According to the widely-read “State of AI Report, work done by Helmholtz AI scientific director Fabian Theis’ lab has potential to transform drug discovery and healthcare.
On the fast-lane to discovering effective drug combinations
Which AI technologies are going to shape future research directions and re-frame the way research is carried out? What AI developments can be considered the most interesting ones of 2021? The work resulting from the collaboration between Helmholtz Zentrum Muenchen and Facebook AI has been highlighted in the 2021 edition of “State of AI Report” as one that fits the bill.
Working with Facebook AI, the team from Fabian Theis’s lab introduces a new AI-based method that will help accelerate discovery of effective new drug combinations. The use of drug combinations is crucial in treating complex diseases such as malignant tumors. Rather than using one drug, the treatment options involve the use of a “drug cocktail.” And one challenging aspect is finding new effective drug treatments because there are nearly infinite ways of combining the drugs!
Here is where AI comes in. The team built the first single AI model that predicts the effects of drug combinations, dosages, timing, and even other types of interventions, such as gene knockout or deletion. The result is an open-sourced model called Compositional Perturbation Autoencoder (CPA) and an easy-to-use API and Python package. The paper with the results is available to the research community as a preprint on bioRxiv.
More about the State of AI Report:
The State of AI Report analyses the most interesting developments in AI. In it, the authors “aim to trigger an informed conversation about the state of AI and its implication for the future.” The authoritative research and industry report is produced by AI investors Nathan Benaich and Ian Hogarth. Information about the work done by the Helmholtz Zentrum Muenchen and Facebook AI is available on slide 25. The report is available for download.
More information on the study:
AI predicts effective drug combinations to fight complex diseases faster
Compositional perturbation autoencoder for single-cell response modeling