Individual Gene Cluster Statistics in Noisy Maps

Abstract

Abstract. Identification of homologous chromosomal regions is important for understanding evolutionary processes that shape genome evolution, such as genome rearrangements and large scale duplication events. If these chromosomal regions have diverged significantly, statistical tests to determine whether observed similarities in gene content are due to history or chance are imperative. Currently available methods are typically designed for genomic data and are appropriate for whole genome analyses. Statistical methods for estimating significance when a single pair of regions is under consideration are needed. We present a new statistical method, based on generating functions, for estimating the significance of orthologous gene clusters under the null hypothesis of random gene order. Our statistics is suitable for noisy comparative maps, in which a one-to-one homology mapping cannot be established. It is also designed for testing the significance of an individual gene cluster in isolation, in situations where whole genome data is not available. We implement our statistics in Mathematica and demonstrate its utility by applying it to the MHC homologous regions in human and fly.

Natural Parameter Values for Generalized Gene Adjacency

“…They obtained an expression of the probability that W\ and W^ share at least m genes, but it is computationally intractable. Using a generating function method, Raghupathy and Durand [26] provided a computationally tractable expression of this probability by constraining all faj (as defined in Section 2.1.1) to take on the same value, 0. The generating function is…”

“…Hoberman, et al [16] developed statistical tests for max-gap clusters found in two different searching strategies, reference region and whole genome comparison. Due to the nature of max-gap definition, max-gap clusters cannot be identified by window sampling [26].…”

Diagnosing duplications – can it be done?

“…Proposed paralogons are then refined by statistical analysis to rule out the null hypothesis: that the cluster of paralogs resulted from several independent SGDs that were inserted in the same region by chance. This is most commonly achieved by randomization, although formal statistical methods are beginning to emerge [54][55][56][57][58].…”

Two Plus Two Does Not Equal Three: Statistical Tests for Multiple Genome Comparison

“…In order to distinguish regions that arose from the same ancestral region from unrelated regions that share homologous gene pairs, it is necessary to show that local similarities in gene content could not have occurred by chance. There is an emerging body of work on statistical tests for this purpose 2,3,4,9,10,14,15,18 . However, this work focuses almost exclusively on tests for comparisons of two regions.…”