This is a tool that is (at the moment) just OcculterCut with some bug fixes. Perhaps one day it will grow into something that can stand on its own, who knows?
The largest 'bugfix' is that OcculterCut predicts GC and AT regions over scaffold gaps, which may skew results of an inattentive user. This tool also parallelises the calculations over the available cores.
More deeply, Ripsaw aims to expand the information it uses for sequence classification. OcculterCut calculates the Jensen–Shannon divergence of a binomical distribution (A+T vs G+C). At present, Ripsaw calculates the J-S divergence of the 14-value distribution of all possible dinucleotides. It finds breakpoints that maximise the dinucleotide distribution divergence (in the same manner as OcculterCut) and then outputs the small bins with their 14-value distibution calculations. From there, I need to classify the bins and then re-join neighboring bins that have the same class.
OcculterCut fits a mixture of Cauchy Distributions to the GC density distribution and finds a value that most cleanly intersects the distributions. There are other possibilities (k-means, SOM, etc) which I have only made tentative attempts at here.
Installation/compilation is
- Clone the repository
go build .
- Writing tests Yeah, I'll get to that eventually. At the moment the project is too simple, but that might change.
Any questions - talk to me - rob.syme@curtin.edu.au