Amplicon sequencing

[CPCantalapiedra]

umm forget the last question, I've seen this topic, where your help is awesome again!

http://seqanswers.com/forums/showthread.php?t=13547

[ajthomas]

Thank you very much for your answers, I'm really impressed with your disposal





That's a great idea, but I guess would be 10!/(8!*2!) + 10 = 55, not 10*10??

Lets say I have 64 samples. With the amplicons, say I have 2 options:

1) Create >1kb amplicons. Say 24 loci * 4 amplicons = 96.
2) Create ~300 b amplicons. Say 8 loci * 12 amplicons = 96.

needing 96 primers in both cases for the initial PCR.

1) number of PCRs = 96*64. (or maybe could be pooled in this step?)
Then shearing and repair.
Then pool (so I got 64 samples).
Ligation RL adapters to each sample using MIDs. So, this is the step you say would cost 100-200 $ each sample? I mean: 6400-12800$ in total?

2) number of PCRs = 96*64 (or maybe could be pooled in this step?),
using primers with the universal tail.
Then PCR step 2, to include the MIDs and adapters = 64 PCRs,
with 12 MIDs (this would be 12!/(2!*10!) + 12=78 ¿?)
being 12 * 2 = 24 primers.
Even I could do both PCRs in just one step... although maybe to messy

It seems obvious that 2) is the best option, but I have lost a lot of loci, and incresing it
would mean to increase the number of PCRs being maybe unmanageable (24 * 12 = 288 amplicons --> 288 * 64 PCRs?!?!)

I'm going to take a serious look at targeted sequencing.

Number of MIDs: The number of samples that can be encoded is simply the number of MIDs used on the forward primer times the number of MIDs used on the reverse primer. For example, Sample1-forward MID1, reverse MID1; Sample2-forward MID1, reverse MID2; Sample3-forward MID1, reverse MID3...Sample100-forward MID10, reverse MID10. Once again, this strategy only works if your amplicons are short enough that you can reliably sequence all the way through.

Your experiment does require a lot of PCRs. Shortening the amplicons certainly does increase the complexity and labor involved. That's also a lot of primer sets to optimize. Here's what I did before switching to using Fluidigm's Access Array: I had only 6 amplicons, but many samples. I was multiplexing 16 samples in each library (8 MIDs on the forward primer and 2 MIDs on the reverse primer). A full sequencing plate had 256 samples, requiring somewhere around 1600 PCRs. (Most sequencing runs, however, were only part of a plate, with the rest filled with other things.) Before I switched, I sequenced about 1000 samples over the course of a year and a half. I had a system set up where the amplicons were laid out in a certain way on the PCR plate to help keep them all straight. After running the PCR, I cleaned them up with AMPure beads in the 96-well plate, then quantified them with a fluorescent reagent. After quantification, I fed that information into a spreadsheet I made to calculate the required volumes needed for pooling, then used that to create a program for a Eppendorf epMotion robot to do the pooling. That robot was invaluable for pooling because there's no way I could have pooled all those samples by hand without making a mistake. In your case, pooling all those PCR products will be a challenge. Increasing the number of amplicons to 288 is not a trivial problem. You can do the setup without too much trouble with multi-channel pipettes, but the pooling has to be done one by one. It will be a significant challenge without some sort of automation.

As for LR-PCR versus sequence capture, I think that LR-PCR is probably a better option in your case. Either way you go, you will have the same number of libraries to ligate, so that cost will be the same either way. Sequence capture can be rather costly. With 24 loci LR-PCR is probably a better approach, but if there are others you would be interested in as well, sequence capture might become more attractive.

[CPCantalapiedra]

thanks again!

I will report when we make decision

[Aaron Cooper]

I am sequencing a 5kb amplicon and I would absolutely recommend AGAINST Shearase. I got extremely uneven coverage, which I was worried about in light of the banding I saw in test reactions. Because I was not doing any PCR enrichment after ligating the adapters, I attribute the coverage bias to Shearase. I also generated a logo for the 6nt surrounding my read initiation sites and was pretty amazed at the level of bias for CGCG. I think I managed to scrape good enough data out anyway, but suffice it to say that I will not be using Shearase again.

[njbayonav]

Number of MIDs: The number of samples that can be encoded is simply the number of MIDs used on the forward primer times the number of MIDs used on the reverse primer. For example, Sample1-forward MID1, reverse MID1; Sample2-forward MID1, reverse MID2; Sample3-forward MID1, reverse MID3...Sample100-forward MID10, reverse MID10. Once again, this strategy only works if your amplicons are short enough that you can reliably sequence all the way through.

