Luke Lambourne, Anupama Yadav, Yang Wang, Alice Desbuleux, Dae-Kyum Kim, Florent Laval, Kerstin Spirohn-Fitzgerald, Tiziana M. Cafarelli, Carles Pons, I. Kovács, Noor Jailkhani, Sadie Schlabach, David De Ridder, Katja Luck, Vladimir V. Botchkarev, Olivia Debnath, