As a computer scientist, I am interested in using computational methods to understand biological systems and other complex systems. My research lies at the intersection of systems biology, bioinformatics and artificial intelligence. I have been an active performer in several DARPA programs including Big Mechanism, Communicating with Computers, and World Modelers.
I have been working on the following topics in recent years:
Automated assembly of mechanistic executable models to explain experimental observations, focusing on cancer signaling. I co-developed the Integrated Network and Dynamical Reasoning Assembler (INDRA) system which assembles information about biochemical mechanisms automatically extracted from the scientific literature into various explanatory and predictive models. We used INDRA to rapidly prototype models of molecular mechanisms directly from English language descriptions written by experts, as described in our paper "From word models to executable models of signaling networks using automated assembly". See also the News & Views piece "Natural language processing: put your model where your mouth is" by Haggerty and Purvis discussing our approach and results.
INDRA is also used to assemble (i.e. correct errors, resolve redundancies, assess reliability, assess relevance, infer missing information) millions of assertions about molecular mechanisms collected by reading the scientific literature (with natural language processing systems such as REACH, TRIPS and Sparser) into executable models that can be used to construct explanations to experimentally observed drug effects. This research is funded under the DARPA Big Mechanism program.
Development of a human-machine communication system for model-driven discovery in molecular biology. I am leading the development of the Bob with Bioagents system (with collaborators at IHMC, SIFT, OHSU and Tufts), and integrated dialogue system with a web-based interface where a human user can talk with a machine partner to learn about molecular mechanisms (e.g. drugs/target relationships, specific mechanisms by which proteins regulate each other, transcription factor/target relationships), and, via dialogue, construct a model of a hypothesis to explain an experimental observation (for instance "What could explain the observation that the amount of FOS decreased when treating colorectal cancer cells with Trametinib?". The automatically assembled model can also be interrogated via dialogue. This research is funded under the DARPA Communicating with Computers program.
Assembly of "World Models" that integrate causal interactions from a variety of domains beyond biology. I am leading the development of INDRA-GEM (INDRA for Generalized Ensemble Modeling), a generalization and extension of the INDRA system to collect and assemble causal interactions in a domain-independent fashion into predictive and explanatory models that can be used to understand global and regional crisis conditions such as food insecurity. This research is funded under the DARPA World Modelers program.
- Modeling uncertainty in dynamical systems: I am interested in the statistics and computation required to represent and reason about uncertainty is models of biological systems, and dynamical systems in general. My PhD thesis explored this topic in detail.