Closed hyanwong closed 1 year ago
Also true for node 246096 / breakpoint 21137.0
There are no nodes in the wide ARG in which the actual breakpoints do not match any in the HMM-reported metadata.
I guess these could be cases where the recombination node has a descendant path to another recombination nodes which does not contain any samples, so that a breakpoint from a child RE node is percolated upwards to the parent RE node. Could that happen?
It's a case where there's so much uncertainty that the forward hmm doesn't quite agree with the arg one. Probably tiny numerical difference in param values used.
I bet the arg version has more mutations
It's all very messy
Ah, right. I forgot that the forward params were a rerun, so are subject to numerical differences.
Node 606551 in the long ARG has a breakpoint at position 26577 but does not have that position listed as a switch point in either of the HMM runs:
Gives: