Skip to main content

Personalized structural biology to guide cancer treatment

Tony Capra, Ph.D., Discovery Grant 2016

Motivation and Relevance:

Our bodies are made up of tens of trillions of cells that come in hundreds of different forms—from bone cells to brain cells—each carrying out its own unique function. These cells exist in an exquisite balance, listening to one another and sharing resources to keep us alive. Orchestrating these interactions requires a complex set of instructions that are encoded in each cell—our genome. Cancer starts when a cell’s instructions get mixed up, or mutated, in a way that makes it stop cooperating with its neighbors and pursue its own growth at the expense of the cells around it. Unfortunately, there are thousands of ways these instructions can get scrambled and ultimately lead to cancer.

As a result, cancer is not a single disease, but a constellation of diseases each with its own evolutionary history and arsenal of defenses. Unfortunately, treatments that work on one patient’s cancer often are ineffective on others, even of the same tissue. Dramatic technological advances are making it possible to identify how our genome’s instructions have been changed in individual patient’s tumors. However, a major roadblock to using this information in treatment decisions is the difficulty of identifying which mutations caused the cancer and how they will influence response to different treatments.


Our goal is to develop computer programs that analyze the mutated instructions found in patient tumors to predict which changes caused the cancer and how the tumor will respond to different treatments. This method will be applicable to any type of cancer, but as proof of concept, we will evaluate and refine it on breast cancers. Ultimately, we have the audacious goal of applying our program to every tumor genome in the world.


Plan and Interpretation:

Our approach will use computer algorithms to integrate diverse publicly available information about mutations previously seen in cancers with models of how proteins—the molecular machines encoded by our genes—function. We will create personalized 3D models of proteins of interest and then simulate the effects of the changes to that protein found in their tumor. In many cases, tumors’ genomes are disrupted in ways that are similar to cancers for which there are effective treatments. However, in others with novel changes, we will model response to a known treatment, such as a drug that interacts with the mutated protein. Computational predictions are of little value without experimental validation. In collaboration with Dr. Carlos Arteaga, we will refine our program’s predictions on patients with mutations in the gene ErbB2, which is commonly amplified in breast cancer and is targeted by several drugs.


Benchmarks of Success:

Our first benchmark will be the development of an efficient, accurate program for evaluating tumor mutations. Our second benchmark will be the successful validation of the program’s predictions on ErbB2 breast cancer mutations. Meeting these goals will result in sufficient preliminary data to submit a proposal to the NIH to support the expansion of this work.


Budget and Timeline:

This grant will enable my postdoctoral scholar to devote six months of committed effort to the development and refinement of the flexible computational pipeline. It will also enable us to collaborate to over the following three to four months to iteratively refine and validate our method with experiments in breast cancer cells.