Introduction et al 2001 et al et al et al et al et al et al et al et al et al et al et al Lehner and Fraser, 2004 et al et al et al et al in vivo et al et al et al et al et al Teichmann and Babu, 2002 et al et al et al et al et al et al et al et al et al physically associated Arabidopsis thaliana Mus musculus Drosophila melanogaster Caenorhabditis elegans Saccharomyces cerevisiae Results Predicting physically associated proteins from patterns of conserved co-expression Figure 1 et al Figure 2 C. elegans et al Figure 2A Figure 2B Figure 2C  P I D P I D I I D P I P I Figure 2C A. thaliana M. musculus D. melanogaster S. cerevisiae Figure 2C Supplementary Figure 1 C. elegans A. thaliana M. musculus D. melanogaster S. cerevisiae Validation of predicted physical protein associations using known interactions et al Figure 3A Figure 3A et al et al Figure 3B et al et al et al et al Lehner and Fraser, 2004 N Validation of predicted physical protein associations by mass spectrometry Figure 4A et al Figure 4B Figure 4C Supplementary Figures 3–5 Figure 4C et al et al Figure 5 Figure 6A Figure 5 et al Figure 6A Figure 6A Figure 6B et al et al Figure 6B et al et al et al et al et al Quantitative estimates of interaction accuracy Figure 7A Supplementary Figure 6 et al 2001 et al et al et al et al et al et al Figure 7B et al et al Figure 7C et al Figure 7D et al Figure 7D Table I Figure 7B Figures 3 Supplementary Table 2 et al et al et al Supplementary Figure 10 Detailed evaluation of ribosome biogenesis proteins Granneman and Baserga, 2004 Doll and Grzeschik, 2001 et al et al Audhya and Emr, 2003 et al et al et al 7 BCP1 7 YHR020W Figure 8B and C et al Figure 8A Figure 4 Figure 8B Supplementary Figure 4 Discussion Characteristics of the newly mapped associations et al Supplementary Table 1 et al et al et al et al et al Brown and Jurisica, 2005 Supplementary Table 1 Table I Supplementary Figure 7 et al Supplementary Figures 8 and 9 Figure 9 et al et al et al P −98 N et al et al P −15 N P −25 N Supplementary Table 3 C. elegans Supplementary Table 4 et al Limitations, false positives, and potential improvements et al Similarly, the mass spectrometry data used to test the CCE associations have some important features and limitations. Primarily, co-sedimentation alone is not proof of physical association—it is possible for unrelated complexes to co-sediment—as reflected in the measured true-positive and false-negative rates for associations inferred solely from these data. These sedimentation-derived associations should thus not be viewed as standalone. However, as a benchmark applied in the manner we present (e.g. analyzed in aggregate form), or when considered in combination with other data, such as incorporated into the BIOS scores of the CCE associations, we find the mass spectrometry data to be extremely valuable. We suggest that benchmarks of this sort could be of great utility for evaluating physical complexes determined by other methods, and could be generally adopted for measuring assay accuracy. Conclusions in vivo in vivo Materials and methods Mapping of orthologs et al A. thaliana C. elegans D. melanogaster M. musculus S. cerevisiae http://bioinformatics.icmb.utexas.edu/idserve M. musculus C. elegans A. thaliana D. melanogaster S. cerevisiae Supplementary Table 5 mRNA expression data Supplementary Table 4 et al et al 2004b et al Calculation of co-expression t  r n t t P t P Removal of cross-hybridization artifacts et al et al et al E −4 Carlson, 2002 Carlson, 2002 et al Kim and Iyer, 2004 Supplementary Figure 2 E −4 Training to extract physical protein associations et al x y Figure 2 naïve Supplementary information The human-only co-expression control set was generated by considering only the human DNA microarray data, ignoring contributions from other organisms and lifting the requirement for each member of a gene pair to have orthologs in the same second organism. Putative associations were identified as for the CCE case, but instead using the log likelihood framework to relate the correlation coefficients across only the human DNA microarray experiments to the likelihood of physically associating. All other calculations were performed identically to the CCE case, including calculation of correlation coefficients, significance testing of correlations, calculation of likelihood values, selection of priors, and filtration for cross-hybridization. Testing for enrichment of known physical associations P I P I P I D P I D P I D P I D Figure 3A Testing for functional similarity et al et al Construction of standard curves for estimating percentages of physical associations Figure 7B and C et al Figure 7B Figure 7B and C Figure 7D et al Figure 7D Supplementary Figure 6 Binary interaction overlap score et al Table I Supplementary Figure 10 Human cell culture and mass spectrometry 2 g 2 g g g g et al et al et al Supplementary information Yeast media and strains et al et al Polysome profile analysis 600 600 600 2 g 260 et al Immunoblotting Precipitated proteins were resuspended in 20 μl Laemmli buffer and 2 μl of each sample was deposited onto nitrocellulose membrane. TAP-tagged proteins were detected with PAP antibody (Rockland Immunochemicals Inc.) and chemiluminescence (ECL; Amersham Biosciences). Supplementary Material Supplementary Figures  Supplementary Tables  Supplementary Information 1  Supplementary Information 2