DrYak
commented
1 year ago Not quite.
LoFreq is mostly done position wise:
- It calls SNVs per position (in a way that is roughly reminiscent in how samtools' pileup and bcftools works).
- It uses some additionnal heurisitc to increase the confidence of the results (it does more than a simple basecount).
ShoRAH at its core is a local haplotype caller, SNVs are a by-product of this:
- it divides all the aligned reads into windows.
- NOTE there is currently a bad interaction between this division into windows and multiplex PCR amplicon protocols (e.g.: ARTIC v4.1) causing to lose window. A new version fixing this should come in the comming months
- within each window, ShoRAH clusters reads together.
- logic is that real SNVs each coming from the same haplotype will cluster together
- whereas sequencing error will always be randomly spread among the read and will not cluster (i.e.: no matter how the sampler creates clusters, the error will fail to group together).
- in each window, local haplotypes are called from the consensus of each such cluster of reads
- (this will "average-out" any sequencing error according to the model).
- SNV are then called simply by comparing these local haplotype with the reference
- (at that stage, sequencing errors have been eliminated by the clustering, any difference should be "true SNV" as far as the model is concerned).
ibseq
Hi all, does shorah algorithms follow the same principle as lofreq?
thanks Ibseq