Researchers from the Luddy School of Informatics, Computing, and Engineering, Instituto Gulbenkian de Ciência, and Northeastern University have developed a novel mathematical framework that increases causal explainability and control of biochemical systems, including those involved in disease.
In a paper featured on the cover of the journal Proceedings of the National Academy of Sciences (PNAS), Professor of Informatics Luis Rocha and Alexander Gates, who earned his Ph.D. in complex networks and systems from the Luddy School, introduce the effective graph to capture the nonlinear logical redundancy present in biochemical network regulation, signaling, and control. Together with Rion Correia, who also earned his Ph.D. in CNS from the Luddy School, and Xuan Wang, a current student of the same program, they demonstrate the utility of the approach with computational models of human cancer cells, showing that the effective graph reveals why some cancer medications are more effective than others in killing breast cancer cells.
“The basic idea is very simple: some interactions are much more important than others,” said Gates, who is now part of the Center for Complex Networks Research at Northeastern University. “Given a biological model, we identify genetic ‘master switches’ which render many of the other known interactions redundant. This allows us to identify prominent pathways that effectively control biochemical regulation and signaling. A surprising consequence is that these pathways can change in the presence of drugs. In other words, drugs can change which molecular interactions act as ‘master switches’.”
Gates and Rocha are lead authors for the paper.
“We have analyzed a large battery of biochemical regulation and signaling networks, some involved in human disease, others in well-known biological processes across organisms,” Rocha said. “We have learned that in all these systems there is an unexpectedly large number of redundant interactions. In short, this means that although a certain cellular component, say a protein, may be potentially regulated by a large number of other components, in the dynamics of the cell it is actually only regulated by a very small number of other molecular products. This massive redundancy is likely in place to make biological organisms resilient to random perturbations while facilitating phenotypic regulation .”
The discovery of the redundancy can be used to find the few components that are actually needed to control the cell. The paper exemplifies this with a large signaling network of genes, proteins, and other components involved in breast cancer.
“Our ultimate goal is to be able to identify genes and other molecular elements that can revert a diseased cell, especially cancer cells, to a healthy state,” Rocha said.“ At the Center for Social and Biomedical Complexity, we have been working on using complex networks and systems methods to design and study systems biology models, focusing on understanding how to control them. This agenda includes developing artificial intelligence, machine learning, dynamical systems, and network science methods that can be used at all levels of this process: from inferring networks of biochemical regulation and signaling from experimental data to synthesizing such data into networks and their subsequent analysis.”
Rocha and his colleagues used 78 experimentally validated models derived from systems biology to demonstrate that redundant pathways are prevalent in biological models of biochemical regulation, that the effective graph provides a probabilistic but precise characterization of multivariate dynamics in a causal graph form, and it provides an accurate explanation of how dynamical perturbation and control signals, such as those induced by cancer drug therapies, propagate in biochemical pathways. The process has the potential to be used in a wide variety of other applications.
“From the complex systems viewpoint, we want to expand the range of models the methodology is applicable to, namely by collaborating with specialists to apply it to epidemiolocal, ecological, eco-evolutionary, and even brain networks,” Rocha said. “From the biomedical complexity viewpoint, we will work to include the methodology in translational workflows whereby we can infer personalized networks that can be analyzed for specific disease cohorts and even individual patients.”
The research in the PNAS study has built upon the foundation laid by previous projects funded by the National Institutes of Health, the National Science Foundation, the Fulbright Scholars Program, and other entities.
“Beyond its scientific importance, this work reflects the winding/twisting paths that research takes,” Gates said. “The core ideas were originally conceived as a secondary idea in my Ph.D. thesis project at IU with Luis. It took 6-7 years and interactions with great collaborators for us to formalize the technical details and think through the implications of the framework, during which the project transformed several times. The result is a tour-de-force that I am particularly proud of. It’s a combination of novel mathematics, applied statistical insights, and a detailed biological case-study.”
“Collaboration is the backbone of research at the Luddy School, and Luis and his colleagues have shown what can be accomplished with expertise at multiple levels,” said Kay Connelly, the associate dean for research at Luddy. “Their developments have the potential to create important breakthroughs in treatment, and they open new pathways for advancement.”