Opened 6 days ago
Closed 6 days ago
#1636 closed enhancement (fixed)
Structural variants are not classified correctly by SigProfiler which causes incorrect signatures
Reported by: | Nicklas Nordborg | Owned by: | Nicklas Nordborg |
---|---|---|---|
Priority: | blocker | Milestone: | Reggie v5.5.2 |
Component: | net.sf.basedb.reggie | Keywords: | |
Cc: |
Description
It seems like the SigProfiler package has a bug in the sv_matrix_generator
code. We are using the the breakpoints.bedpe
file as input and SigProfiler uses the strand1
and strand2
columns to classify the variants (if the two locations are on the same chromosome). See the table below.
On the other hand, Viola-SV (see https://viola-sv.readthedocs.io/en/1.0.x-stable-doc/userguide/sv_position_specification.html) indicates a different mapping.
strand1 | strand2 | SigProfiler | Viola-SV |
+ | - | inversion | deletion |
- | + | inversion | tandem-duplication |
+ | + | deletion | inversion |
- | - | tandem-duplication | inversion |
Lyckily, SigProfiler also accepts a bedpe-like formation where the two strandX
columns have been replaced with svclass
. It should be relatively easy to implement the same classification as Viola-SV in our own code before sending it to SigProfiler.
In 8008: