Computational detection of natural selection in gene family expansion and contraction

Here we use a model of stochastic birth and death for mapping gene family evolution onto a phylogeny, and show that it can be efficiently applied to multispecies comparisons. The model offers the opportunity for researchers to make stronger inferences regarding the role of natural selection and changing duplication and deletion rates in gene family expansion or contraction. The work is performed as a part of the first author’s MS thesis under the second author’s supervision.
Nội dung trích xuất từ tài liệu:
Computational detection of natural selection in gene family expansion and contraction’Tap ch´ Tin hoc v` Diˆu khiˆ n hoc, T.23, S.1 (2007), 1—14ıee.. a `.COMPUTATIONAL DETECTION OF NATURAL SELECTIONIN GENE FAMILY EXPANSION AND CONTRACTIONCHI NGUYEN1 , NELLO CRISTIANINI 21 Departmentof Computer Science, University of California, Davis, USA2 Departmentof Statistics, University of California, Davis, USAAbstract. Researchers have generally attributed ostensibly large differences in family size to theeffects of natural selection in promoting either the expansion or contraction of families along specificlineages, without a strong statistical basis. Here we use a model of stochastic birth and death formapping gene family evolution onto a phylogeny, and show that it can be efficiently applied to multispecies comparisons. The model offers the opportunity for researchers to make stronger inferencesregarding the role of natural selection and changing duplication and deletion rates in gene familyexpansion or contraction. The work is performed as a part of the first author’s MS thesis under thesecond author’s supervision.´´’ `T´m t˘t. Thˆng thu.`.ng c´c nh` nghiˆn c´.u g˘n su. kh´c nhau l´.n theo ve bˆ ngo`i cua k´ c˜.oaooaae u aaoea ’ıch o..i su. anh hu.o.ng cua viˆc lu.a chon tu. nhiˆn trong qu´ tr` th´ c dˆ y su. mo. rˆng hay thu’’’ea ınh u a . ’ oe .ınh ogia d` v´ . ’....´´´´ .’’ oınhuhep cua gia d` trong nh˜.ng thˆ hˆ sau m` khˆng du.a trˆn mˆt co. so. thˆng kˆ ch˘c ch˘ n. Tronge aae ea oeo.... dung mˆt mˆ h` thˆng kˆ sinh-tu. dˆ ´nh xa su. tiˆn ho´ cua c´c ho gien’´´’ eaoo ınh oea ’ a .b`i b´o n`y, ch´ ng tˆi su .a a auo ’.. . e’`’t´.i mˆt cˆy phˆn lo`i, v` chı ra r˘ ng mˆ h` n`y c´ thˆ du.o.c ´p dung mˆt c´ch c´ hiˆu qua cho su.oo aaa a ’ao ınh a o eao ao e.......a ra mˆt co. hˆi cho c´c nh` nghiˆn c´.u thu.c hiˆn nh˜.ng suy diˆn ch´˜ooaae ueueınhao ınhso s´nh da lo`i. Mˆ h` dua....´ .x´c ho.n liˆn quan t´.i vai tr` cua su. lu.a chon tu. nhiˆn v` tˆc dˆ nhˆn dˆi hay loai bo trong su. mo.aeoo ’e a o o a o. ’. .... ’`´’’a a oaa ay ’ arˆng v` thu hep ho gien. Kˆt qua nghiˆn c´.u n`y l` mˆt phˆn trong luˆn v˘n thac s˜ cua t´c giaoaee u....... nhˆ t du.´.i su. hu.´.ng dˆ n cua t´c gia th´. hai.˜´’ a’ uth´uao .oa1. INTRODUCTIONOne of the major goals of evolutionary biology has been to identify the genetic changesunderlying phenotypic differences between organisms, and to distinguish the evolutionaryforces responsible for these changes. Past studies have necessarily focused on small numbersof nucleotide differences between orthologous genes, largely because of technical limitationson DNA sequence collection. The recent sequencing of many whole genomes, however, haserased this limitation. Researchers may now focus on large-scale genomic differences betweenorganisms that play an important role in adaptive evolution, including large changes in thesize of gene families (e.g. Tatusov et al. 1997; Lander et al. 2001; Snel et al. 2002).While the newfound ability to observe gene family expansions and contractions has stimulated many new hypotheses, we still lack a statistical framework that would allow for stronginferences regarding gene family evolution. Especially interesting to evolutionary studies arethe causes of changes in gene family size. Unlike the analysis of nucleotide sequence evolu-2CHI NGUYEN, NELLO CRISTIANINItion–where there are well-accepted methods for testing for the action of natural selection(e.g.Yang and Bielawski 2000)–there are no such methods in the analysis of gene familyevolution. Generally, researchers have ascribed large differences in gene family size betweengenomes to natural selection, without any consideration of the expected difference in size dueto random gene gain or loss over long periods of time (e.g. Copley et al. 2003; Lander et al.2001; Lespinet et al. 2002; Oakeshott et al. 1999). While many of these comparisons maycertainly be due to natural selection promoting the expansion or contraction of gene familysize, most are simple comparisons in which one species is found to have a larger or smallernumber of genes.We believe that the inability to make statistical inferences about the role of natural selection in the evolution of gene family size is due to the lack of a null model. With no expectationfor how similar or different in size families are likely to be, researchers are unable to makeprobabilistic statements about observed disparities. While simple statements about the equivalence of two numbers can be made with tests of homogeneity, these tests do not take intoaccount the time since divergence of two taxa. Observing a gene family with 100 members inone taxa and 50 in another is certainly striking if they have diverged for 5 million years, butif they have not shared a common ancestor for 250 million years the biological significanceof the difference is less obvious. In addition, when data are available on gene family sizein more than two taxa, it would be informative to use phylogenetic relationships among thespecies to identify lineage- or branch-specific expansions and contractions (e.g. Lespinet etal. 2002). A statistical model of gene family evolution that allows for both hypothesis testingand phylogenetic inference, therefore, would be very useful.We propose to use the well-studied stochastic birth and death (BD) process as a model forgene family evolution. Birth and death models have been widely studied in statistics (Bailey1964; Darwin 1956; Karlin and Taylor 1975), and have also found use in population geneticsand phylogenetics (e.g. Sims and McConway 2003). The observation in multiple genomes thatboth gene family sizes and gene duplicate ages are approximately Poisson-Dirichlet distributedsuggested that they could be expl ...