Home >
Community >
Does anyone have high-fidelity polymerase recommendations for...
Upvote
22
Downvote
+ Illumina sequencing
+ Dna polymerase
+ Chemistry
+ Polymerase chain reaction
+ Sanger sequencing
Posted by
Mark Hoheisel
Does anyone have high-fidelity polymerase recommendations for...
Hi I came across your question while searching for the problem that I have been facing with the Platinum H F supermen. I have tried so many time the same PCR I am getting on the regular TAQ polymerase and I am not getting any band while working with Platinum supermix. I have tried with dilution, Anealing temprature and it was a stress not getting any band. I set a different profile and I got some bands but it still not work with all my plants DNA.It gives me a hope when I saw that it is not something I am dealing with. Maybe you have got some solutions if so please share it with me. I have to sequence my plants DNA that is why I am doing the PCR with High Fidelity polymerase.
Hi I came across your question while searching for the problem that I have been facing with the Platinum H F supermen. I have tried so many time the same PCR I am getting on the regular TAQ polymerase and I am not getting any band while working with Platinum supermix. I have tried with dilution, Anealing temprature and it was a stress not getting any band. I set a different profile and I got some bands but it still not work with all my plants DNA.It gives me a hope when I saw that it is not something I am dealing with. Maybe you have got some solutions if so please share it with me. I have to sequence my plants DNA that is why I am doing the PCR with High Fidelity polymerase.
I came to this thread because of the same concern of having inosine in my COI primer jgHCO2198 amplified with the KAPA HiFi polymerase after reading about it in another metabarcoding paper. I used this reverse primer in my zooplankton metabarcoding paper and was able to get amplification but it worries me about how well it performed with having the proofreading ability in the polymerase. After a lot of digging and advice from this thread, it looks like New England Biolabs offers a uracil and inosine compatible polymerase with their Q5U Hot Start High Fidelity DNA Polymerase (https://www.neb.com/products/m0515-q5u-hot-start-high-fidelity-dna-polymerase#Product%20Information) product numbers M0515S and M0515L. Has anyone used this polymerase yet with NGS?
The pricing I'm seeing for NEB Q5U is $171 for 100 units or $660 for 500 units compared to the KAPA HiFi Uracil+ which is $282 for 1.25 mL (50 x 50 uL reaction volumes) or $1124 for 6.25 mL (250 x 50 uL reactions). Of course, prices may vary for researchers but it appears NEB offers a cheaper product.
I came to this thread because of the same concern of having inosine in my COI primer jgHCO2198 amplified with the KAPA HiFi polymerase after reading about it in another metabarcoding paper. I used this reverse primer in my zooplankton metabarcoding paper and was able to get amplification but it worries me about how well it performed with having the proofreading ability in the polymerase. After a lot of digging and advice from this thread, it looks like New England Biolabs offers a uracil and inosine compatible polymerase with their Q5U Hot Start High Fidelity DNA Polymerase (https://www.neb.com/products/m0515-q5u-hot-start-high-fidelity-dna-polymerase#Product%20Information) product numbers M0515S and M0515L. Has anyone used this polymerase yet with NGS?
The pricing I'm seeing for NEB Q5U is $171 for 100 units or $660 for 500 units compared to the KAPA HiFi Uracil+ which is $282 for 1.25 mL (50 x 50 uL reaction volumes) or $1124 for 6.25 mL (250 x 50 uL reactions). Of course, prices may vary for researchers but it appears NEB offers a cheaper product.
Apologies for not addressing Rumakanta Sapkota's answer/question in a timely manner. Hopefully you figured out a workable solution by now?
Arsh Bibi: I had the exact same problem and same headaches with no bands. I tested multiple kinds of HiFi Polymerases including the Platinum HiFi. Unfortunately, in my experience, if you really want to use a "high fidelity polymerase", you may have to change the primer set to one without inosine bases. Although, I just saw another intriguing answer above (Tarik Khan's answer) about using a Uracil+ version of Kapa HiFi, which I have not looked into but certainly will now.
In my case (mtCOI metabarcoding) I switched from Geller et al. 2013 (in Molecular Ecology Resources) to primers that are first cited in Lobo et al. 2013 (in BMC Ecology) as LoboR1 - small story here, I actually designed the same exact primer set in my own work after evaluating the Geller primers, switching out the Insoines for degenerate bases that would target the broadest range of known metazoans, and found out after the fact that my primers are the same as in the Lobo paper. That is all to say - there may be an alternate primer choice out there already that would work better with a HiFi polymerase.
Of course there are risks to doing this but you will have to decide which potential source of error is more important. Either by 1) changing your primers, or 2) using a different polymerase. There are plenty of papers out there that make the polymerase compromise, so you are not ruined if you choose the option to keep the Inosine primers. We just wouldn't know how that decision impacts your estimates of diversity, unless you test both options simultaneously.
