Open cginer opened 6 years ago
In a first step to implement this alternative reciprocal overlap criteria, a new Batch Query option is being tested to be added to the Search menu. The tool accepts the specification of a confidence interval to be added to the query breakpoints, a custom percentage of reciprocal overlap and gives the option to consider either the internal part of the inversion, or the inversion with its breakpoints.
As a reminder: the number of inversions found that is given in the section header sometimes is wrong.
To be more clear on what each of the searches is doing, I think the different options should be reestructured a bit:
The 2 only options should be region match and breakpoint match (overlap is a modification of region match).
Not sure if it's clear enough to what type of search the modifiers are referring to.
For the reciprocal overlap we should add "minimal or >" and maybe we can have 2 clicks to clarify how the reciprocal overlap is calculated: (1) Take into account all the region of inversion (including breakpoints), and (2) Take into account only the internal part of the inversion (excluding breakpoints).
With the current breakpoint-based merging algorithm, large inversions predicted by optical mapping or strand-seq are not identified as known ones, even when it seems they are the same. We have the feeling that adding very large error intervals, although should solve that issue, would also have a high risk of merging different inversions. Perhaps an alternative reciprocal overlap criteria (e.g. >80%) would work better for Mb-long inversions.
[Issue raised by Marta and Claudia]