Summary module swaps LR to RL orientation

[zhemingfan]

Currently calls which are correctly called as LR are being swapped to RL in the summary module.

Traceback (most recent call last):
File "/projects/jfan_prj/jfan_prj/MAVIS/mavis/venv/bin/mavis", line 33, in
sys.exit(load_entry_point('mavis', 'console_scripts', 'mavis')())
File "/projects/jfan_prj/jfan_prj/MAVIS/mavis/src/mavis/main.py", line 295, in main
raise err
File "/projects/jfan_prj/jfan_prj/MAVIS/mavis/src/mavis/main.py", line 255, in main
summary_main.main(
File "/projects/jfan_prj/jfan_prj/MAVIS/mavis/src/mavis/summary/main.py", line 217, in main
grouped, removed = group_by_distance(bpp_set, distances)
File "/projects/jfan_prj/jfan_prj/MAVIS/mavis/src/mavis/summary/summary.py", line 191, in group_by_distance
grouped_calls.append(group_events(pairs))
File "/projects/jfan_prj/jfan_prj/MAVIS/mavis/src/mavis/summary/summary.py", line 137, in group_events
raise AssertionError('cannot group events differing on key elements', bpp, new_bpp)
AssertionError: ('cannot group events differing on key elements', 'BPP(Breakpoint(1:3522792-3522852L), Breakpoint(1:3522805-3522839R), opposing=False, tracking_id=vcf-cuteSV.INS.74)', 'BPP(Breakpoint(1:3522792-3522839R), Breakpoint(1:3522792-3522852L ), opposing=False)')

[creisle]

After our debugging discussion offline we have narrowed down the cause of this error. When breakpoints are being grouped in the summary module a new breakpoint is created from the grouped breakpoints. However when a new breakpoint pair is created we auto-sort the breakpoints before assigning then 1/2 here

mavis/src/mavis/breakpoint.py

Line 263 in 479fdeb

if b1.key[:3] > b2.key[:3]:

This sort method is insufficient when the breakpoint intervals are imprecise ranges. Overall there are 2 things to fix for this bug

• change the breakpoint sort in BreakpointPair to not sort when the ranges overlap: if the breakpoints have the same chrom but their ranges overlap do not sort the breakpoints in the init function
• change the convert module to not allow start ranges which encompass areas past the end range

In this instance we are seeing the bizzarre SV that looks like this

ref       -------------------------------------------------------------------------------
break1 ------------------------=======================--------------
break2 ------------------------==========-------------------------------------

which should be impossible