Apologies for not addressing Rumakanta Sapkota's answer/question in a timely manner. Hopefully you figured out a workable solution by now?
Arsh Bibi: I had the exact same problem and same headaches with no bands. I tested multiple kinds of HiFi Polymerases including the Platinum HiFi. Unfortunately, in my experience, if you really want to use a "high fidelity polymerase", you may have to change the primer set to one without inosine bases. Although, I just saw another intriguing answer above (Tarik Khan's answer) about using a Uracil+ version of Kapa HiFi, which I have not looked into but certainly will now.
In my case (mtCOI metabarcoding) I switched from Geller et al. 2013 (in Molecular Ecology Resources) to primers that are first cited in Lobo et al. 2013 (in BMC Ecology) as LoboR1 - small story here, I actually designed the same exact primer set in my own work after evaluating the Geller primers, switching out the Insoines for degenerate bases that would target the broadest range of known metazoans, and found out after the fact that my primers are the same as in the Lobo paper. That is all to say - there may be an alternate primer choice out there already that would work better with a HiFi polymerase.
Of course there are risks to doing this but you will have to decide which potential source of error is more important. Either by 1) changing your primers, or 2) using a different polymerase. There are plenty of papers out there that make the polymerase compromise, so you are not ruined if you choose the option to keep the Inosine primers. We just wouldn't know how that decision impacts your estimates of diversity, unless you test both options simultaneously.
Just to clarify, when I talked about "quantitative value" I actually meant "relative read abundance". I am also working on eukaryotic COI-metabarcoding. :-) To put some numbers in perspective. Considering an average 1% error rate for Illumina reads (a conservative estimation), if you use Phusion+MiSeq on your data, (11 Mreads, 300 bp on average) you will get 33,000,300 erroneous bases. If you use AmpliTaq+MiSeq, you will get 33,033,000 erroneous bases. And that's actually a very conservative estimation for the total number of errors. In my view, the fidelity gain from using Phusion Taq does not compensate for the loss in undetectable species derived from using less degenerate primers. But, of course, if it works for your particular target taxonomic group, then that's fine.
Just to clarify, when I talked about "quantitative value" I actually meant "relative read abundance". I am also working on eukaryotic COI-metabarcoding. :-) To put some numbers in perspective. Considering an average 1% error rate for Illumina reads (a conservative estimation), if you use Phusion+MiSeq on your data, (11 Mreads, 300 bp on average) you will get 33,000,300 erroneous bases. If you use AmpliTaq+MiSeq, you will get 33,033,000 erroneous bases. And that's actually a very conservative estimation for the total number of errors. In my view, the fidelity gain from using Phusion Taq does not compensate for the loss in undetectable species derived from using less degenerate primers. But, of course, if it works for your particular target taxonomic group, then that's fine.
Hi, just as a small addition: the problem of proofreading polymerases with inosines is their 3'->5' exonuclease activity (Knittel & Picard, 1993; DOI:10.1101/gr.2.4.346). Since the inosines cause a "wobbeling" bais-pairing, the polymerase does not recognize these pairs as good matches and removes them. As a result, no PCR product is generated. In principle you can use any taq-based polymerase that has no 3'->5' exonuclease activity. For illumina sequencing we use DreamTaq from ThermoFisher which gives quite decent results.
Hi, just as a small addition: the problem of proofreading polymerases with inosines is their 3'->5' exonuclease activity (Knittel & Picard, 1993; DOI:10.1101/gr.2.4.346). Since the inosines cause a "wobbeling" bais-pairing, the polymerase does not recognize these pairs as good matches and removes them. As a result, no PCR product is generated. In principle you can use any taq-based polymerase that has no 3'->5' exonuclease activity. For illumina sequencing we use DreamTaq from ThermoFisher which gives quite decent results.
Hi Owen - Thank you for your thoughtful response. I believe it depends on the application and how much one cares about that extra error added by the TAQ being used - that is still two orders of magnitude higher error with one vs the other. Maybe for your application, you are ok with the added error. I would rather be conservative and minimize error wherever possible when I know it exists. Also to clarify before I go on, my work is not microbial, eDNA, or free DNA, it is metazoan community metabarcoding to the species-level, so quantification is not really in the cards - relative sequence abundance across samples is what it is best for at this point, and I can do that just fine.
