a cracking combination
TRANSCRIPT
© 2001 Macmillan Magazines Ltd
H I G H L I G H T S
NATURE REVIEWS | GENETICS VOLUME 2 | DECEMBER 2001 | 915
As genome sequencing and high-throughput functional genomicsapproaches generate more andmore data, researchers neednew ways to tease out biologi-cally relevant information.Two important advances inbioinformatics techniques —both of which rely on the com-bination of data sets — are nowreported that allow better predic-tions of gene and protein function.Ge et al. have shown how gene-expression and protein–protein inter-action data can be integrated to refinepredictions of genetic interactions.And Dietmann and Holm havedevised an improved computationalmethod for detecting remote homol-ogy on the basis of structural similar-ities between proteins.
Ge et al. began their study byinvestigating whether genes that areco-expressed at the transcriptionallevel are more likely to encode pro-teins that interact with each other.They tested this in yeast, using tran-scriptional microarray data and acombination of genome-wide two-hybrid data, as well as individual pro-tein data from literature searches. Ingeneral, clusters of co-expressedgenes do indeed have a higher chanceof giving a positive score in the two-hybrid assays and in experimentaldata derived from the literature. Thislends weight to the idea that co-expression and protein interactionare both useful predictors of func-tional interaction. More importantly,the combination of these two types ofdata can also provide new insightsinto functionally related groups ofgenes. The authors provide an exam-ple in which a group of genes that isinvolved in a network of stress-response protein–protein interactionsis divided into two co-expressionclusters. This refines the view of howthe genes interact and leads to the development of new, testablehypotheses.
Important clues about functioncan come from protein structure,especially if the proteins in questionare distantly related and their amino-
acid sequenceshave diverged. Dietmann and Holmcombine sequence, structural andfunctional information to create atree of protein structures, in whichthe most structurally similar proteinslie on the same branch of the tree.This approach overcomes some ofthe shortcomings of other bioinfor-matics tools. Their analysis can alsoinfer homology relationshipsbetween proteins and group theminto ‘superfamilies’. The results com-pare very favourably with a proteinclassification system that is manuallycurated by experts (SCOP). Finally,the authors show that their automat-ed system could be used to predict thesuperfamily membership, and there-fore a putative function, of a new pro-tein structure that might be deter-mined as part of a structuralgenomics project.
Such computational approachesare essential if we are to predict andunderstand complex genetic net-works from data generated by high-throughput genomic studies. And asthe experimental technologiesimprove and generate more compre-hensive and complementary datasets, the prospects for further combi-natorial approaches to data analysiswill be bright.
Magdalena Skipper
References and linksORIGINAL RESEARCH PAPERS Ge, H. et al.Correlation between transcriptome andinteractome mapping data from Saccharomycescerevisiae. Nature Genet. 10.1038/ng776 (2001) |Dietmann, S. & Holm, L. Identification of homologyin protein structure classification. Nature Struct.Biol. 8, 953–957 (2001)FURTHER READING Lichtarge, O. Getting pastappearances: the many-fold consequences ofremote homology. Nature Struct. Biol. 8, 918–920(2001) | Brenner, S. E. A tour of structuralgenomics. Nature Rev. Genet. 2, 801–809 (2001)
A cracking combination
F U N C T I O N A L G E N O M I C S IN BRIEF
Random monoallelic expression of three genesclustered within 60 kb of mouse t complex genomic DNA.Sano, Y. et al. Genome Res. 15, 1833–1841 (2001)
There are three mechanisms of gene-dosage control inmammals: random X inactivation, parent-of-origin autosomalgene imprinting and random autosomal inactivation. This lastis very rare, but Sano et al. now report a cluster of three genes— Nubp2, Igfals and Jsap1 — on mouse chromosome 17 thatundergo this process. In single cells, these genes show X-like,random monoallelic expression, but can switch between activeand inactive states during cell division — monoallelicexpression correlates with the 50% methylation status of thegenomic region. The authors discuss possible mechanisms forsuch gene expression patterns and their involvement in thebiology of the t complex.
An apolipoprotein influencing triglycerides in humans and mice revealed by comparativesequencing.Pennacchio, L. A. et al. Science 294, 169–173 (2001)
Apolipoproteins are known to affect plasma lipid levels inhumans an important factor in susceptibility to heartdisease. Pennacchio et al. explored the already knownapolipoprotein gene cluster on chromosome 11 for newsusceptibility loci. By comparing mouse and human sequence,they identified a new locus, which when knocked out in miceleads to a substantial increase in plasma lipid levels, but whenoverexpressed causes these levels to drop below wild-typelevels. This marked effect prompted the authors to look forSNPs in the human locus all three rare alleles that theystudied were associated with high lipid levels.
Reciprocal mouse and human limb phenotypescaused by gain- and loss-of-function mutationsaffecting Lmbr1. Clark, R. M. et al. Genetics 159, 715–726 (2001)
Most of the dominantly inherited preaxial polydactly andsyndactyly phenotypes, which affect limb-digit number, mapto human chromosome 7q36 and to a homologous region inmice. By using deletion chromosomes, these authors show that limb defects that map to this region in mice result fromgain-of-function mutations, rather than by haploinsufficiency.Mice with loss-of-function mutations in limb region 1(Lmbr1), which might be allelic to the known limbmorphology mutants Hemimelic extra toes and Hammertoe,have fewer limb digits than normal. Because this phenotype is reciprocal to polydactly, the authors propose that levels of Lmbr1 activity control the morphology of the vertebratelimb skeleton.
D E V E LO P M E N TA L B I O LO G Y
H U M A N G E N E T I C S
G E N E E X P R E S S I O N