Project Quine
An AI system designed for biological research, built on a two-layer architecture comprising specialized biology models and a scheduling harness. The system supports joint reasoning across multiple biological levels and modalities, assisting scientists in large-scale drug screening and experimental design.
Timeline (last 90 days)
Project Quine consists of two core components: a model that operates across multiple levels and modalities in biology, and a scheduling harness built on top of that model.
Quine's scheduling harness draws on academic literature on one hand and invokes the capabilities of the model being trained on the other, forming an iterative back-and-forth interaction with scientists.
Project Quine can compress multiple levels of biology together for joint reasoning, simultaneously incorporating protein sequences, RNA, imaging data, and patient-related environmental features.
The Project Quine team collaborates with institutions such as the Broad Institute, bringing together oncologists, biologists, and computational biologists.
The Project Quine team applied the system to drug repurposing research for pancreatic cancer, searching for potential efficacy among already-approved known compounds.
The candidate drug space for pancreatic cancer drug repurposing numbers as many as seven to eight thousand compounds, and Quine narrows this space down to ten to twenty candidates most worthy of laboratory validation.
The team obtained a preliminary drug nomination list from a single weekend of model runs.
The team placed the drugs nominated by the Quine system into wet lab experiments for validation, confirming that these drugs produce the expected efficacy shifts.
Project Quine can simulate the trajectory of experiments, predict the likely outcome of each experiment and subsequent responses, and design the experiment with the maximum information gain based on all possible trajectories.
All Facts (9)
Project Quine consists of two core components: a model that operates across multiple levels and modalities in biology, and a scheduling harness built on top of that model.
50%UnverifiedQuine's scheduling harness draws on academic literature on one hand and invokes the capabilities of the model being trained on the other, forming an iterative back-and-forth interaction with scientists.
50%UnverifiedProject Quine can compress multiple levels of biology together for joint reasoning, simultaneously incorporating protein sequences, RNA, imaging data, and patient-related environmental features.
50%UnverifiedThe Project Quine team collaborates with institutions such as the Broad Institute, bringing together oncologists, biologists, and computational biologists.
50%UnverifiedThe candidate drug space for pancreatic cancer drug repurposing numbers as many as seven to eight thousand compounds, and Quine narrows this space down to ten to twenty candidates most worthy of laboratory validation.
50%UnverifiedProject Quine can simulate the trajectory of experiments, predict the likely outcome of each experiment and subsequent responses, and design the experiment with the maximum information gain based on all possible trajectories.
50%UnverifiedThe team obtained a preliminary drug nomination list from a single weekend of model runs.
50%UnverifiedThe Project Quine team applied the system to drug repurposing research for pancreatic cancer, searching for potential efficacy among already-approved known compounds.
50%UnverifiedThe team placed the drugs nominated by the Quine system into wet lab experiments for validation, confirming that these drugs produce the expected efficacy shifts.
50%