To elaborate on my concern for the added error: I do not agree with your assessment of negligible error rates and the impact on the quality, or rather confidence, in the OTU's. In my application, I am aiming for species-level discrimination of taxa where a (roughly) >3% difference in base pairs across a ~300 bp amplicon means it is likely to be a different species. That is equal to >9 total base pair differences between two different amplicons that are possibly different species, technically. So, why would I want to increase the artificial bp differences in my dataset? I don't. I want the most accurate representation of what is real. Illumina error rates are going to be the same with either polymerase, and I can control how the TAQ I use impacts my estimation of species. Why not decrease the potential error where you can to result in more accurate representation of the sample being analyzed?
A two order of magnitude increase in error in a MiSeq dataset with e.g. 11,000,000 reads of 300 bp in length corresponds to nearly 33,000 bp more (vs. 300) misrepresentations of the nucleotides in samples by going with the lower quality polymerase. While proportionally it doesn't seem important, I think it is important to minimize any potential for artificially inflating OTU's and genetic haplotypes in datasets like this so that we don't have to regenerate data in the future just because we realize that we could have been more accurate.
As for the replacing of the troublesome Inosines: In my case I did not replace the Inosines with N's. At the 5' end I chose a nucleotide that was represented in a majority of target species, and at the 3' end degenerate bases that best suited the range of taxa that are being targeted (but not N's). Yes, I was making a compromise and taking a chance, and I agree that simply using N's would have caused trouble. I used my expertise and background in testing and designing primers to make a call at each Inosine. We will see how it turns out. So far my data looks fantastic. I am also validating the dataset at a level that will help identify if I have missed anything important by my decision i.e. I am chipping away at individually barcoding many organisms present to verify that they are in the metabarcode dataset (or not), which is more than many do at this point. False negatives are always a concern, even the Inosines won't solve that problem.
And, for quantification comparison we have counts of morphologically ID'd taxa, so we will see how it compares. So far, things are also looking relatively comparable (a.k.a as good as one could hope with community metabarcode data). Confident quantification in the type of sample I have is also not likely regardless of primer sets used - body size varies by many orders of magnitude (from rotifers to mysid shrimp). These are not microbial samples or free eDNA, they are metazoans. Quantification is a pipe dream until we can take 100% of the collection-extraction-amplification-sequencing-analysis biases out of the equation, and we have a long way to go for that to happen. Relative sequence abundance across samples is the best we can do and I can easily get that from the lovely dataset I have produced.
I hope that sufficiently clarifies my reasoning. I look forward to sharing my results when they are published.
Hi Owen - Thank you for your thoughtful response. I believe it depends on the application and how much one cares about that extra error added by the TAQ being used - that is still two orders of magnitude higher error with one vs the other. Maybe for your application, you are ok with the added error. I would rather be conservative and minimize error wherever possible when I know it exists. Also to clarify before I go on, my work is not microbial, eDNA, or free DNA, it is metazoan community metabarcoding to the species-level, so quantification is not really in the cards - relative sequence abundance across samples is what it is best for at this point, and I can do that just fine.
To elaborate on my concern for the added error: I do not agree with your assessment of negligible error rates and the impact on the quality, or rather confidence, in the OTU's. In my application, I am aiming for species-level discrimination of taxa where a (roughly) >3% difference in base pairs across a ~300 bp amplicon means it is likely to be a different species. That is equal to >9 total base pair differences between two different amplicons that are possibly different species, technically. So, why would I want to increase the artificial bp differences in my dataset? I don't. I want the most accurate representation of what is real. Illumina error rates are going to be the same with either polymerase, and I can control how the TAQ I use impacts my estimation of species. Why not decrease the potential error where you can to result in more accurate representation of the sample being analyzed?
A two order of magnitude increase in error in a MiSeq dataset with e.g. 11,000,000 reads of 300 bp in length corresponds to nearly 33,000 bp more (vs. 300) misrepresentations of the nucleotides in samples by going with the lower quality polymerase. While proportionally it doesn't seem important, I think it is important to minimize any potential for artificially inflating OTU's and genetic haplotypes in datasets like this so that we don't have to regenerate data in the future just because we realize that we could have been more accurate.
As for the replacing of the troublesome Inosines: In my case I did not replace the Inosines with N's. At the 5' end I chose a nucleotide that was represented in a majority of target species, and at the 3' end degenerate bases that best suited the range of taxa that are being targeted (but not N's). Yes, I was making a compromise and taking a chance, and I agree that simply using N's would have caused trouble. I used my expertise and background in testing and designing primers to make a call at each Inosine. We will see how it turns out. So far my data looks fantastic. I am also validating the dataset at a level that will help identify if I have missed anything important by my decision i.e. I am chipping away at individually barcoding many organisms present to verify that they are in the metabarcode dataset (or not), which is more than many do at this point. False negatives are always a concern, even the Inosines won't solve that problem.
