Physical Coverage

[d3mux]

Hi everyone,
still working on a university project.
Now i have a doubt: i have several mate pairs with one mate (call it A) aligning once and the other one (B) aligning twice.
If i want to calculate the physical coverage on a reference genome, do i have to consider both the matepairs i'd have joining A with both the aligning sites of B?
Or i have to choose one between the two sites where B aligns?
Thanks

[westerman]

The answer is "both" or "one". Depends. You ask a general question and get a general answer.

I assume that all alignments are strong and valid, that the reference is complete, that the reference only has one A and one B, and that your mate pair library is deep enough to not skip potential pairings. In that case let us think about how such a situation can occur. The reference should look like:

...AAAA.....BBBB...

And your sample looks like (where 'B1' and 'B2' are the same B-mate only numbered differently):

case X) ...AAAA....B1B1B1...B2B2B2...
or
case Y) ...B2B2B2...AAAAA...B1B1B1

Or some variant on the above. In other words region B has been duplicated without A also being duplicated. In 'case X' you would only want to count the distance between A and B2. Adding in the distance between A and B1 would add too much distance. In 'case Y' you would want to count both distances.

If my original assumptions are not valid then there are many other cases which would explain your findings.

[GenoMax]

If you know the insert size then only one of the two "B" alignment sites should be correct. If both B alignment sites are within margin of insert size variation then you would have a multi-mapper situation.

[d3mux]

Yes probably is a region who is repeated ad some matepairs have only right (or left) read aligning twice in the repeated regions.
Well... so probably the best approach is to generate two different tracks, one with all possible mates and the other introducing a "bias" to consider valid only the mp sizes between a range (maybe mean +- 3*sigma).
It is indeed useful to show that probabily i have one more clue about a possible repeat.
Thanks