Abstract
Genetic Algorithms are optimization methods aiming at solving complex problems. They are likely to play an interesting role in proteomics. This discipline is a quite new one which studies the proteins in individuals. It provides high dimension data. <br />The first part deals with history, working of genetic algorithms and introduces some theoretical results. In the next part the building of a genetic algorithm is presented to solve biomarker selection in mass spectrometry and 2D electrophoresis gels alignment. This part focuses on the difficulty to choose an appropriate criterion to optimize. The last part deals with theoretical results. Convergence of elitist genetic algorithms is proved for non homogeneous case and orientated mutations. Then we built a convergence criterion mixing theoretical basements and appliability, which is based on occurrences of the locally optimal solution. Finally, the efficiency of introducing catastrophic events to avoid some convergence problems is shown.