docs for paired reads in Mutect2 somatic genotyping #6264
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
takutosato
approved these changes
Nov 15, 2019
docs/mutect/mutect.tex
Outdated
| \subsection{Handling Paired Reads} | ||
| The fundamental unit of evidence in paired-end sequencing is not the read but the fragment of DNA from which two reads were sequenced. If we apply the somatic likelihoods model below directly to the read-vs-haplotype likelihoods matrix from Pair-HMM our model would in effect allow a read and its mate to come from two different biological haplotypes. To prevent this absurdity, we transform the read-vs-haplotype likelihoods matrix into a fragment-vs-haplotype likelihood matrix. The simplest approach, which \code{Mutect2} uses, is to multiply the likelihoods of all paired reads in a fragment (in log space, add) to obtain the fragment's likelihood. That is, $P({\rm fragment} | {\rm haplotype}) = P({\rm read~1} | {\rm haplotype}) P({\rm read~2} | {\rm haplotype})$. Multiplying likelihoods is justified as far as sequencing error is concerned, since sequencing errors on paired reads are statistically independent. There are, of course shared covariates such as sequence context that influence errors on both reads. These, however, are the domain of \code{FilterMutectCalls}'s downstream filtering. The somatic likelihoods model of \code{Mutect2} is \textit{only} concerned with distinguishing sequencing errors (in the narrow sense of an error that occurred on the sequencer itself) from possible somatic variants. \code{FilterMutectCalls} is responsible for distinguishing somatic variants from all other errors, such as those that occur during sample preparation and alignment. | ||
|
|
||
| Simply multiplying likelihoods in this way is not justified when paired reads overlap because while sequencing errors are independent, PCR errors are not. That is, a substitution occurring on both reads may be due to independent sequencing errors or to the amplification of a single PCR error. In order to force the possibility of PCR error into a model of independent reads, we enforce that the effective, multiplied, likelihood must not exceed a bound given by the probability of PCR error. To achieve this, at bases where read pairs overlap, \code{Mutect2} caps the total (that is, the sum, because qualities are measured in a logarithmic phred scale) base and indel qualities to user-defined PCR SNV and indel qualities. This can be adjusted for PCR-free protocols. For example, if the PCR quality is 40 and two reads overlap with base qualities of 25 and 30, each base quality is replaced with $40/2 = 20$. If the base qualities were both, say, 15, no adjustment would be needed because the probability of sequencing error dominates the probability of PCR error. This caping of overlapping base qualities occurs before Pair-HMM, while merging reads into fragments occurs after Pair-HMM. |
There was a problem hiding this comment.
This caping -> This capping
There was a problem hiding this comment.
Hanging comma in "There are, of course..."
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
@takutosato I should have done this earlier.
@droazen This is as requested by our friends at Illumina.