Discovering pathway cross-talks based on functional relations between pathways

Abstract

Background: In biological systems, pathways coordinate or interact with one another to achieve a complex biological process. Studying how they influence each other is essential for understanding the intricacies of a biological system. However, current methods rely on statistical tests to determine pathway relations, and may lose numerous biologically significant relations.

Results: This study proposes a method that identifies the pathway relations by measuring the functional relations between pathways based on the Gene Ontology (GO) annotations. This approach identified 4,661 pathway relations among 166 pathways from Pathway Interaction Database (PID). Using 143 pathway interactions from PID as testing data, the function-based approach (FBA) is able to identify 93% of pathway interactions, better than the existing methods based on the shared components and protein-protein interactions. Many well-known pathway cross-talks are only identified by FBA. In addition, the false positive rate of FBA is significantly lower than others via pathway co-expression analysis.

Conclusions: This function-based approach appears to be more sensitive and able to infer more biologically significant and explainable pathway relations.

Figure 1

Distributions of functional similarity (funSim) scores. (A) The distribution of functional similarity scores of the pathway pairs and random sets, respectively. Each data point represents the fraction of the pathway pairs that have a funSim more than the value on the x-axis. (B) The distribution of the best funSim scores for all pathways.

Figure 2

Overlap of predicted pathway cross-talk pairs by different methods. The different prediction methods were applied to all possible combinations of 168 pathways. FRPs are identified by the function-based approach (FBA), and SOPs and SIPs are determined by the physical entity-based approach (PEBA).

Figure 3

Comparison of co-expression values among sets of pathway pairs inferred by different methods. The distributions of coexpression values were compared among common pathway pairs, which are identified by FBA and PEBA, unique pairs by FBA, and unique pairs by PEBA, respectively.

Figure 4

Interaction between S1P4 and RHOA signaling pathways. The pathways are integrated and visualized via the Pathway Integration Tool (PINT) [36]. The oval nodes denote the proteins, and the octagon nodes denote the protein complexes. The edges between nodes represent the interactions or reactions. The green and blue nodes correspond to the components in S1P4 and ROHA signaling pathways, respectively. The yellow node is the common component between pathways.

Figure 5

Pathway cross-talk cliques. The nodes indicate the pathway, and the edges are the pathway pairs which have no common gene and are not SIPs, but are identified as highly functional relations by FBA.

Figure 6

Reproducibility analysis of function-based approach. The left figure (A) is the results of using pathway pairs from the same database, and the right figure (B) is the results of using pathway pairs cross databases. Each data point represents funSim score of a pathway pair. The dash lines represent the cut-off value (funSim = 0.5). If the points are located in the left-bottom and right-top corners, the prediction results of these pathway pairs are consistent. In contrast, the points located in the left-top and right-bottom corners indicate the inconsistent results.