Autonomous phenomics for compound profiling
The theoretical experimental space of biomedical research is vast. Even when restricting it to e.g. find optimal combinations and concentrations of drugs for treatment, it becomes infeasible to search exhaustively using brute force. As an example, evaluating all combinations of 3 drugs from a panel of 500 would require almost 24 M experiments. This constitutes a search space that is infeasible to conquer using conventional methods where experiments and experimental design are carried out by humans. To overcome this challenge, we need intelligent methods for selecting the most informative experiments, integrated with automated lab equipment to perform them.
The purpose of this project is to improve studies of mechanisms and pathways for compounds, such as drug or environmental contaminants, with intelligent data generation, automation and AI. To this end we will develop an intelligent system for phenotypic cell profiling (phenomics) that is able to autonomously suggest the next most informative experiment, perform it using an automated lab, learn from the results, and iterate.
Overview of the iterative, AI-controlled process. Based on a scientific question or hypothesis to test, the AI makes predictions on existing data (external data integrated with in-house data) and designs new experiments to improve the hypothesis testing. New data is obtained, models are improved, and the process is iterated until desired confidence reached or termination criteria fulfilled. Complementing the system is human-controlled confirmatory or exploratory experiments. The produced images, data and models will all be made available online.
AI-driven, autonomous profiling of cells has the potential to radically speed up biological discoveries, offers a means for detailed understanding of cell states, and allows for testing and optimization of previously impossible treatment strategies.
Our key applications of autonomous cell-based phenomics will be in compound profiling for drug safety, for assessment of environmental toxicants, and for phenotypic screening of drugs. We will target both individual compounds and combinations of compounds, as well as multiple cell lines, primary cells, and co-cultures.
The project is partially funded by group leader Ola Spjuth receiving a 3 MSEK grant from the FORMAS Future Research Leaders call in 2018.