(A) Full network containing 9,780 association lines connecting 876 protein nodes. Lines are shaded by the degree of STRING confidence in the association. Nodes are colored according to the following: known mitochondrial-localized proteins (reference set) correctly predicted by the linear classifier, green; known mitochondrial-localized proteins not captured by the linear classifier, light green; predicted proteins not annotated as mitochondrial-localized (mitochondrial candidates), orange; proteins predicted as additional interactors by the network (interactor candidates), blue; mitochondrial candidates recently annotated as mitochondrial-localized (MitoP2 database) or verified by mitochondrial import assay, red.
(B) Module map of 46 modules with five or more proteins. Modules were named and localized based on GO terms, with the following abbreviations: asm, assembly; biogen, biogenesis; cyt, cytoplasmic; dehy, dehydrogenase; met, metabolism; mito, mitochondrial; org, organization; proces, processing; syn, synthesis. The localization of modules in three different compartments—nucleus, mitochondria, and cytoplasm—is indicated by sectors of different colors. When the module contains a mixture of proteins with different localization it is annotated as shared between the different compartments. Module shared between mitochondria and nucleus or mitochondria and cytoplasm belong to green and yellow sectors, respectively. Cytoplasm refers to all of the contents of a cell excluding mitochondrion and nucleus but including the plasma membrane and other sub-cellular structures. The identity of all proteins and their functional links can be found in Figure S4 and Figure S5 for (A and B), respectively, where the standard gene names are shown within the nodes and are hyperlinked to STRING.

Orthology was used to identify functional processes in yeast that originate from bacteria and are conserved to humans. Modules of five or more proteins are shown as single nodes at the circumference of the circle and the degree of evidence connecting modules is shown by lines in different gray tone. The connection between modules is calculated as the average of the STRING interaction scores for all protein-pairs. Modules are distributed according to their localization. Color in the inner rings reflects for each module the percentage of proteins that have proteobacterial orthology (red) or human orthology (blue). The number of proteobacterial orthologs is also indicated.

Samples were derived by incubating radiolabeled proteins with isolated mitochondria in the presence or absence of a membrane potential and of proteinase K. Cases where import was accompanied by removal of the signal peptide are marked as ‘‘SP-processing'' (+). Su9(1–69)DHFR and AAC serve as positive controls for a processed matrix protein and a non-processed inner membrane protein, respectively. The score reflects the likelihood of mitochondrial localization for tested candidates as predicted by the linear classifier. MP, membrane potential; PK, proteinase K; SP, signal peptide.

Cumulative frequency of the fitness scores of single gene deletion mutants on NF, green line, and F, black line, conditions are plotted for each functional module. Genes annotated in SGD as inviable were assigned a fitness value of zero. As an example, the graph depicted in the legend shows that close to half of the genes are essential (lower half of y-axis), the other half differ in their quantitative deletion phenotypes and range from poor fitness (severe growth defect) to high fitness (no growth defect). On the x-axis, the offset of the curves of different color shows that the distribution of growth defects of the non-essential genes differs between the two growth conditions, and in this case shows that more severe defects were observed under NF than F conditions. This designates a mitochondrial specific phenotype. The one-sided paired Wilcoxon rank test was used to assess whether the deletion mutant strains for genes of a module have a significantly impaired fitness under NF versus F conditions. Significance is indicated by star (p < 0.05). In the modules, node color indicates the mRNA expression level difference between growth on NF (YPL, SCL, and YPE) versus F (YPD and SCD) conditions (red gradient, higher expression; and blue gradient, reduced expression during NF growth). A stringent criterion was applied to calculate a single expression value for the difference between NF and F growth: genes with a consistent direction of the three ratios were assigned the least-fold difference; genes that show no differential regulation or show fold ratios opposite in direction between any two conditions were assigned a fold ratio of zero. Gray nodes had no expression measurements. The identity of single proteins can be found in Figure S6. cf, cumulative frequency.

(A) Five modules enriched in human disease gene orthologs. Node color identifies human orthologs to yeast genes with and without associated Mendelian diseases (OMIM database). Proteins that belong to physical complexes are shown by overlapping nodes or in some cases are connected by solid lines; functional associations are shown by dotted lines. Disease genes within the same functional module had a tendency to have similar clinical phenotypes: glutaricaciduria II (NAD metabolism/tricarboxylic acid cycle), glycine encephalopathy (folate and glycine metabolism), mainly susceptibility to hereditary pheochromocytoma and paraganglioma (RCC-II), and variants of inherited disease of porphyrin metabolism (heme biosynthesis). Table S6 contains descriptions for each disease gene.
(B) Conservation of disease genes to proteobacteria. Venn diagram of the overlap of yeast genes having human orthologs, proteobacterial orthologs, and human disease gene orthologs. Of the proteins with a disease ortholog, 31% (31/99) have a proteobacterial ortholog—whereas only 18% (100/565) of all human orthologs have a proteobacterial ortholog (hypergeometric test, p < 10⁻⁴).
(C) Fitness distribution of yeast orthologs of disease genes. Empirical cumulative distributions of NF fitness scores are shown for all human disease gene orthologs compared to the remaining human orthologs. NF fitness represents the growth and survival of single gene deletion mutants under respiratory conditions (NF carbon source): higher fitness represents a milder growth defect. As a group, disease orthologs have higher fitness (i.e., milder mutant phenotypes) (t-test, p < 0.01).