PNN for Molecular Level Selection Detection.

Abstract

Contemporary population genetics has developed several statistical tests designed for the detection of natural selection at the molecular level. However, the appropriate interpretation of the test results is often hard. This is because such factors as population growth, migration, and recombination can produce similar values for some of these tests. To overcome these difficulties, the author has proposed a so-called multi-null methodology, and he has used it in search of natural selection in ATM, RECQL, WRN, and BLM, i.e., in four human familial cancer genes. However, this methodology is not appropriate for fast detection because of the long-lasting computer simulations required for estimating critical values under nonclassical null hypotheses. Here the author presents the results of another study based on the application of probabilistic neural networks for the detection of natural selection at the molecular level. The advantage of the proposed method is that it not so time-consuming and, because of the good recognition abilities of probabilistic neural networks, it gives low decision error levels in cross validation.