And, for quantification comparison we have counts of morphologically ID'd taxa, so we will see how it compares. So far, things are also looking relatively comparable (a.k.a as good as one could hope with community metabarcode data). Confident quantification in the type of sample I have is also not likely regardless of primer sets used - body size varies by many orders of magnitude (from rotifers to mysid shrimp). These are not microbial samples or free eDNA, they are metazoans. Quantification is a pipe dream until we can take 100% of the collection-extraction-amplification-sequencing-analysis biases out of the equation, and we have a long way to go for that to happen. Relative sequence abundance across samples is the best we can do and I can easily get that from the lovely dataset I have produced.
I hope that sufficiently clarifies my reasoning. I look forward to sharing my results when they are published.
Hi. Sorry if this answer arrives a bit late. Why do you need an ultra high-fidelity polymerase at all, when you can get the same results using a medium-quality polymerase such as AmpliTaq Gold 360? Keep in mind that the PCR error rate of AmpliTaq Gold is 1 error in every 10^5 bases. Where Phusion Taq arrives to 1 in every 10^7 bases. However, the MiSeq sequencing error rates are typically around 1 per cent (and increasing with sequencing cycle up to 5% after 250 cycles!). That is at least 3 orders of magnitude higher than possible PCR errors! My point is that the number of errors introduced by PCR is going to be absolutely negligible, compared to errors introduced by the sequencer. So you can get perfectly publishable and accurate metabarcoding data using a medium-performance Taq, such as AmpliTaq Gold 360, that has no problem at all in dealing with inosines. I am routinely using AmpliTaq Gold 360 Master Mix in all my COI metabarcoding protocols, with excellent results.
Hi. Sorry if this answer arrives a bit late. Why do you need an ultra high-fidelity polymerase at all, when you can get the same results using a medium-quality polymerase such as AmpliTaq Gold 360? Keep in mind that the PCR error rate of AmpliTaq Gold is 1 error in every 10^5 bases. Where Phusion Taq arrives to 1 in every 10^7 bases. However, the MiSeq sequencing error rates are typically around 1 per cent (and increasing with sequencing cycle up to 5% after 250 cycles!). That is at least 3 orders of magnitude higher than possible PCR errors! My point is that the number of errors introduced by PCR is going to be absolutely negligible, compared to errors introduced by the sequencer. So you can get perfectly publishable and accurate metabarcoding data using a medium-performance Taq, such as AmpliTaq Gold 360, that has no problem at all in dealing with inosines. I am routinely using AmpliTaq Gold 360 Master Mix in all my COI metabarcoding protocols, with excellent results.
Thanks Michelle for the question, and also thanks to Safia and Yvonne . After three rounds of PCR using High Fidelity type polymerase with inosine-base primer, finally I figure out why the reaction was not working. I still want to use inosine-base primers, what do you recommend- DreamTaq or regular TAQ polymerase?
Thanks Michelle for the question, and also thanks to Safia and Yvonne . After three rounds of PCR using High Fidelity type polymerase with inosine-base primer, finally I figure out why the reaction was not working. I still want to use inosine-base primers, what do you recommend- DreamTaq or regular TAQ polymerase?
The solution that you are proposing of replacing inosines by Ns in your primers is not a good idea. You will get amplicon bands for some species, for sure, but you are introducing additional unknown biases for some other species present in the samples, which will not get amplified by your N-loaded primers. The effective concentration of matching primers acting during the PCR reaction for these species affected by the primer bias is reduced by several orders of magnitude. The result is that you will get read abundances with no quantitative value at all. In my view, it is much, much better to use a less fidelity polymerase, and keep the inosines.
The solution that you are proposing of replacing inosines by Ns in your primers is not a good idea. You will get amplicon bands for some species, for sure, but you are introducing additional unknown biases for some other species present in the samples, which will not get amplified by your N-loaded primers. The effective concentration of matching primers acting during the PCR reaction for these species affected by the primer bias is reduced by several orders of magnitude. The result is that you will get read abundances with no quantitative value at all. In my view, it is much, much better to use a less fidelity polymerase, and keep the inosines.
Hi I came across your question while searching for the problem that I have been facing with the Platinum H F supermen. I have tried so many time the same PCR I am getting on the regular TAQ polymerase and I am not getting any band while working with Platinum supermix. I have tried with dilution, Anealing temprature and it was a stress not getting any band. I set a different profile and I got some bands but it still not work with all my plants DNA.It gives me a hope when I saw that it is not something I am dealing with. Maybe you have got some solutions if so please share it with me.
I have to sequence my plants DNA that is why I am doing the PCR with High Fidelity polymerase.
Hi I came across your question while searching for the problem that I have been facing with the Platinum H F supermen. I have tried so many time the same PCR I am getting on the regular TAQ polymerase and I am not getting any band while working with Platinum supermix. I have tried with dilution, Anealing temprature and it was a stress not getting any band. I set a different profile and I got some bands but it still not work with all my plants DNA.It gives me a hope when I saw that it is not something I am dealing with. Maybe you have got some solutions if so please share it with me.
