The Genom of Homo sapiens.pdf

INTEGRATIVE GENOMICS 433with high-density pinning devices. Failure to recover adouble-mutant combination in SGA identifies a candidatesynthetic lethal interaction, which must then be confirmedby direct tetrad analysis to rule out false-positiveand -negative interactions. The SGA approach is readilyapplied to mapping of dominant mutations, dosage lethalityscreens, and chemical genetic screens (Tong et al.2001; Jorgensen et al. 2002b). Recently, an alternativesystematic method termed synthetic lethal analysis by microarray(SLAM) has been developed (Ooi et al. 2003).SLAM employs high-efficiency transformation of thepool of all viable deletion strains with a marked query allele,followed by identification of nonviable combinationsby barcode microarray analysis of the selected population.The SGA and SLAM approaches thus enablecomprehensive HTP mapping of the yeast genetic interactionnetwork.An initial implementation of the SGA method for eightgenes implicated in cytoskeletal control and the DNAdamage response recovered 291 bona fide synthetic lethalinteractions (Tong et al. 2001). Recent application ofSGA to >100 query mutations in a variety of cellular processeshas identified nearly 4,000 synthetic lethal interactions,fourfold more than has previously been reported inthe entire biomedical literature (Tong et al. 2004). On average,each SGA screen uncovers >30 synthetic lethal interactions,considerably more than might have been anticipatedfrom conventional forward screens (Hartwell etal. 1997; Tong et al. 2001, 2004). If the general principlesof genetic networks are conserved, the unexpected densityof genetic interactions revealed in yeast will greatlycomplicate the detection of human polygenic disorders bysingle-nucleotide polymorphism (SNP) mapping (Hartmanet al. 2001; Tong et al. 2004). Genetic interactionmaps reveal functional connections between and withinpathways, and thereby assign roles to uncharacterizedgenes by their position in the network. As data sets increasein size, such connections are readily revealed byclustering algorithms (Tong et al. 2004).PROTEIN INTERACTIONSMost genetic interactions reflect either direct or indirectfunctional connections between the encoded proteinproducts. Protein interactions span a vast range of affinities,from the stable complexes that form cellular machinessuch as the ribosome, RNA polymerases, or theproteasome, to the ephemeral regulatory interactions suchas those that control cell division, signaling, transcription,ubiquitin-dependent proteolysis, directed transport, andvesicle fusion. Transient protein interactions are inevitablydictated by reversible posttranslation modificationssuch as phosphorylation, methylation, acetylation,and ubiquitination. Modified protein sequences are usuallybound in a reversible manner by dedicated recognitiondomains on interaction partners, which thereby bringpathway dynamics under posttranslational control (Pawsonand Nash 2003). Mapping of protein interactions byvarious biochemical means has thus been a major enginefor pathway discovery.Methods for detection of protein interactions have becomeincreasingly sophisticated and sensitive, such thatit is usually no longer necessary to spend arduous monthsin the cold room fractionating complex cell lysates fordesired activities. Two primary approaches are amenableto HTP detection of interactions; namely, the two-hybridsystem and mass spectrometric analysis of protein complexes.The two-hybrid system senses a binary interactionbetween bait and prey fusion proteins by virtue oftranscriptional activation of reporter genes or restorationof other forms of reporter protein activity (Fields andSong 1989; Uetz 2002). The cost-effectiveness of thetwo-hybrid system and its ability to delineate interactionregions through partial-length clones have made it an attractivesystem for HTP interaction studies. Mass spectrometricmethods have recently been applied to HTPcharacterization of protein complexes. In this approach,an epitope-tagged protein complex is affinity-purifiedfrom cell lysate, resolved into its constituents by SDS-PAGE, and each species is identified by mass spectrometry(Fig. 2). Despite the conceptual simplicity of thisscheme, HTP implementation of protein tagging, complexisolation, mass spectrometry, and informatics is farfrom trivial.TWO-HYBRID PROTEININTERACTION MAPSThe first systematic studies for mapping protein interactionnetworks were based on the two-hybrid method.Large-scale two-hybrid screens may be carried out eitherin a library pool format, whereby interacting clones areselected from mass transformation of a cDNA or genomiclibrary, or in an array-based format, whereby each bait isdirectly tested against all possible prey fusions via arobotic mating procedure. Library-based methods arerapid and convenient, whereas array-based formats allowsystematic coverage and are less prone to spurious interactions.The yeast proteome served as a test bed for comprehensivelibrary pool and array-based two-hybridscreens. In the pioneering study by Uetz et al. (2000), 957interactions were recovered from screens with 1,004baits, whereas extensive library pool screens by Ito et al.(2001) yielded a set of 4,549 interactions. Together, thesetwo sets of two-hybrid data contain >5,506 potential proteininteractions, providing both an important resource ofinteraction information for biological studies and abenchmark for subsequent HTP studies. Because of relativeease and low cost, systematic two-hybrid analysishas been carried for other unicellular organisms, includingthe gut pathogen Helicobacter pylori (Rain et al.2001). Recently, the first large-scale two-hybrid proteininteraction map for a metazoan species was reported forDrosophila melanogaster (Giot et al. 2003). Using apooled library approach, 12,000 baits cloned in a recombinationalvector system were used to identify ~20,000interactions, some 5,000 of which were considered to behighly probable based on their local connectivity. Thisstudy elaborated known and novel pathways in manyareas of metazoan cell biology.