On January 9th, Tsinghua University released a heavyweight research achievement: Professor Lan Yanyan from the Intelligent Industry Research Institute (AIR) of the university, together with a team from the School of Life Sciences and the Department of Chemistry, developed the AI driven ultra high throughput drug virtual screening platform DrugCLIP. The related research results have been published in the international academic journal Science. This platform has increased the speed of new drug screening by millions of times, achieving full coverage screening of human genome level targets for the first time, and promoting new drug research and development into a new stage of precision and efficiency. Accurate and efficient, breaking through the bottleneck of new drug screening speed, the core challenge of new drug development is to match and adapt small molecules to disease targets, the process is like searching for a needle in a haystack. There are over 20000 types of proteins in the human body, many of which are key targets for disease resistance, but currently only 10% of them can be targeted with targeted drugs; The large number of small molecules that can be used in pharmaceuticals makes it difficult for traditional screening methods to fully cover them. Traditional screening is like screening one by one in the boundless chemical universe. If 10000 targets are screened and each target matches 1 billion candidate molecules, it would take hundreds of years for a computer to run continuously. Therefore, existing drug development can only be carried out on a small number of targets and small-scale molecules, and a large number of potential therapeutic targets and drugs are buried. ”Lan Yanyan introduced in an interview with a reporter from Science and Technology Daily. DrugCLIP provides precise navigation for new drug screening, breaking through the speed bottleneck in one fell swoop. This achievement breaks away from traditional screening pathways and converts proteins and small molecules into exclusive signals that can be quickly recognized by computers, without the need to gradually simulate the binding process, achieving automatic and accurate matching of small molecules and targets. Lan Yanyan admitted that in the early stages of development, the team was not fully confident in the accuracy of screening, but the first laboratory experiment was successful, preliminarily verifying the effectiveness of the platform. Research has shown that the speed advantage of DrugCLIP in drug screening is extremely impressive. On ordinary high-performance computers, DrugCLIP can complete 31 trillion matching calculations in a single day, screening 1 million candidate molecules in just 0.02 seconds, which is millions of times faster than traditional methods. The workload of the past few hundred years can now be completed in a single day on a single computer, completely ending the speed dilemma of new drug screening. Benefiting people's livelihoods, the accuracy of DrugCLIP, from free and open source to co building an industrial ecosystem, can also be tested in practical applications. According to Lan Yanyan, the joint team first conducted experiments on depression related targets, allowing the platform to accurately screen 100 potential molecules from 1.6 million candidate molecules, of which 15% were proven to effectively act on the targets. The binding effect of 12 molecules was better than that of commonly used antidepressants, and the effectiveness was verified through biological and chemical experiments. The team also overcame a scientific research "hard bone": a target closely related to tumors and Parkinson's disease that had never been found to be suitable for small molecules before. The team first predicted its protein structure with AI, and then screened it with DrugCLIP. Among the 50 potential molecules, 10 could successfully bind, and 2 could effectively inhibit the activity of the target, confirming that the platform is suitable for various targets without ready-made structures and has a wide range of applications. Previously, AI prediction systems solved the problem of 'seeing protein structures clearly', while DrugCLIP focused on solving the problem of 'finding suitable small molecules for proteins', bridging the entire chain from protein structure analysis to new drug development. This is a key breakthrough in the industry's development. ”Lan Yanyan said. Relying on DrugCLIP, the joint team set a new record: completed full coverage screening of approximately 10000 targets and 20000 key loci within the human genome, analyzed over 500 million candidate small molecules, enriched over 2 million potential effective molecules, and built the world's largest drug target matching database, which is open and shared for free by researchers worldwide. Lan Yanyan stated that in the future, DrugCLIP will collaborate deeply with scientific research and industrial ecosystem partners, focusing on fields such as cancer, infectious diseases, and rare diseases, to accelerate the discovery of new targets and First in class drugs. The team will continue to optimize engine performance, expand support modes, and help build a more intelligent, efficient, and inclusive global drug innovation ecosystem. (New Society)