Statistical methods of DNA sequence analysis: detection of intragenic recombination or gene conversion

Simple but exact statistical tests for detecting a cluster of associatednucleotide changes in DNA are presented. The tests are based on the lineardistribution of a set of s sites among a total of n sites, where the ssites may be the variable sites, sites of insertion/deletion, orcategorized in some other way. These tests are especially useful fordetecting gene conversion and intragenic recombination in a sample of DNAsequences. In this case, the sites of interest are those that correspond toparticular ways of splitting the sequences into two groups (e.g., sequencesA and D vs. sequences B, C, and E-J). Each such split is termed aphylogenetic partition. Application of these methods to a well-documentedcase of gene conversion in human gamma-globin genes shows that sitescorresponding to two of the three observed partitions are significantlyclustered, whereas application to hominoid mitochondrial DNAsequences--among which no recombination is expected to occur--shows noevidence of such clustering. This indicates that clustering ofpartition-specific sites is largely due to intragenic recombination or geneconversion. Alternative hypotheses explaining the observed clustering ofsites, such as biased selection or mutation, are discussed.