A new AI model can predict human responses to drug compounds, transforming medicine
City College of New York's new AI model will be able to predict accurately human response to novel drug compounds. Moreover, it is less costly and faster.
Published in Nature Mature Intelligence on October 17, this technique might significantly speed up precision medicine and medication development.
According to research, the new CODE-AE model can screen brand-new medication molecules and reliably forecast their effectiveness in people. In tests, it was also able to find potentially more effective tailored medications for over 9,000 patients.
To find safe and effective therapies and choose an existing medicine for a particular patient, accurate and reliable predictions of patient-specific responses to a new chemical molecule are essential.
However, directly testing a drug's early efficacy on humans is immoral and impossible. To assess a pharmacological molecule's therapeutic efficacy, cell or tissue models of the human body are frequently used, said the statement.
Unfortunately, the treatment efficacy and toxicity in actual patients frequently do not match the pharmacological impact in a disease model. This knowledge gap primarily causes the high prices and low rates of drug discovery productivity.
“Our new machine learning model can address the translational challenge from disease models to humans,” said Lei Xie, a professor of Computer Science, Biology, and Biochemistry at the CUNY Graduate Center and Hunter College and the paper’s senior author.
“CODE-AE uses biology-inspired design and takes advantage of several recent advances in machine learning. For example, one of its components uses similar techniques in Deepfake image generation.”
A new solution
The new approach may give a solution to the issue of not having enough patient data to train a broad machine-learning model.
“Although many methods have been developed to utilize cell-line screens for predicting clinical responses, their performances are unreliable due to data incongruity and discrepancies,” a CUNY Graduate Center Ph.D. student and co-author of the paper.
“CODE-AE can extract intrinsic biological signals masked by noise and confounding factors and effectively alleviated the data-discrepancy problem.”
When predicting patient-specific medication responses from cell-line compound screens, CODE-AE greatly outperforms state-of-the-art approaches in terms of accuracy and robustness.
Supported by the National Institute of General Medical Sciences and the National Institute on Aging, the study's next task for the research team is to establish a method for CODE-AE to accurately forecast the impact of a new drug's concentration and metabolization in human bodies. The researchers also pointed out that the AI model might be modified to precisely anticipate adverse medication effects in humans.
Accurate and robust prediction of patient-specific responses to a new compound is critical to personalized drug discovery and development. However, patient data are often too scarce to train a generalized machine-learning model. Although many methods have been developed to utilize cell-line screens for predicting clinical responses, their performances are unreliable owing to data heterogeneity and distribution shift. Here we have developed a novel context-aware deconfounding autoencoder (CODE-AE) that can extract intrinsic biological signals masked by context-specific patterns and confounding factors. Extensive comparative studies demonstrated that CODE-AE effectively alleviated the out-of-distribution problem for the model generalization and significantly improved accuracy and robustness over state-of-the-art methods in predicting patient-specific clinical drug responses purely from cell-line compound screens. Using CODE-AE, we screened 59 drugs for 9,808 patients with cancer. Our results are consistent with existing clinical observations, suggesting the potential of CODE-AE in developing personalized therapies and drug response biomarkers.