Our research spans many biomedical application areas, and a common theme is using biological networks to connect diverse data and provide a cohesive view of a process. We develop new computational methods and also work with collaborators to apply them to study specific conditions and diseases. We also specialize in analyzing time series datasets where gene expression or protein activity is tracked over time, as described in this Morgridge Institute story. As part of the John W. and Jeanne M. Rowe Center for Research in Virology at the Morgridge Institute, we are particularly interested in applications in viral infection and virus-induced cancers. Representative examples of our network biology work appear below.

In addition, we create machine learning methods to guide biochemistry experiments, especially for protein engineering and drug discovery. Example studies include Gelman et al. 2021 and Alnammi et al. 2023.

Phosphorylation timing

TPS algorithm Figure from Köksal et al. 2018

The timing of events within cells can teach us how proteins pass messages to respond to changes in the cellular environment. Our Temporal Pathway Synthesizer (TPS) algorithm evaluates which possible signaling pathways could have generated an observed phosphorylation response. It uses constraints such as the timing of phosphorylation changes to rule out pathways in which late-responding proteins pass messages to (phosphorylate) early-responding proteins.

Pathways disrupted by cancer

Multi-PCSF algorithm Figure created by Tobias Ehrenberger in Ernest Fraenkel's lab

Cancer is primarily caused by genomic alterations such as DNA mutations, and when specific genes are frequently mutated in many tumors it is well-understood that they may be linked to tumor formation in some manner. However, many tumors do not contain any of these frequent alterations. Our group uses network algorithms like the multi-sample prize-collecting Steiner forest (Multi-PCSF) technique to identify pathways that may be commonly perturbed by mutations in different genes. In the example above, diverse mutations from six medulloblastoma tumors are connected via protein-protein interactions through novel, unmutated genes revealing a pathway that may be disrupted in all patients. This project was a collaboration with Ernest Fraenkel, Scott Pomeroy, and Jill Mesirov.

See Anthony's Morgridge Factor guest blog post for more information.

Linking signaling and transcriptional networks

SDREM algorithm Figure from Gitter et al. 2013

When cells respond to external stresses, messages are rapidly passed through signaling networks to activate downstream transcription factors, which adjust gene expression levels. Our Signaling and Dynamic Regulatory Events Miner (SDREM) algorithm integrates the initial proteins that detect an external stress and temporal measurements of the gene expression changes that are caused by the stress with protein-protein and protein-DNA interaction networks to reconstruct the reacting pathways. We have used SDREM to study yeast stress response, like the osmotic stress example above, and the human immune response to influenza infection.