WormPhenotypeSimilarity

Figures

Figure 1. Comparison of metric precisions

1. Using GO as the gold standard Fig_PrecRecGo-4.doc
2. Using Kegg as the gold standard Fig_PrecRecKegg-4.doc
3. Using Co-expression as the gold standard Fig_PrecRecExpression-4.doc
4. Using PPIs as the gold standard Fig_PrecRecP2P-4.doc

Figure 2. Overview of networks and modules

1. SuperImposed? IDF/uncCorr network Fig_Networks-3.doc
2. Comparison of number of shared phenotypes per linked gene pair Fig_NumPhenosPerLink-2.doc
3. Same figure, restricted to gene pairs that share EMB, the highest frequency phenotype (53% of genes have it.) This underscores the problem of Unc Corr linking genes that share few common phenotypes Fig_NumPhenosPerLink_EMB-2.doc
4. High level view of modules identified using each metric Fig_modules_with_phenos.doc
5. Comparison of number of enriched phenotypes per module Fig_ModulePhenoEnrichment.doc

Figure 3. Utility of cophenotype data in functional prediction

• Precision/Recall plots of ClueGene results on GoProcess100? categories using:

1. Modules identified in Coexpression and PPI Networks
2. Modules identified in Coexpression, PPI, and IDF Networks
3. Modules identified in Coexpression, PPI, and uncCorr Networks

Tables

Table 1. Tested metrics

• Name of metric, formula, and reference

Table 2. Knockout Phenotypes

• Abbreviation of phenotype, explanation, frequency, and references

Table 3. Modules identified in IDF and uncCorr networks

• Coherence score (average pairwise score using that metric), coherence score of other metric, enriched phenotypes, support in GO (still need to add Coexpression and PPI coherence scores)

1. IDF modules Tab_Modules-Idf.xls
2. uncCorr modules Tab_Modules-UncCorr.xls