Your experiment does require a lot of PCRs. Shortening the amplicons certainly does increase the complexity and labor involved. That's also a lot of primer sets to optimize. Here's what I did before switching to using Fluidigm's Access Array: I had only 6 amplicons, but many samples. I was multiplexing 16 samples in each library (8 MIDs on the forward primer and 2 MIDs on the reverse primer). A full sequencing plate had 256 samples, requiring somewhere around 1600 PCRs. (Most sequencing runs, however, were only part of a plate, with the rest filled with other things.) Before I switched, I sequenced about 1000 samples over the course of a year and a half. I had a system set up where the amplicons were laid out in a certain way on the PCR plate to help keep them all straight. After running the PCR, I cleaned them up with AMPure beads in the 96-well plate, then quantified them with a fluorescent reagent. After quantification, I fed that information into a spreadsheet I made to calculate the required volumes needed for pooling, then used that to create a program for a Eppendorf epMotion robot to do the pooling. That robot was invaluable for pooling because there's no way I could have pooled all those samples by hand without making a mistake. In your case, pooling all those PCR products will be a challenge. Increasing the number of amplicons to 288 is not a trivial problem. You can do the setup without too much trouble with multi-channel pipettes, but the pooling has to be done one by one. It will be a significant challenge without some sort of automation.

As for LR-PCR versus sequence capture, I think that LR-PCR is probably a better option in your case. Either way you go, you will have the same number of libraries to ligate, so that cost will be the same either way. Sequence capture can be rather costly. With 24 loci LR-PCR is probably a better approach, but if there are others you would be interested in as well, sequence capture might become more attractive.

Hi ajthomas!! I am very interested in your post because at our lab we are trying to implement the Access Array Fluidgm technology to amplify 20-30 nuclear loci in a wide set of samples (over 2000), but we would like first to prove if the methodology is appropriate for adequate genotype fragments between 100-200pb. I have made some quotations for the pilot experiment and the cost is around $130 per sample (assuming we are going to amplify 48 loci in 48 samples, and the cost does include primers and MID's) which I think is cheap but means a lot of money when we are talking just about an experiment. So I am curious about where did you develop your experiments with Acces Array?, here in Mexico is difficult to have access to these technologies. Also I would like to know how much would be the deep of sequencing for an adequate genotype of a nuclear loci, I assume that you are working with nuclear genes, and sequence your amplicons to genotype (as I read using 454, I am thinking in use Illumina), which coverage would you recommend?

[ajthomas]

Hi ajthomas!! I am very interested in your post because at our lab we are trying to implement the Access Array Fluidgm technology to amplify 20-30 nuclear loci in a wide set of samples (over 2000), but we would like first to prove if the methodology is appropriate for adequate genotype fragments between 100-200pb. I have made some quotations for the pilot experiment and the cost is around $130 per sample (assuming we are going to amplify 48 loci in 48 samples, and the cost does include primers and MID's) which I think is cheap but means a lot of money when we are talking just about an experiment. So I am curious about where did you develop your experiments with Acces Array?, here in Mexico is difficult to have access to these technologies. Also I would like to know how much would be the deep of sequencing for an adequate genotype of a nuclear loci, I assume that you are working with nuclear genes, and sequence your amplicons to genotype (as I read using 454, I am thinking in use Illumina), which coverage would you recommend?

We had a BioMark we had bought a few months earlier, so we got the Access Array loaders and developed the assay ourselves. We had already been doing the amplifications in 96-well plates, so it only a matter of making a few adjustments here and there and scaling it down to work in the Access Array chips. If you're looking for looking for somewhere to develop/do your assays, PM me and we might be able to work some way we could work with you to do it as a service.

As for the depth of sequencing required, that depends on what exactly your primers amplify. If it's a standard nuclear locus, with only two copies per genome, you won't need very deep coverage at all (low double digits should be fine) as there will be no more than two sequences present in each sample. But if that were the case, I don't think you would be doing deep sequencing to genotype. In my case, one of my primer pairs amplifies as many as 10 different loci per haplotype, and I find I have enough data with a few hundred sequences per sample. To answer your question, you just need to figure out about how many different sequences might be present in any given sample and then figure out how many reads you need in order to be sure you see them all.