UNIVERSITY OF HERTFORDSHIRE COMPUTER SCIENCE RESEARCH COLLOQUIUM presents "The LSA Machine Applied to Cancer Diagnosis by using Gene-expression Data" by Dr. Andreas Albrecht (University of Hertfordshire) 13 February 2002 (Wednesday) Lecture Theatre LC108 3 - 4 pm Coffee/tea and biscuits will be available. EVERYONE IS WELCOME TO ATTEND! Abstract: Within the Human Genome Project it has to become clear that there is a need to develop new methodologies to utilise the genome data in a meaningful way, detecting variations in gene expressions both between individuals and different disease processes and to elucidate further the function of thus far unidentified genes. In the present talk we report the results from the application of our LSA Machine (LSA = Logarithmic Simulated Annealing) to the following problem: Given are gene-expression data from cDNA microarrays containing 2308 genes for each of 63 training samples and 25 test samples of four different types of Small, Round Blue Cell Tumours (SRBCTs). It is important to distinguish between the four types because of their similar appearance in routine histology, while treatment options and responses to therapy vary widely depending on the accurate diagnosis. With a new modification of the LSA Machine we were able to classify correctly all 25 test samples. The results are stable, and the training time on the 63x2308 input is relatively short. The modification of the LSA Machine allows to identify genes which might be significant for the particular cancer type. We provide a comparison to Neural Network Algorithms applied to the same problem, and to Vector Support Machines (applied to gene-expressions of another cancer type). On-line: http://homepages.feis.herts.ac.uk/~nehaniv/colloq/albrecht.txt