A new machine learning model can screen novel drug compounds to accurately predict efficacy in humans

A new machine learning model can screen novel drug compounds to accurately predict efficacy in humans

The journey between the identification of a potential therapeutic agent and the approval of a new drug by the Food and Drug Administration can take well over a decade and cost more than a billion dollars. A research team at the CUNY Graduate Center has developed an artificial intelligence model that could significantly improve the accuracy and reduce the time and cost of the drug development process. The new model, called CODE-AE, described in a newly published paper in Nature Machine Intelligence, can screen novel drug compounds to accurately predict efficacy in humans. In testing, it was also able to theoretically identify personalized medications for over 9,000 patients that could better treat their conditions. The researchers expect the technique to significantly accelerate drug discovery and precision medicine.

Accurate and reliable prediction of patient-specific responses to a new chemical compound is crucial for discovering safe and effective therapeutics and selecting an existing drug for a particular patient. However, it is unethical and unfeasible to conduct early efficacy testing of a drug directly on humans. Cell or tissue models are often used as a surrogate for the human body to evaluate the therapeutic effect of a drug molecule. Unfortunately, drug action in a disease model often does not correlate with drug efficacy and toxicity in human patients. This knowledge gap is a major factor in the high costs and low productivity rates of drug discovery.

Our new machine learning model can address the translational challenge from disease models to humans. CODE-AE uses a design inspired by biology and leverages several recent advances in machine learning. For example, one of its components uses similar techniques when creating deepfake images.”

Lei Xie, Professor of Computer Science, Biology and Biochemistry, CUNY Graduate Center and senior author of the Hunter College and Paper

Drug Discovery E-Book

Compilation of the top interviews, articles and news from the last year. Download a copy today

The new model can get around the problem of having enough patient data to train a generalized machine learning model, said You Wu, Ph.D. of the CUNY Graduate Center. Student and co-author of the work. “Although many methods have been developed to use cell line screens to predict clinical responses, their performances are unreliable due to data inconsistency and discrepancies,” Wu said. “CODE-AE can extract intrinsic biological signals masked by noise and confounding factors, effectively alleviating the problem of data discrepancy.”

As a result, CODE-AE improves the accuracy and robustness over state-of-the-art methods in predicting patient-specific drug responses solely from cell line junction screens.

The research team's next challenge in advancing the use of the technology in drug discovery is to develop a way for CODE-AE to reliably predict the effect of the concentration and metabolism of a new drug in the human body. The researchers also noted that the AI ​​model could potentially be optimized to accurately predict human side effects of medications.

This work was supported by the National Institute of General Medical Sciences and the National Institute on Aging.

Source:

The Graduate Center, CUNY

Reference:

Er, D., et al. (2022) A context-aware disentangling autoencoder for robust prediction of personalized clinical drug response from cell line compound screening. Nature-machine intelligence. doi.org/10.1038/s42256-022-00541-0.

.