Photo credit Robert Hood / Fred Hutch News Service


BBI Faculty Conversations. Check back in soon for more chats with BBI members from our partner institutions. Get an inside view of their work and where they think the field of precision medicine is heading. Today we are joined by BBI member Dr. Taran Gujral, Assistant Professor in the Human Biology Division at the Fred Hutchinson Cancer Research Center.


BBI:  To start, could you tell us a little bit about your background and the focus of your work?

Dr. Gujral: I had a very broad education and training in mathematics, cell biology, biochemistry, and computational biology. This diverse training inspired what I try to create in my lab – bringing in minds from various backgrounds. We have a bioinformatics specialist, a computational biologist to focus on machine learning, mathematician, an immunologist, a virologist, a pharmacologist, a biochemist, and a cancer biologist. All working together!

I try to bring the expertise from these different fields to solve complex questions in cancer biology. We work to understand how cancer cells respond to various signals. These signals could be from the cell microenvironment, from other cell types, from small proteins or chemicals, or signals, such as those coming from potential therapies or antibodies. We monitor what these signals tell a cancer cell to do – i.e. proliferate, die, or move and metastasize. Our overall goal is to understand and model how cancer cells make these decisions.


BBI: This sounds exciting. How do you bring your team together to work on precision medicine?

Dr. Gujral: Overall, the drug development process is tedious and slow, especially in cancer. It takes decades to get a drug from the lab to patients. Some of the reason for this is that the drug discovery process itself is very slow. First, a potentially druggable target or domain has to be identified. In the case of cancer, this target could be an oncogene. Second, it requires screening a large library of hundreds to millions of compounds or biologics against that target. Through various stages of optimization and screening, you may identify a set of possible compounds. All this has to happen before a potential drug is tested in a living human being.

As I oversee a small lab, I think about this inefficient process a lot, and ask myself how we can contribute to this complex process. Obviously, we cannot do better than big pharma. We do not have a million-compound library and the capability to do such screening. Instead, our angle is to take another look at already existing drugs as opposed to novel compounds. These are chemicals that have already passed some level of safety screening, but failed efficacy for a particular disease, or are already in use for another disease but may be repurposed.

Finally, we look at combinations of these compounds instead of seeing them as a single agent.

Cancer cells are very smart. You block one pathway, and they activate a different one to survive. The ideal therapy would likely be a combination of drugs to block as many survival pathways at once to kill cancer cells and prevent the emergence of resistance. So, our computational models are designed to predict drug combinations that can be tested in the lab.


BBI: As a smaller lab, how do you competitively test these compounds?

Dr. Gujral: This is where our lab has built a machine learning-based screening platform. Unlike big private labs, we don’t screen hundreds of thousands of compounds. We screen with a small set of 25 to 30. Any small lab can do this. With this small number of compounds, we have a machine learning model that can predict the behavior of the other hundreds of thousands of compounds against a target. The model can take days to weeks to complete as it runs possible computations. We rely on the power of computation to tell us what drugs and combinations might be effective or ineffective, which we can then test experimentally to verify.


BBI: Fascinating. Can you give us an example where you have put the model to work?

Dr. Gujral: We have a project funded by the BBI to study fibrolamellar cancer (FLC). Usually cancer of the liver is caused by injury, or often alcohol, or a virus, or a combination of environmental factors. In contrast, FLC occurs mostly in children without the underlying environmental exposure. It is a very rare form of cancer occurring in approximately 1 to 5 patients per year in Seattle. A few years ago, a consortium of scientists led a study finding that the key driver of FLC in most children had a fusion of two genes, PKA and DNAJB1. These genes are difficult to target in cancer cells because they have large roles to play in normal cell function as well.

So, the search continued for a druggable target for FLC. Working with our surgical colleagues Drs. Ray Yeung and Kim Riehle, we received tumor samples directly from affected patients. We deployed our 25-compound machine learning system against the tumor pieces and identified the gene PLK1 that is important for growth of FLC cells, but not normal liver cells. Because of the model, we identified a drug that had failed efficacy against rheumatoid arthritis in clinical trials, but had passed safety trials for use in humans. It showed efficacy against FLC cells in our model. We tested the drug against tumor cells and in a mouse model, and the drug slowed FLC tumor growth substantially.

This is an excellent story about how precision medicine can work. We came in with little knowledge about FLC other than what a key gene driver of the cancer seemed to be. We worked with a colleague in a completely different field to apply our model to tissue samples. It is an exciting example of the power of precision medicine. BBI helped formalize the partnership and we’re in the process of publishing this work now.


 [The Fred Hutch recently covered Dr. Gujral’s AI platform and how it helps enable finding new drugs to treat COVID. Take a look at link]