Artificial intelligence tools accelerate drug and protein research breakthroughs

07:50

By: Dakir Madiha

Artificial intelligence tools accelerate drug and protein research breakthroughs

A new generation of artificial intelligence tools is transforming biomedical research, enabling scientists to analyze genetic regulation, decode protein structures, and design drug compounds in a fraction of the time previously required. Recent studies show that machine learning systems can compress months or even years of laboratory work into days.

One recent study published in Nature introduced a machine learning system capable of analyzing tens of thousands of chemical structures to predict how molecules will assemble during drug synthesis. Developed by researchers from the University of Utah and the University of California, Los Angeles, the system reduces the lengthy process of optimizing chemical reactions, which often takes months of experimentation.

The tool addresses a major challenge in applying artificial intelligence to chemistry. AI models typically require massive datasets, but producing high quality experimental chemistry data is expensive and time consuming. According to Matthew Sigman, a chemist at the University of Utah, the system allows researchers to work with smaller datasets while still generating reliable predictions. The model can also transfer its predictions to chemical reactions it has not previously encountered.

In related work, researchers at Yale University collaborating with pharmaceutical company Boehringer Ingelheim created an AI platform called MOSAIC. Reported by Nature in January, the platform identified more than 35 new compounds, including pharmaceutical and cosmetic ingredients, achieving a success rate of about 71 percent.

Artificial intelligence is also improving the study of protein structures. On March 10, the Lawrence Berkeley National Laboratory announced a program called AQuaRef, described in Nature Communications, that combines quantum computing techniques with AI to determine protein structures more accurately while reducing computational costs.

Tests on 71 protein structures showed improved performance compared with existing methods. The system was also able to correctly determine proton positions in DJ-1, a human protein linked to certain forms of Parkinson’s disease that has been difficult to map using conventional techniques.

Researchers at the National University of Singapore separately reported progress with their AI system D-I-TASSER, which predicts complex protein structures with about 13 percent greater accuracy than previous leading methods.

Advances are also emerging in the study of gene regulation. Scientists at the Joint BioEnergy Institute of Lawrence Berkeley National Laboratory developed a high throughput platform that can test thousands of plant genetic switches in a single experiment. These DNA sequences control when genes are activated or silenced, and identifying them has been a major bottleneck in plant synthetic biology.

While CRISPR technology allows precise gene editing, identifying the regulatory elements to modify has remained slow. The new platform aims to accelerate that process.

Meanwhile, researchers at the Broad Institute of MIT and Harvard created an AI framework that automatically identifies shared cellular information across multiple measurement types. The approach gives scientists a more integrated view of cellular states involved in diseases such as cancer, Alzheimer’s disease, and metabolic disorders.

These developments arrive as AI designed drugs move toward late stage clinical testing. According to Drug Target Review, 2026 could become a decisive year for AI driven drug discovery as several treatments identified using artificial intelligence enter critical phase III clinical trials.