AI is a viable alternative to high throughput screening: a 318-target study

• Main Authors: Wallach, Izhar, Bernard, Denzil, Nguyen, Kong, Ho, Gregory, Morrison, Adrian, Stecula, Adrian, Rosnik, Andreana, O’Sullivan, Ann Marie, Davtyan, Aram, Samudio, Ben, Thomas, Bill, Worley, Brad, Butler, Brittany, Laggner, Christian, Thayer, Desiree, Moharreri, Ehsan, Friedland, Greg, Truong, Ha, van den Bedem, Henry, Ng, Ho Leung, Stafford, Kate, Sarangapani, Krishna, Giesler, Kyle, Ngo, Lien, Mysinger, Michael, Ahmed, Mostafa, Anthis, Nicholas J., Henriksen, Niel, Gniewek, Pawel, Eckert, Sam, de Oliveira, Saulo, Suterwala, Shabbir, PrasadPrasad, Srimukh Veccham Krishna, Shek, Stefani, Contreras, Stephanie, Hare, Stephanie, Palazzo, Teresa, O’Brien, Terrence E., Van Grack, Tessa, Williams, Tiffany, Chern, Ting-Rong, Kenyon, Victor, Lee, Andreia H., Cann, Andrew B., Bergman, Bastiaan, Anderson, Brandon M., Cox, Bryan D., Warrington, Jeffrey M., Sorenson, Jon M., Goldenberg, Joshua M., Young, Matthew A., DeHaan, Nicholas, Pemberton, Ryan P., Schroedl, Stefan, Abramyan, Tigran M., Gupta, Tushita, Mysore, Venkatesh, Presser, Adam G., Ferrando, Adolfo A., Andricopulo, Adriano D., Ghosh, Agnidipta, Ayachi, Aicha Gharbi, Mushtaq, Aisha, Shaqra, Ala M., Toh, Alan Kie Leong, Smrcka, Alan V., Ciccia, Alberto, de Oliveira, Aldo Sena, Sverzhinsky, Aleksandr, de Sousa, Alessandra Mara, Agoulnik, Alexander I., Kushnir, Alexander, Freiberg, Alexander N., Statsyuk, Alexander V., Gingras, Alexandre R., Degterev, Alexei, Tomilov, Alexey, Vrielink, Alice, Garaeva, Alisa A., Bryant-Friedrich, Amanda, Caflisch, Amedeo, Patel, Amit K., Rangarajan, Amith Vikram, Matheeussen, An, Battistoni, Andrea, Caporali, Andrea, Chini, Andrea, Ilari, Andrea, Mattevi, Andrea, Foote, Andrea Talbot, Trabocchi, Andrea, Stahl, Andreas, Herr, Andrew B., Berti, Andrew, Freywald, Andrew, Reidenbach, Andrew G., Lam, Andrew, Cuddihy, Andrew R., White, Andrew, Taglialatela, Angelo
• Other Authors: School of Biological Sciences
• Format: Article
• Language: English
• Published: 2024
• Subjects: Medicine, Health and Life Sciences; High throughput screening; AI
• Online Access: https://hdl.handle.net/10356/179577
• Institution: Nanyang Technological University

High throughput screening (HTS) is routinely used to identify bioactive small molecules. This requires physical compounds, which limits coverage of accessible chemical space. Computational approaches combined with vast on-demand chemical libraries can access far greater chemical space, provided that the predictive accuracy is sufficient to identify useful molecules. Through the largest and most diverse virtual HTS campaign reported to date, comprising 318 individual projects, we demonstrate that our AtomNet® convolutional neural network successfully finds novel hits across every major therapeutic area and protein class. We address historical limitations of computational screening by demonstrating success for target proteins without known binders, high-quality X-ray crystal structures, or manual cherry-picking of compounds. We show that the molecules selected by the AtomNet® model are novel drug-like scaffolds rather than minor modifications to known bioactive compounds. Our empirical results suggest that computational methods can substantially replace HTS as the first step of small-molecule drug discovery.