Scientists code ChatGPT to design new drugs

Generative AI platforms, from ChatGPT to Midjourney, have made headlines in 2023. But GenAI can do more than create collaged images and help write emails: It can also design new drugs to treat diseases.

Today, scientists use cutting-edge technology to design new synthetic drug compounds with appropriate properties and characteristics, also known as “de novo drug design.” However, current methods can be labor intensive, time consuming and costly.

Inspired by the popularity of ChatGPT and wondering whether this approach could speed up the drug design process, scientists at the Schmid College of Science and Technology at Chapman University in Orange, California, decided to create their own GenAI model, detailed in a new article, “De Novo Drug Design using Transformer-based Machine Translation and Reinforcement Learning of Adaptive Monte-Carlo Tree Search,” published in the journal Drugs.

Dony Ang, Cyril Rakovski, and Hagop Atamian coded a model to learn a massive data set of known chemicals, how they bind to target proteins, and the rules and syntax of broader chemical structure and properties.

The end result can generate countless unique molecular structures that follow essential chemical and biological constraints and bind efficiently to their targets, which promises to significantly accelerate the process of identifying viable drug candidates for a wide range of diseases, at a fraction of the cost.

To create this revolutionary model, researchers integrated for the first time two cutting-edge AI techniques from the fields of bioinformatics and computational pathology: the famous “encoder-decoder transformer architecture” and “reinforcement learning via Monte Carlo tree search” (RL-SCTM). The platform, aptly named “drugAI,” allows users to input a target protein sequence (e.g., a protein typically involved in cancer progression).

DrugAI, trained on data from the comprehensive public BindingDB database, can generate unique molecular structures from scratch and then iteratively refine candidates, ensuring that finalists exhibit strong binding affinities with the respective drug targets , which is crucial for the effectiveness of potential drugs. The model identifies 50 to 100 new molecules that could inhibit these particular proteins.

“This approach allows us to generate a potential drug that has never been imagined,” said Dr. Atamian. “This has been tested and validated. Today we are seeing magnificent results.”

The researchers evaluated the molecules generated by drugAI against several criteria and found that drugAI’s results were of similar quality to those of two other common methods, and in some cases, better. They found that drugAI’s drug candidates had a 100% validity rate, meaning that none of the generated drugs were present in the training set.

DrugAI’s drug candidates were also measured for drug-likeness, or the similarity of a compound’s properties to those of oral drugs, and drug candidates were at least 42% and 75% higher than other models. Additionally, all molecules generated by drugAI exhibited strong binding affinities with their respective targets, comparable to those identified via traditional virtual screening approaches.

Ang, Rakovski and Atamian also wanted to see how drugAI’s results for a specific disease compared to existing drugs known for that disease. In another experiment, screening methods generated a list of natural products that inhibited COVID-19 proteins; drugAI generated a list of new drugs targeting the same protein to compare their characteristics. They compared drug resemblance and binding affinity between natural molecules and drugAI, and found similar measurements in both cases, but drugAI was able to identify them much more quickly and less expensively.

Additionally, the scientists designed the algorithm to have a flexible structure that allows future researchers to add new functions. “That means you’re going to end up with more refined drug candidates with an even higher probability of becoming a real drug,” Dr. Atamian said. “We are excited about the possibilities moving forward.”