acutkosky / cs273a Goto Github PK

doIt.py

Takes as input a correlation matrix and a desired number of clusters.
Clusters the correlation matrix using spectral clustering.
For each cluster it identifies the marker that is best predicted (scored by least squares) from the other markers in the cluster via linear regression.
Outputs clusters, and best linear regression data for each clusters.

testmodel.py

Tests a regression obtained from doIt.py on chromosome 1 of the IMR90 cell line (or other cell lines/chromosomes if desired).
Outputs the R^2 correlation score.
This is given by
1- v/u
where v is the sum of squared errors, u is the sum of squared errors if the regression always were to guess the average of the data.
So a score of 1.0 is perfect, a score of 0.0 is only as good as always guessing the average.