Necessary to filter out nanopore (or PacBio) reads before mapping?

Question

[BioinformaticsLad]

I have some nanopore reads I want to map to hg38 human genome using minimap2. In the past I would filter out mean QV<7 reads but lately I've been thinking: if the reads are long enough, there's little chance that (despite errors) they map to the wrong regions. Is it really necessary to filter these out? If the coverage is high enough, the consensus should cover up low quality bases. Anyone have any opinions?

Answer 1

[lieven.sterck]

I think there is indeed a growing consensus that you should just map them and if they map OK then use them. In order words, let the mapping do the 'filtering' rather than on basecall quality scores.

[h.mon]

In addition, PacBio Sequel don't use quality scores, and all raw reads have a Phred score of zero.