New Clustering and Classification Techniques for Detecting Gene Expression Patterns
Dr. Marc Strickert
Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), Gatersleben/Germany


Micro- and macro-array technologies enable multi-parallel probing of genes subject to the living conditions of the considererd organism. The challenging research task is the characterization of specifically grown individuals with their genetic fingerprints and their interrelationships. Unsupervised and supervised artificial neural networks facilitate the analysis. Two unsupervised methods, high-throughput multidimensional scaling (HiT-MDS) and distance-preserving projection pursuit (DiPPP), are introduced for gene profile and experiment clustering; these help detecting similar functional classes. Additonally, generalized relevance learning vector quantization (GRLVQ) with correlation measure is presented for supervised classification and rating of genes. Results are given for barley seed development and for leukemia cancer data.