I have to sequence my plants DNA that is why I am doing the PCR with High Fidelity polymerase.
More
VOTE
I came to this thread because of the same concern of having inosine in my COI primer jgHCO2198 amplified with the KAPA HiFi polymerase after reading about it in another metabarcoding paper. I used this reverse primer in my zooplankton metabarcoding paper and was able to get amplification but it worries me about how well it performed with having the proofreading ability in the polymerase. After a lot of digging and advice from this thread, it looks like New England Biolabs offers a uracil and inosine compatible polymerase with their Q5U Hot Start High Fidelity DNA Polymerase (https://www.neb.com/products/m0515-q5u-hot-start-high-fidelity-dna-polymerase#Product%20Information) product numbers M0515S and M0515L. Has anyone used this polymerase yet with NGS?
The pricing I'm seeing for NEB Q5U is $171 for 100 units or $660 for 500 units compared to the KAPA HiFi Uracil+ which is $282 for 1.25 mL (50 x 50 uL reaction volumes) or $1124 for 6.25 mL (250 x 50 uL reactions). Of course, prices may vary for researchers but it appears NEB offers a cheaper product.
I came to this thread because of the same concern of having inosine in my COI primer jgHCO2198 amplified with the KAPA HiFi polymerase after reading about it in another metabarcoding paper. I used this reverse primer in my zooplankton metabarcoding paper and was able to get amplification but it worries me about how well it performed with having the proofreading ability in the polymerase. After a lot of digging and advice from this thread, it looks like New England Biolabs offers a uracil and inosine compatible polymerase with their Q5U Hot Start High Fidelity DNA Polymerase (https://www.neb.com/products/m0515-q5u-hot-start-high-fidelity-dna-polymerase#Product%20Information) product numbers M0515S and M0515L. Has anyone used this polymerase yet with NGS?
The pricing I'm seeing for NEB Q5U is $171 for 100 units or $660 for 500 units compared to the KAPA HiFi Uracil+ which is $282 for 1.25 mL (50 x 50 uL reaction volumes) or $1124 for 6.25 mL (250 x 50 uL reactions). Of course, prices may vary for researchers but it appears NEB offers a cheaper product.
More
VOTE
Hi, I am a bit late to the party, but I have had good success with:
KAPA HIFI HotStart Uracil+
http://sequencing.roche.com/en/products-solutions/by-category/library-preparation/library-amplification/kapa-hifi-uracil-kit.html
"The Uracil+ version enabled efficient high-fidelity
amplification of TAK multiplex primer set that had deoxyinosines."
http://advances.sciencemag.org/content/2/3/e1501371
Hi, I am a bit late to the party, but I have had good success with:
KAPA HIFI HotStart Uracil+
http://sequencing.roche.com/en/products-solutions/by-category/library-preparation/library-amplification/kapa-hifi-uracil-kit.html
"The Uracil+ version enabled efficient high-fidelity
amplification of TAK multiplex primer set that had deoxyinosines."
http://advances.sciencemag.org/content/2/3/e1501371
More
VOTE
Apologies for not addressing Rumakanta Sapkota's answer/question in a timely manner. Hopefully you figured out a workable solution by now?
Arsh Bibi: I had the exact same problem and same headaches with no bands. I tested multiple kinds of HiFi Polymerases including the Platinum HiFi. Unfortunately, in my experience, if you really want to use a "high fidelity polymerase", you may have to change the primer set to one without inosine bases. Although, I just saw another intriguing answer above (Tarik Khan's answer) about using a Uracil+ version of Kapa HiFi, which I have not looked into but certainly will now.
In my case (mtCOI metabarcoding) I switched from Geller et al. 2013 (in Molecular Ecology Resources) to primers that are first cited in Lobo et al. 2013 (in BMC Ecology) as LoboR1 - small story here, I actually designed the same exact primer set in my own work after evaluating the Geller primers, switching out the Insoines for degenerate bases that would target the broadest range of known metazoans, and found out after the fact that my primers are the same as in the Lobo paper. That is all to say - there may be an alternate primer choice out there already that would work better with a HiFi polymerase.
Of course there are risks to doing this but you will have to decide which potential source of error is more important. Either by 1) changing your primers, or 2) using a different polymerase. There are plenty of papers out there that make the polymerase compromise, so you are not ruined if you choose the option to keep the Inosine primers. We just wouldn't know how that decision impacts your estimates of diversity, unless you test both options simultaneously.
Best of luck.
Apologies for not addressing Rumakanta Sapkota's answer/question in a timely manner. Hopefully you figured out a workable solution by now?
