Network topology is a common tool in systems biology. The focus of our work were disordered proteins whose importance in signaling have long been discussed. A good tool for studying these regulatory proteins is to study triangles i.e., three-nodal signaling motifs where each node is related to the other two, which infers a closer connection between the nodes. Our research shows that oncotherapeutical targets and disordered proteins often form three-nodal motifs together. Our aim was to analyse of these motifs for better understanding the complex relationship of these proteins with oncotherapeutical targets.
The directed edges of three networks were extracted, and motifs were identified using the FANMOD program. For network analysis, Cytoscape and the Python NetworkX package were used. Then, disordered proteins were identified using the DisProt database, and CIViCmine were used for identifying biomarkers. Then, binary classification models were developed.
We have shown that disordered proteins were more frequently in triangles and in regulatory motifs like unbalanced triangles and cycles. They form common motifs with known oncotherapeutical targets, where many disordered proteins are predictive biomarkers for the target (Human Cancer Signaling Network: 23%). Certain topological parameters are associated with predictive properties. Thus, we developed a machine learning model based on network topological data and biological characteristics of 110 target-disordered protein pairs, whose predictive properties couldbe identified from the literature . With this model we are able tocan predict potential predictive biomarker properties for the other identified 694 pairs in our networks.
We show based on network topology analysis that intrinsically disordered proteins have great potential as predictive biomarkers. Our results imply that investigation into these motifs may help to identify novel predictive biomarkers, which can be further studied individually and validated experimentally. As predictive biomarkers are vital for therapeutical decision making, developing a tool for predictive biomarker identification may have effect on the daily clinical practice.
The authors.
EFOP-3.6.3-VEKOP-16-2017-00009.
All authors have declared no conflicts of interest.