Machine learning in computational and systems biology

Jean-Philippe Vert

Malaysia Genome Institute
November 8-12, 2010

In this 5-day workshop, we introduce several machine learning techniques and illustrate their use in a variety of applications in computational and systems biology. We emphasize in particular support vector machines (SVM) and kernel methods. Applications include the classification of biological sequences, small molecules or microarray data, as well as de novo and supervised reconstruction of biological networks.

Slides

Schedule

Practical sessions

Prerequisites

• R, with packages kernlab, lars, BioDist.
• Cytoscape, to visualize networks

P0: R basics

• Goal: learn basic R command. R is the language we will use for all other practical sessions.
• R reference card
• R tutorial

P1: SVM and kernel methods basics

• Goal: learn and manipulate SVM and kernel PCA, understand how they work on simple data, play with kernels and parameters.
• Application: cancer diagnosis from gene expression data.
• Notes
• Useful R functions
• Code

P2: Using your own kernels

• Goal: Use precomputed kernels, and define your own kernels.
• Notes
• Code

P3: Classification of sequences with string kernels

• Goal: understand and test a few string kernels, for classification of protein and DNA sequences
• Notes
• Code

P4: Reconstruction of regulatory networks from expression

• Goal: understand and implement several methods to reconstruct the gene regulatory network from a matrix of expression data.
• data.tar.gz, the data set of expression data and regulation of E. coli used by Faith et al.
• Notes
• Code
• Code for stability selection

P5: Reconstruction of PPI and metabolic networks

• Goal: understand and test several methods for the prediction of protein-protein interactions and edges in the metabolic network.
• ppimetabo.tar.gz: yeast datasets (PPI and metabolic networks)
• Notes
• Code