Finally, based on the regions previously identified, an unsupervised
classification with model selection criteria allows a status to be
assigned for each region (gain, loss or normal).
Our algorithm involves three steps:
- First, for each chromosome, regions are grouped in classes, each class containing regions of the same expected (but unknown) DNA copy number.
- Second, the resulting classes for all chromosomes are clustered to produce superclasses, of same expected DNA copy number.
- Finally, each superclass is given a status: gain, normal or loss.