The Matrix Coalescent and an Application to Human Single-Nucleotide Polymorphisms

The “matrix coalescent” is a reformulation of the familiar coalescent process of population genetics. It ignores the topology of the gene tree and treats the coalescent as a Markov process describing the decay in the number of ancestors of a sample of genes as one proceeds backward in time. The matrix formulation of this process is convenient when the population changes in size, because such changes affect only the eigenvalues of the transition matrix, not the eigenvectors. The model is used here to calculate the expectation of the site frequency spectrum under various assumptions about population history. To illustrate how this method can be used with data, we then use it in conjunction with a set of SNPs to test hypotheses about the history of human population size.