Understanding the nature of signaling pathways — networks of molecules in a cell that work together to control a cell’s response to its environment — is an increasingly important part of biomedical research and helpful, for example, in enhancing our understanding of how cancer cells live or die.
A signaling pathway typically starts at one or more receptors, which are proteins on the cell’s surface that sense the presence of external signals. When such an external molecule binds to and activates a receptor, it triggers an internal cascade of reactions in the cell. Ultimately, these events activate transcription factors, which control the production of other proteins that help the cell respond to the signal.
In the last 20 years, databases have been developed to store the interactions in different signaling pathways. But these databases are incomplete and depend upon manual curation, which can be slow. Moreover, different databases may store different versions of the same signaling pathway.
In a paper titled, Pathways on demand: automated reconstruction of human signaling networks, published in Systems Biology and Applications, a Nature partner journal, T. M. Murali, professor in the Department of Computer Science at Virginia Tech, in collaboration with Shiv Kale, a research scientist at the Biocomplexity Institute of Virginia Tech present a new computational algorithm called PathLinker that automatically reconstructs signaling pathways from a background network of molecular interactions.
The algorithm starts by identifying the receptors and transcription factors (proteins that bind to DNA and help control the signaling process) in a specific pathway. It then reconstructs that pathway by finding multiple short paths in the background network that connect receptors to transcription factors.
When checked against comprehensive gold-standard sets of signaling pathways in existing, manually curated databases, PathLinker was found to be extremely accurate.
To understand how PathLinker works, Murali suggests thinking of signaling pathways as routes in Google Maps. Manual curation would be like looking at a physical map drawn on paper and identifying all the streets that are in an area such as Manhattan with your finger. Computational reconstruction is akin to automatically tracing multiple routes from one end of Manhattan to another. Other computational algorithms can find a skeleton road network that only detail a small part of the entire Manhattan map.
“Most of the previous methods focus on finding a really compact network,” said Murali. “They do not seek to find more than one way to get from the sources to the targets. With PathLinker, we make it a point to compute many different paths, in fact, as many as the user wants. The goal is not just to find the best way to get there, but to reconstruct all the alternate routes the cell might use to transfer the signal. This principle is important since cells have evolved several redundant paths to convey signals.”
Anna Ritz, a post-doctoral fellow in Murali’s group and now an assistant professor of biology at Reed College, led the PathLinker research efforts.
“An exciting and desirable outcome of PathLinker is that the ‘mistakes’ it makes may point to proteins and interactions that have not yet been assigned to a specific signaling pathway, added Ritz. “We can prioritize such proteins and interactions for experimental validation in the order in which PathLinker unveils them.”
Murali’s and Kale’s group collaborated on validating that PathLinker indeed has the power to reveal new roles for proteins.
“We set out to determine if PathLinker could aid us in our studies on Wnt Mediated Signaling” explained Kale. “Wnt signaling pathways are a group of signal pathways made of proteins passing signals into a cell through cell surface receptors. Application of PathLinker generated several novel predictions in regards to aspects of Wnt signaling. We were able to test and verify these predictions in the laboratory, which ultimately lead to new insights into the molecular mechanism of their signaling.”
As Murali wrote in the paper, “PathLinker provides a promising framework for reconstructing a well-studied signaling pathway given relatively little information about its components. It may serve as a powerful approach for discovering the structure of poorly studied processes and prioritizing both proteins and interactions for experimental study.”