A Bayesian method to smooth-out the noise in a coverage dataset, revealing the underlying average behaviour which is constrained to lie on `Harmonics' of a fundamental coverage value.
The current statistical model assumes 100% clonality in the sample -- this is visibly making things worse than we would like, and so it is necessary to at least consider how subclonality will affect our data.
This is a very complex mathematical issue though, and if taken in a "clever but stupid" direction runs the risk of becoming ludicrously complex & non-convergent.
A simpler means of inference seems like it would be for the best -- some kind of marginalisation in the PDF?
The current statistical model assumes 100% clonality in the sample -- this is visibly making things worse than we would like, and so it is necessary to at least consider how subclonality will affect our data.
This is a very complex mathematical issue though, and if taken in a "clever but stupid" direction runs the risk of becoming ludicrously complex & non-convergent.
A simpler means of inference seems like it would be for the best -- some kind of marginalisation in the PDF?