Arsh Bibi: I had the exact same problem and same headaches with no bands. I tested multiple kinds of HiFi Polymerases including the Platinum HiFi. Unfortunately, in my experience, if you really want to use a "high fidelity polymerase", you may have to change the primer set to one without inosine bases. Although, I just saw another intriguing answer above (Tarik Khan's answer) about using a Uracil+ version of Kapa HiFi, which I have not looked into but certainly will now.
In my case (mtCOI metabarcoding) I switched from Geller et al. 2013 (in Molecular Ecology Resources) to primers that are first cited in Lobo et al. 2013 (in BMC Ecology) as LoboR1 - small story here, I actually designed the same exact primer set in my own work after evaluating the Geller primers, switching out the Insoines for degenerate bases that would target the broadest range of known metazoans, and found out after the fact that my primers are the same as in the Lobo paper. That is all to say - there may be an alternate primer choice out there already that would work better with a HiFi polymerase.
Of course there are risks to doing this but you will have to decide which potential source of error is more important. Either by 1) changing your primers, or 2) using a different polymerase. There are plenty of papers out there that make the polymerase compromise, so you are not ruined if you choose the option to keep the Inosine primers. We just wouldn't know how that decision impacts your estimates of diversity, unless you test both options simultaneously.
Best of luck.
More
VOTE
Just to clarify, when I talked about "quantitative value" I actually meant "relative read abundance". I am also working on eukaryotic COI-metabarcoding. :-)
To put some numbers in perspective. Considering an average 1% error rate for Illumina reads (a conservative estimation), if you use Phusion+MiSeq on your data, (11 Mreads, 300 bp on average) you will get 33,000,300 erroneous bases. If you use AmpliTaq+MiSeq, you will get 33,033,000 erroneous bases. And that's actually a very conservative estimation for the total number of errors. In my view, the fidelity gain from using Phusion Taq does not compensate for the loss in undetectable species derived from using less degenerate primers. But, of course, if it works for your particular target taxonomic group, then that's fine.
Just to clarify, when I talked about "quantitative value" I actually meant "relative read abundance". I am also working on eukaryotic COI-metabarcoding. :-)
To put some numbers in perspective. Considering an average 1% error rate for Illumina reads (a conservative estimation), if you use Phusion+MiSeq on your data, (11 Mreads, 300 bp on average) you will get 33,000,300 erroneous bases. If you use AmpliTaq+MiSeq, you will get 33,033,000 erroneous bases. And that's actually a very conservative estimation for the total number of errors. In my view, the fidelity gain from using Phusion Taq does not compensate for the loss in undetectable species derived from using less degenerate primers. But, of course, if it works for your particular target taxonomic group, then that's fine.
More
VOTE
Hi,
just as a small addition: the problem of proofreading polymerases with inosines is their 3'->5' exonuclease activity (Knittel & Picard, 1993; DOI:10.1101/gr.2.4.346). Since the inosines cause a "wobbeling" bais-pairing, the polymerase does not recognize these pairs as good matches and removes them. As a result, no PCR product is generated. In principle you can use any taq-based polymerase that has no 3'->5' exonuclease activity. For illumina sequencing we use DreamTaq from ThermoFisher which gives quite decent results.
Hi,
just as a small addition: the problem of proofreading polymerases with inosines is their 3'->5' exonuclease activity (Knittel & Picard, 1993; DOI:10.1101/gr.2.4.346). Since the inosines cause a "wobbeling" bais-pairing, the polymerase does not recognize these pairs as good matches and removes them. As a result, no PCR product is generated. In principle you can use any taq-based polymerase that has no 3'->5' exonuclease activity. For illumina sequencing we use DreamTaq from ThermoFisher which gives quite decent results.
More
VOTE
Hi Owen -
Thank you for your thoughtful response. I believe it depends on the application and how much one cares about that extra error added by the TAQ being used - that is still two orders of magnitude higher error with one vs the other. Maybe for your application, you are ok with the added error. I would rather be conservative and minimize error wherever possible when I know it exists. Also to clarify before I go on, my work is not microbial, eDNA, or free DNA, it is metazoan community metabarcoding to the species-level, so quantification is not really in the cards - relative sequence abundance across samples is what it is best for at this point, and I can do that just fine.
To elaborate on my concern for the added error:
I do not agree with your assessment of negligible error rates and the impact on the quality, or rather confidence, in the OTU's. In my application, I am aiming for species-level discrimination of taxa where a (roughly) >3% difference in base pairs across a ~300 bp amplicon means it is likely to be a different species. That is equal to >9 total base pair differences between two different amplicons that are possibly different species, technically. So, why would I want to increase the artificial bp differences in my dataset? I don't. I want the most accurate representation of what is real. Illumina error rates are going to be the same with either polymerase, and I can control how the TAQ I use impacts my estimation of species. Why not decrease the potential error where you can to result in more accurate representation of the sample being analyzed?
