Dominating balanced protein interaction networks in cancer / Rianna Patricia S. Cruz [and three others].

As available proteomic data grows, so does our need for computational methods to process such data for practical applications — such as drug and therapeutic development. This is critical particularly in cancer treatments, where multiple mutations may obscure driver proteins and pathways to target for potential treatments.
To identify these driver proteins and pathways, we explore cancer networks’ minimum connected dominating sets (MCDS), a set
of topologically significant nodes of a network. We build on existing heuristic algorithms to find driver proteins of selected cancer
networks via their MCDS.
From sets of known cancer driver proteins (𝑛 = [8, 10]) and essential proteins (𝑛 = [991, 1415]) of breast, ovarian, and pancreatic
cancer, we generated protein interaction networks for each selected cancer, using balanced and directed graphs to model regulatory function.
We identified each interaction networks’ driver proteins (𝑛 = [40, 100]) from their MCDS and validated each against sets of posi-
tive control driver proteins derived by other methods. From these driver protein sets, we performed pathway analysis to identify pathways enriched by these proteins. We then verified whether these proteins had a documented association with cancer.
Our driver proteins had measures of centrality (betweenness, degree centrality) higher than those of positive control proteins of the same cancer networks. This confirms their topological significance in their respective networks.
Pathway analysis identified over 300 pathways enriched with statistical significance. A survey on these pathways found that
79 − 80% of these pathways are linked to cancer. They were also almost twice as likely to have a documented association with cancer than those not enriched by our identified driver proteins.
We not only identify specific potential driver proteins in cancer networks but also validate the potential of minimum connected
dominating set-finding algorithms to identify driver proteins in protein regulatory networks. We validate the potential of balanced
signed directed graphs in modeling regulatory functions of protein interaction networks.