UCSF

A complex network of molecular interactions underpins cellular physiology, with each interaction contributing to the cell’s overall function. In normal physiological states, these networks are tightly regulated, but in disease, their structure and dynamics can shift, leading to dysregulations and pathogenesis. Predicting the structure of disease-relevant networks has the potential to enhance therapeutic target identifications, improve disease prognosis predictions, and refine models of complex molecular systems. In the first half of this work we develop, implement, benchmark, and illustrate Integrative Network Modeling, an algorithm for modeling disease relevant protein interaction networks based on affinity purification mass spectrometry (AP-MS) data. We find AP-MS experiments contain more information about a protein’s direct protein interactions than previously thought. In the second half of this work, we predict the presence of protein interactions in the epidermal growth factor receptor (EGFR) molecular neighborhood using proximity labeling mass spectrometry. We apply a deep-learning model to predict the three-dimensional structure of EGFR binary complexes; we identify multiple proteins in complex with EGFR. The computational methods developed and applied in these studies are aimed at modeling complex molecular systems based on the integration of information from mass spectrometry and protein structure. Together, they are a step towards bridging the gap between structural and systems biology.