SVFactorie is an attempt to build a conditional random field (CRF) based predictor of genomic structural variations from data generated by high-throughput sequencing. The features used include those generated by our Cloudbreak Hadoop-based SV caller, as well as a new change point detection algorithm, simple read counts, and some genomic annotations like repeat masker and segmental duplication tracks. It uses the very cool Factorie library (http://factorie.cs.umass.edu/ ) to train CRFs and make inferences.
This is still very experimental software, without an official release.