Display physical coverage in Icarus in addition to reads coverage about quast HOT 6 CLOSED

Could you explain use cases where physical coverage track will be useful?

When a contig/scaffold ends or is misassembled, physical coverage helps troubleshoot whether the assembly, the reads, or the reference assembly is to fault.

When a region has no read coverage but does have physical coverage, it indicates a region where the assembly cannot assemble a contig, but should be able to scaffold over.

If a region has low/no read coverage and low/no physical coverage, it indicates a region where the assembly does not expect a contig nor scaffold.

If a region has typical read coverage but has low physical coverage, it can indicate a region of reference misassembly.

from quast.

Should we consider single reads or improperly paired reads somehow?

Single reads and improperly paired reads should be ignored in physical coverage calculation.

Reads pairs for which both reads are multimapped, can be useful to either count or ignore depending on the situation. Ignoring them is useful to indicate genomic positions where scaffolds should end, because the repeat at that position is larger than the sequencing fragment size.

from quast.

Should we check insert size deviation somehow and, e.g. skip remotely mapped reads?

Filtering out improperly paired reads should be good enough, as long as that flag is set. Super long incorrectly paired fragments could possibly dominate the physical coverage count if not filtered out.

from quast.

How do you calculate physical coverage usually? samtools/bedtools/some ad hoc script?

Filter as above, create a BED file of (read start pos, read start pos + isize) and use bedtools genomecov.

from quast.

We added physical coverage to coverage track. @sjackman, please take a look at the attached example and feel free to checkout the latest source code. We are waiting for your comments :)
alignment_viewer.html.zip

Note that we calculate physical coverage as you suggested (extracting properly paired reads and counting the gaps between the paired-end reads and reads itself as covered fragments). If you find this feature useful in some cases, please send us few examples -- it will be very interesting to look at them!.

from quast.

Thanks! I'll test it out and get back to you. In the example I notice that there are spots were read coverage drops quickly but it is smoother in physical coverage. That's a good example of why scaffolding works, because there are positions with poor read coverage, but still spanned by paired-end reads.

from quast.