An artificial neural network for estimating haplotype frequencies

Abstract

The problem of estimating haplotype frequencies from population data has been considered by numerous investigators, resulting in a wide variety of possible algorithmic and statistical solutions. We propose a relatively unique approach that employs an artificial neural network (ANN) to predict the most likely haplotype frequencies from a sample of population genotype data. Through an innovative ANN design for mapping genotype patterns to diplotypes, we have produced a prototype that demonstrates the feasibility of this approach, with provisional results that correlate well with estimates produced by the expectation maximization algorithm for haplotype frequency estimation. Given the computational demands of estimating haplotype frequencies for 20 or more single-nucleotide polymorphisms, the ANN approach is promising because its design fits well with parallel computing architectures.

Data mining

“…In this case, however, the haplotypes, and therefore the diplotypes, in the output layer were unknown. Thus, Cartier and Baechle [2005] trained the network against a distribution of haplotypes that was consistent with the observed genotype data, and arbitrarily assumed that each of the possible diplotypes consistent with a given genotype had an equal probability of being correct. These methods were applied to the first 10 loci from 2,047 individual records selected from replicate 1 of the combined Aipotu, Danacaa, and Karangar data sets in the simulated data.…”

“…While not done in this paper, prediction scores could then be used as a measure for linkage and/or association studies. Cartier and Baechle [2005] used ANN for a distinctly different purpose, i.e., estimation of haplotype frequencies. Thinking of this as a multivariate classification problem, they developed a neural network with two hidden layers that used genotypes as the input layer, and diplotypes (haplotype pairs) as the output layer.…”