Utility and Limitations of Using Gene Expression Data to Identify Functional Associations
Fig 2
Relationship between pathway ECs, annotation quality and similarity measures.
(A) Relationship between the EC calculated for pathway genes that are annotated based on experimental evidence (ECexp) and EC calculated for pathway genes that are annotated only based on computational evidence (ECcomp). The genes used to calculate ECexp and ECcomp do not overlap. Each dot represents one pathway. Dashed line: y = x line. (B) Heatmap of correlations between pathway EC percentiles calculated with: partial correlations estimated with the corpcor method, Spearman’s rank correlation coefficient (Spearman), Pearson Correlation Coefficient (PCC), adjusted and normalized Mutual Information (MI), partial correlation calculated with the partialcorr method, and transformed p-values of Bayesian Network (BN) (C) Percent pathways that have high EC using different similarity measures. (D) Heatmap of pathway EC percentiles calculated using different similarity measures. Color represents EC percentiles. White dotted rectangles: high EC pathways that are specific to one measure.