Ravi Shankar, Studio of Computational Biology & Bioinformatics, CSIR-Institute of Himalayan Bioresource Technology
16 July 2013
Well, the authors are correct that their is a huge apprehension in buying PacBio sequencers and they are surrounded by some negative publicity. However, this article appeared to be an academic attempt to bring PacBio sequencers in a good light. The entire article misses depth in dealing and really did not address a much technically to prove their points. If sequencing error is there with 14% or so, it is per base chance. Their error correction method is not convincing one, as they claim that they increase coverage to minimize the sequencing errors. This is not a right way and it can't remove innate errors per reads. It is like constituting jury of wrong people in majority, whose majority vote is counted to decide the point of agreement. So, these all jury members will finalize a sequence always with 14% error possibility or so.
PacBio should instead focus upon reducing the innate sequencing errors. None is going to invest on it just for coverage, as its coverage data is actually of no use, and its better to supplement it with Illumina runs, which are cheaper manifolds. Therefore, this is essential for them to reduce the innate sequencing error per read. Length is their forte, no doubt.
A simple question: If PacBio is so great, why not we are seeing frequent reporting of genome or even transcriptome sequencing? Let me also make a point that with such kind of errors per read, this sequencer is logically unreliable for phenomics and variability studies also. Yes this can be good to fill in the patches, and to some extent sort out repetitive DNA menace in assembling.
I am also a bit suprised to see that Genome Biology entertained such articles, which I found substantially lacking on technical front. Had the authors arrived with some comparative and reasonable data with sound arguments, it could be understandable. But this article...I am a bit surprised to see it here. Everyone is eagerly waiting for the sequencers which could just shoot the genomes down and make the lives of computational biologists a bit easy, which will definately pave the ways for easier lives for rest of the guys!
Competing interests
None.
Error rate is random
Mauricio Carneiro, Broad Institute of MIT and Harvard
30 August 2013
Hi Ravi,
The key concept you are missing here is that the error (despite high at 14%) is random, therefore proportional coverage helps makes the error rate irrelevant, no error correction method is needed other than a good variant caller with a bayesian model (hint: GATK).
Technical justification missing in the article
16 July 2013
Well, the authors are correct that their is a huge apprehension in buying PacBio sequencers and they are surrounded by some negative publicity. However, this article appeared to be an academic attempt to bring PacBio sequencers in a good light. The entire article misses depth in dealing and really did not address a much technically to prove their points. If sequencing error is there with 14% or so, it is per base chance. Their error correction method is not convincing one, as they claim that they increase coverage to minimize the sequencing errors. This is not a right way and it can't remove innate errors per reads. It is like constituting jury of wrong people in majority, whose majority vote is counted to decide the point of agreement. So, these all jury members will finalize a sequence always with 14% error possibility or so.
PacBio should instead focus upon reducing the innate sequencing errors. None is going to invest on it just for coverage, as its coverage data is actually of no use, and its better to supplement it with Illumina runs, which are cheaper manifolds. Therefore, this is essential for them to reduce the innate sequencing error per read. Length is their forte, no doubt.
A simple question: If PacBio is so great, why not we are seeing frequent reporting of genome or even transcriptome sequencing? Let me also make a point that with such kind of errors per read, this sequencer is logically unreliable for phenomics and variability studies also. Yes this can be good to fill in the patches, and to some extent sort out repetitive DNA menace in assembling.
I am also a bit suprised to see that Genome Biology entertained such articles, which I found substantially lacking on technical front. Had the authors arrived with some comparative and reasonable data with sound arguments, it could be understandable. But this article...I am a bit surprised to see it here. Everyone is eagerly waiting for the sequencers which could just shoot the genomes down and make the lives of computational biologists a bit easy, which will definately pave the ways for easier lives for rest of the guys!
Competing interests
None.
Error rate is random
30 August 2013
Hi Ravi,
The key concept you are missing here is that the error (despite high at 14%) is random, therefore proportional coverage helps makes the error rate irrelevant, no error correction method is needed other than a good variant caller with a bayesian model (hint: GATK).
Competing interests
No competing interests