A two order of magnitude increase in error in a MiSeq dataset with e.g. 11,000,000 reads of 300 bp in length corresponds to nearly 33,000 bp more (vs. 300) misrepresentations of the nucleotides in samples by going with the lower quality polymerase. While proportionally it doesn't seem important, I think it is important to minimize any potential for artificially inflating OTU's and genetic haplotypes in datasets like this so that we don't have to regenerate data in the future just because we realize that we could have been more accurate.
As for the replacing of the troublesome Inosines:
In my case I did not replace the Inosines with N's. At the 5' end I chose a nucleotide that was represented in a majority of target species, and at the 3' end degenerate bases that best suited the range of taxa that are being targeted (but not N's). Yes, I was making a compromise and taking a chance, and I agree that simply using N's would have caused trouble. I used my expertise and background in testing and designing primers to make a call at each Inosine. We will see how it turns out. So far my data looks fantastic. I am also validating the dataset at a level that will help identify if I have missed anything important by my decision i.e. I am chipping away at individually barcoding many organisms present to verify that they are in the metabarcode dataset (or not), which is more than many do at this point. False negatives are always a concern, even the Inosines won't solve that problem.
And, for quantification comparison we have counts of morphologically ID'd taxa, so we will see how it compares. So far, things are also looking relatively comparable (a.k.a as good as one could hope with community metabarcode data). Confident quantification in the type of sample I have is also not likely regardless of primer sets used - body size varies by many orders of magnitude (from rotifers to mysid shrimp). These are not microbial samples or free eDNA, they are metazoans. Quantification is a pipe dream until we can take 100% of the collection-extraction-amplification-sequencing-analysis biases out of the equation, and we have a long way to go for that to happen. Relative sequence abundance across samples is the best we can do and I can easily get that from the lovely dataset I have produced.
I hope that sufficiently clarifies my reasoning. I look forward to sharing my results when they are published.
Hi Owen -
Thank you for your thoughtful response. I believe it depends on the application and how much one cares about that extra error added by the TAQ being used - that is still two orders of magnitude higher error with one vs the other. Maybe for your application, you are ok with the added error. I would rather be conservative and minimize error wherever possible when I know it exists. Also to clarify before I go on, my work is not microbial, eDNA, or free DNA, it is metazoan community metabarcoding to the species-level, so quantification is not really in the cards - relative sequence abundance across samples is what it is best for at this point, and I can do that just fine.
To elaborate on my concern for the added error:
I do not agree with your assessment of negligible error rates and the impact on the quality, or rather confidence, in the OTU's. In my application, I am aiming for species-level discrimination of taxa where a (roughly) >3% difference in base pairs across a ~300 bp amplicon means it is likely to be a different species. That is equal to >9 total base pair differences between two different amplicons that are possibly different species, technically. So, why would I want to increase the artificial bp differences in my dataset? I don't. I want the most accurate representation of what is real. Illumina error rates are going to be the same with either polymerase, and I can control how the TAQ I use impacts my estimation of species. Why not decrease the potential error where you can to result in more accurate representation of the sample being analyzed?
A two order of magnitude increase in error in a MiSeq dataset with e.g. 11,000,000 reads of 300 bp in length corresponds to nearly 33,000 bp more (vs. 300) misrepresentations of the nucleotides in samples by going with the lower quality polymerase. While proportionally it doesn't seem important, I think it is important to minimize any potential for artificially inflating OTU's and genetic haplotypes in datasets like this so that we don't have to regenerate data in the future just because we realize that we could have been more accurate.
As for the replacing of the troublesome Inosines:
In my case I did not replace the Inosines with N's. At the 5' end I chose a nucleotide that was represented in a majority of target species, and at the 3' end degenerate bases that best suited the range of taxa that are being targeted (but not N's). Yes, I was making a compromise and taking a chance, and I agree that simply using N's would have caused trouble. I used my expertise and background in testing and designing primers to make a call at each Inosine. We will see how it turns out. So far my data looks fantastic. I am also validating the dataset at a level that will help identify if I have missed anything important by my decision i.e. I am chipping away at individually barcoding many organisms present to verify that they are in the metabarcode dataset (or not), which is more than many do at this point. False negatives are always a concern, even the Inosines won't solve that problem.
