Integrated XGBoost-Based Virtual Screening and Molecular Dynamics Simulation for the Identification of Potential CCR5 Antagonists
This study identifies FDB020301 as a promising next-generation anti-HIV lead by integrating XGBoost-based virtual screening, ADMET profiling, molecular docking, and molecular dynamics simulations to demonstrate its superior binding affinity and stability against the CCR5 receptor compared to the clinical drug Maraviroc.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Human Immunodeficiency Virus, or HIV, remains a persistent global health challenge, requiring a constant search for new ways to stop the virus from infecting human cells. The virus does not enter cells on its own; it needs a specific set of keys to unlock the door. One of the most important keys is a protein on the surface of immune cells called the CCR5 receptor. When the virus attaches to this receptor, it gains entry and begins to replicate. Scientists have long known that blocking this receptor can stop the infection, and a drug called Maraviroc was developed to do exactly that. However, like many treatments, the virus can eventually learn to bypass this block, and the drug does not work for every strain of HIV. This reality drives researchers to look for new, stronger blockers that can outsmart the virus and keep immune cells safe.
To find these new blockers without spending years testing thousands of chemicals in a lab, a team of researchers at Kakatiya University in India turned to the power of computers. They combined two powerful approaches: machine learning, which is a type of artificial intelligence that learns patterns from data, and molecular simulation, which acts like a high-speed movie showing how molecules move and stick together. Their goal was to sift through a massive library of food-derived chemicals to find a few promising candidates that could bind tightly to the CCR5 receptor and shut it down. By using a computer model trained on known active and inactive drugs, they narrowed down a list of nearly seventy thousand compounds to just a handful of top contenders, eventually identifying one molecule that showed exceptional promise in their digital tests.
The researchers began their search with a vast collection of chemical structures known as FooDB, which contains information on more than seventy thousand compounds found in food. They first used a machine learning tool called XGBoost to predict which of these chemicals might be able to stop the CCR5 receptor. This tool had been trained on a dataset of over thirteen thousand known chemicals, learning to distinguish between those that successfully blocked the receptor and those that did not. When the team fed the food database into this trained model, the computer flagged over ten thousand compounds as potentially active. To make the list manageable, they focused on the two thousand most likely candidates and checked them against standard rules for drug safety and absorption. This process, known as filtering for drug-likeness, ensured that the chemicals were small enough and had the right chemical balance to be absorbed by the human body if taken as a pill.
After this initial screening, the team applied a second layer of checks to predict how the chemicals would behave inside the human body. They looked for signs of toxicity, such as the potential to cause cancer or damage the heart, and checked how well the body would absorb and process them. Only eleven compounds passed all these safety and absorption hurdles. These eleven were then subjected to a detailed computer simulation called molecular docking. In this step, the researchers placed each of the eleven chemicals into a digital model of the CCR5 receptor to see how well they fit. They compared the results to Maraviroc, the current standard drug. One compound, identified as FDB020301, stood out immediately. It fit into the receptor with a binding energy of negative eleven point four kilocalories per mole, which is a measure of how tightly the two stick together. This score was stronger than that of Maraviroc, which scored negative ten point five, suggesting that FDB020301 might hold on more securely.
To ensure that this tight fit was not just a lucky snapshot but a stable reality, the researchers ran a ten-nanosecond molecular dynamics simulation. This is a computer experiment that watches how the drug and the receptor move and interact over time, accounting for the constant jiggling of atoms and the presence of water molecules. The simulation showed that the complex remained stable throughout the entire run. The receptor did not lose its shape, and the drug stayed firmly in place, forming hydrogen bonds and other interactions with specific parts of the receptor, such as amino acids named TYR37 and GLN280. The researchers also calculated the energy required to pull the drug away from the receptor, finding a value of negative thirty-five point nineteen kilocalories per mole, which further confirmed that the bond was strong and energetically favorable.
The study concludes that FDB020301 is a highly promising candidate for a new HIV treatment, but the researchers are careful to note that these results exist only within the computer. The molecule has not yet been tested in a living cell or a human body. While the digital evidence suggests it is a strong blocker with good safety profiles, the next steps require actual laboratory experiments to prove that it works against the virus in the real world. Until then, this work serves as a powerful demonstration of how modern computing can guide scientists toward new possibilities, turning a massive sea of chemical data into a single, focused target for future medical breakthroughs.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.