Dr. Robin Gras

Dr. Robin Gras is Associate Professor and Canadian Research Chair in Probabilistic Heuristics and Bioinformatics at the School of Computer Science of the University of Windsor. He is also cross-appointed by the Biological Department at the University of Windsor. He was senior scientist, from 2000 to 2004, in the Swiss Institute of Bioinformatics, Geneva Switzerland after being post-doctorant from 1998 to 2000 in the same institute and lecturer, in 1998, at the University of Rennes, France. He received his B.Sc. and his M.Sc. in computer science at the University of Rennes. He completed his Ph.D. in computer science applied to bioinformatics at INRIA of Rennes in 1997, and obtained his Habilitation a Diriger la Recherche in 2004 in the University of Rennes. From 2000 to 2002 he was also consultant for GeneProt Inc. concerning the automation of protein identification and characterization process.

Dr. Gras research focuses on analyzing and modeling complex biological systems. Most of the biological processes involve a dynamic system of interacting components. In general, the network of interactions between these components is partially or completely unknown. As the number of components involves is very large and the complexity of the network is very high, no exact analysis methods can provide a result in a reasonable time. He works on heuristics approaches based on the building of probabilistic models of the data and simulation of dynamic interacting systems to provide good approximations of the underlying studied processes’ model. The uniqueness of his researches comes from two directions. First, he improves and combines several efficient methods to discover dependencies in a dataset and use these information for feature selection and to build high accuracy predictors. This is particularly important to be able to understand the new data coming from system biology (gene expression data and proteomics) and from clinical measurement. Second, he has conceived a very detail simulation framework based on low level interaction between the agents of the system. It is the only existing simulation able to represent cooperative and competitive agents with complex evolving behavior, emergence and death of species based on genomic set representation and learning capacity of the agents.

His current projects include:

1. Analysis of system biology data (e.g. micro-array, proteomics) to provide diagnosis tools and models of the interactions between the molecules (RNA or proteins) involved in the biological mechanisms. The major difficulties in this problem come from the very large number of parameters (number of genes or proteins) concerned (more than 10000) at a time, the noise level of the data and the small size of the data set. They develop a wrapping machine learning approach using a Probabilistic Model Building Genetic Algorithm (PMBGA) to filter the none-essential parameters and to provide a relevant classifier.
   
2. Conception of a probabilistic model building genetic programming algorithm (PMBGP). The major difference compared with PMBGA is that in this case the search is in the space of functions not in the space of discrete values vectors. They will apply this algorithm to learn discriminating functions for classification. They will therefore by able to discover dependencies between sub-functions and then provide biological insight about the exiting interactions in the system they study.
   
3. Applying a PMBGA approach to solve combinatorial optimization bioinformatics problems such as: none-unique probe selection problem for micro-array, contact-map overlap problem for protein alignment and prediction of protein-protein interaction sites from a set of sequence features.
   
4. Conceiving and implementing an agent-based predator-prey evolving ecosystem simulation based on Fuzzy Cognitive Map as agents’ behavioral model. The behavioral model of the agents evolves during the simulation allowing efficient and competitive behaviors to naturally emerge. This simulation also involves a new model for speciation based on relative genomics distances between individuals and species. This simulation will be used to study problems of invasive species, difference between speciation mode for sexual and asexual organism, the species diversity patterns

 Several Master and PhD positions are available for these projects.