And, for quantification comparison we have counts of morphologically ID'd taxa, so we will see how it compares. So far, things are also looking relatively comparable (a.k.a as good as one could hope with community metabarcode data). Confident quantification in the type of sample I have is also not likely regardless of primer sets used - body size varies by many orders of magnitude (from rotifers to mysid shrimp). These are not microbial samples or free eDNA, they are metazoans. Quantification is a pipe dream until we can take 100% of the collection-extraction-amplification-sequencing-analysis biases out of the equation, and we have a long way to go for that to happen. Relative sequence abundance across samples is the best we can do and I can easily get that from the lovely dataset I have produced.
I hope that sufficiently clarifies my reasoning. I look forward to sharing my results when they are published.
More
VOTE
Hi. Sorry if this answer arrives a bit late. Why do you need an ultra high-fidelity polymerase at all, when you can get the same results using a medium-quality polymerase such as AmpliTaq Gold 360? Keep in mind that the PCR error rate of AmpliTaq Gold is 1 error in every 10^5 bases. Where Phusion Taq arrives to 1 in every 10^7 bases. However, the MiSeq sequencing error rates are typically around 1 per cent (and increasing with sequencing cycle up to 5% after 250 cycles!). That is at least 3 orders of magnitude higher than possible PCR errors! My point is that the number of errors introduced by PCR is going to be absolutely negligible, compared to errors introduced by the sequencer. So you can get perfectly publishable and accurate metabarcoding data using a medium-performance Taq, such as AmpliTaq Gold 360, that has no problem at all in dealing with inosines. I am routinely using AmpliTaq Gold 360 Master Mix in all my COI metabarcoding protocols, with excellent results.
Hi. Sorry if this answer arrives a bit late. Why do you need an ultra high-fidelity polymerase at all, when you can get the same results using a medium-quality polymerase such as AmpliTaq Gold 360? Keep in mind that the PCR error rate of AmpliTaq Gold is 1 error in every 10^5 bases. Where Phusion Taq arrives to 1 in every 10^7 bases. However, the MiSeq sequencing error rates are typically around 1 per cent (and increasing with sequencing cycle up to 5% after 250 cycles!). That is at least 3 orders of magnitude higher than possible PCR errors! My point is that the number of errors introduced by PCR is going to be absolutely negligible, compared to errors introduced by the sequencer. So you can get perfectly publishable and accurate metabarcoding data using a medium-performance Taq, such as AmpliTaq Gold 360, that has no problem at all in dealing with inosines. I am routinely using AmpliTaq Gold 360 Master Mix in all my COI metabarcoding protocols, with excellent results.
More
VOTE
Thanks Michelle for the question, and also thanks to Safia and Yvonne . After three rounds of PCR using High Fidelity type polymerase with inosine-base primer, finally I figure out why the reaction was not working. I still want to use inosine-base primers, what do you recommend- DreamTaq or regular TAQ polymerase?
Thanks Michelle for the question, and also thanks to Safia and Yvonne . After three rounds of PCR using High Fidelity type polymerase with inosine-base primer, finally I figure out why the reaction was not working. I still want to use inosine-base primers, what do you recommend- DreamTaq or regular TAQ polymerase?
More
VOTE
Hi, I am a bit late to the party, but I have had good success with:
KAPA HIFI HotStart Uracil+
http://sequencing.roche.com/en/products-solutions/by-category/library-preparation/library-amplification/kapa-hifi-uracil-kit.html
"The Uracil+ version enabled efficient high-fidelity
amplification of TAK multiplex primer set that had deoxyinosines."
http://advances.sciencemag.org/content/2/3/e1501371
Hi, I am a bit late to the party, but I have had good success with:
KAPA HIFI HotStart Uracil+
http://sequencing.roche.com/en/products-solutions/by-category/library-preparation/library-amplification/kapa-hifi-uracil-kit.html
"The Uracil+ version enabled efficient high-fidelity
amplification of TAK multiplex primer set that had deoxyinosines."
http://advances.sciencemag.org/content/2/3/e1501371
More
VOTE
The solution that you are proposing of replacing inosines by Ns in your primers is not a good idea. You will get amplicon bands for some species, for sure, but you are introducing additional unknown biases for some other species present in the samples, which will not get amplified by your N-loaded primers. The effective concentration of matching primers acting during the PCR reaction for these species affected by the primer bias is reduced by several orders of magnitude. The result is that you will get read abundances with no quantitative value at all. In my view, it is much, much better to use a less fidelity polymerase, and keep the inosines.
The solution that you are proposing of replacing inosines by Ns in your primers is not a good idea. You will get amplicon bands for some species, for sure, but you are introducing additional unknown biases for some other species present in the samples, which will not get amplified by your N-loaded primers. The effective concentration of matching primers acting during the PCR reaction for these species affected by the primer bias is reduced by several orders of magnitude. The result is that you will get read abundances with no quantitative value at all. In my view, it is much, much better to use a less fidelity polymerase, and keep the inosines.
More
VOTE