Projects
Project Summary and Goals
We are designing machine learning algorithms to acquire probabilistic models of metabolic and signaling networks in cancer by integrating multiple sources of data. These include flow cytometry measurements of multiple phosphorylated protein and phospholipid components in cells, SELDI-ToF proteomic data, as well as mRNA expression analysis through microarrays. We are designing new computational methods for learning quantitative models of cell signaling and metabolic activity from these data sources. Our goal is to not only to learn network models of metabolism and signaling in cells, but also to identify the regulatory components that control them.
We have computationally identified key changes in the glutathione pathway in prostate cancer cells as well as interaction between the glutathione and urea pathways which explains the overexpression of putrescene in cancer cells. We have reconstructed the T-cell signaling pathway from flow cytometry data of Sachs et. al. and found new crosstalk mechanisms, several of which have been experimentally validated in the recent literature. Using SELDI-ToF data, we have been able to identify key biomarkers that help in accurate differential diagnosis of colorectal cancer from other bowel diseases.
Our ultimate goal is to use patient-specific microarray, proteomic and genomic information to robustly reconstruct key metabolic and signaling pathways from data. Identifying malfunctions in regulatory networks is a crucial first step to developing a system-level, patient-specific model of cancer. The next phase of our work is experimental validation of the computed network perturbations. With new data gathered from experiments, we will further refine our computational models, which we hope will lead to effective targeted therapies for cancer.
Our sponsors
We are supported by a grant from the The Gulf Coast Consortium for Computational Cancer Research funded the John & Ann Doerr Fund for Computational Biomedicine.
