Material Detail

This video was recorded at Solomon seminar. One of the pressing open problems of computational systems biology is the elucidation of the topology of genetic regulatory networks. In this talk, we first present a method, called GENIE3, for the unsupervised inference of gene regulatory networks from expression data. This method decomposes the prediction of a regulatory network between p genes into p different feature selection (regression) problems and solves each of these problems using feature importance scores obtained from tree-based ensemble methods such as Random Forests. After a presentation of the method, we discuss its performance in the context of the Dialogue for Reverse Engineering Assessments and Methods (DREAM) challenges, which is an annual international competition aiming at the evaluation of GRN inference algorithms on benchmarks of simulated and real data. GENIE3 was best performer on the DREAM4 In Silico Multifactorial challenge and on the DREAM5 Network Inference challenge. We then present two adaptations of the method for handling time series expression and systems genetics data. In the latter case, the good performance of the method is illustrated on the data from the DREAM5 systems genetics challenge and the more recent StatSeq benchmark.

Disciplines:

• Science and Technology  / Computer Science