our problem is not the protocol otherwise I do not think that ABI would have released it.

We have done all sorts of computational analysis. The most telling is to analyze the v.3 run as if it were a v.2 run with only 25 bases -- our restricted set of starting points also shows up in this analysis.

[pmiguel]

The WFA run looked fine.

It has been a while since I have posted to this thread. We have been talking with ABI/Lifetech about our run. I believe that the final conclusion is that the run itself was good but that the number of starting points is much less than it should have been. Thus we have a very high coverage on a small sub-set of the genome. How this happened is still up to debate but there seems to be a limited number possibilities: (1) the new version 3 protocol does not work well on our type of genomes or (2) our PCR reactions selected for a limited amount of the initial sequence (2) the lab techs made a mistake. #1 seems unlikely. #2 we are trying to test for. #3 is being denied by the poor lab techs who seem to get all of our abuse in the first place.

After much hand-wringing, the problem was likely caused by a choice I made and then forgot. The DNA sheared into a smaller size range than expected. But we had so much, I thought it would be fine just to extract a slice representing maybe the top 10% of the smear that was in the correct size range. None of the subsequent lab QC showed any problems.

I now think there is some relationship between the size that DNA fragments to and the sequence composition of that DNA. That is, while the break points introduced by shearing may be fairly random, any given narrow slice of that smear may contain a very biased subset of the entire genome. Especially if the slice being chosen is from an extreme (tail) of the distribution.

--
Phillip