Unravelling gene interactions to find the cause of artherosclerosis, a multigenic disease, using an artificial neural network

To understand the etiology of multigenic diseases like atherosclerosis, a polymerase chain reaction (PCR) based gene array containing 65 single nucleotide polymorphisms (SNPs) was analyzed. To asses the possibilities of pattern recognition techniques in detecting unfavorable genetic combinations, two approaches were analysed. A selection of these 65 SNPs formed the input both to binary logistic regression models and to self-learning artificial neural networks (ANNs). Repeated analyses showed that both methods performed equally well. Further research to improve the differentiating power of both methods should focus first on decreasing the number of otherwise indeterminable polymorphisms.