Figure 1.
Strategy for constructing tissue-specific networks and predicting phenotype-associated genes.
Diverse functional genomic datasets such as expression, protein-protein interactions and phenotype information were integrated in a Bayesian framework to generate tissue-specific networks. Input datasets were probabilistically “weighted” based on how informative they were in reflecting known co-functional proteins that are both expressed in a given tissue. To account for overlap in information in multiple datasets (especially the large number of gene expression microarray datasets), mutual information-based regularization was used to down-weight datasets showing significant overlap with each other. These networks were then used as input into a Support Vector Machine classifier to predict phenotype related genes. Finally, we implemented a web interface that allows network comparison between tissues.
Figure 2.
Tissue-specific networks are more accurate than the global network in reflecting protein functional relationships.
A. 107 tissues were grouped into major body systems according to the anatomical hierarchical structure maintained in GXD [20]. Through three-fold cross-validation, the performance of tissue-specific networks was compared against the global network and the percentage improvement of tissue-specific networks over the global network was plotted. All tissue-specific networks out-performed the global network in this cross-validation analysis. Improvements were consistent across tissues belonging to all major organ systems. Candle-stick plots (minimum, 25%, median, 75% and maximum) represent the distribution of percentage AUC improvement for all tissues in a specific system. B. Example precision recall curves of tissue-specific and the global network, generated using three-fold cross-validation. Across the entire precision-recall space, tissue-specific networks performed better than the global network. Complete precision-recall figures for all networks are included in Dataset S2.
Figure 3.
Top connected genes of Wnt10b in muscle-specific and bone-specific networks.
In A, blue-highlighted genes are directly involved in skeletal muscle development. In B, blue-highlighted genes are involved in bone minerization or bone structure formation. The enrichment of genes involved in the above processes reflects the differential roles of Wnt10b in skeletal muscle and bone.
Table 1.
Example enriched Gene Ontology terms in the tissue MA:0000016 nervous system.
Figure 4.
Tissue-specific networks perform better than the global network in predicting genes related to different phenotypes.
By mapping phenotypes to different tissues according to their terminology and description, we are able to compare the performance of tissue-specific networks and the global network in predicting phenotype-related genes. Candle-stick plots (minimum, 25%, median, 75% and maximum) show the distribution of percentage AUC improvement when predicting phenotype-related genes. A. Phenotypes were grouped according to the number of annotated genes. Tissue-specific functional networks show consistent improvement across different phenotype sizes. B. Phenotypes were grouped according to major organ systems of their corresponding tissue. Improvements were consistent across all major systems. C. Example precision-recall curves for “abnormal osteogenesis” (MP:0000057), “abnormal nervous system electrophysiology” (MP:0002272), “abnormal spleen white pulp morphology” (MP:0002357), and “abnormal CNS glial cell morphology” (MP:0003634) using both tissue-specific networks (shown in red) and global networks (shown in green). For phenotypes such as these, tissue-specific networks are necessary to make accurate predictions.
Figure 5.
Prediction and verification of infertility-related genes through male reproductive system-specific networks.
A. Local functional relationship network of the gene Mybl1 in the male reproductive system. The top 18 genes connected to the query set with connection weights higher than 0.634 are displayed. These top functionally related proteins include well characterized male infertility genes such as Dmc1, Ddx4, and Cyct. B. Histological cross-sections of oval seminiferous tubules show that wild type (Mybl1+/+) testis tubules contain many developing germ cells, while mutant (Mybl1repro9/repro9) testis tubules contain many fewer germ cells and more empty space, indicative of infertility.
Figure 6.
Top connected genes to Atcay in the cerebellum-specific network reveals likely ataxia candidates.
Edges with weight greater than 0.9 are shown. In the cerebellum network (A), Grm1 and Cacn1a are the top predicted connections to Atcay, with confidences of 0.902 and 0.943, respectively. Both genes are closely connected to Atcay and its top 10 neighbors. In the global network (B), Grm1 and Cacn1a are much more weakly connected to Atcay (0.763 and 0.647, respectively), and are not identified as top connectors to Atcay. Grm1 and Cacn1a are not connected to Atcay or any of its top 10 neighbors in the global network.
Table 2.
Evidence for top 10 predictions for ataxia-causing genes using mouse cerebellum-specific networks.