Motif recognition in biomolecular sequences
Thomas M. Liebling,
This talk is based on Nicolas Moeri's PhD thesis and will describe some results of a long research collaboration between our unit at DMA and the biologist's group of Philip Bucher from ISREC, the Swiss Institute for Applied Cancer Research.
The problems at issue are how to characterize a "motif" in a family of biomolecular sequences (DNA or protein) with some biological features in common and how to efficiently discriminate between sequences that have and those that don't have the motif.
The motif characterization leads to an interesting hard, combinatorial optimization problem, while the recognition problem is easy.
A new model for motif description using generalized profiles and an efficient heuristic to optimize the associated parameters, given the corresponding sample "learning" sequences will be presented.
Associated preliminary computational results are very encouraging.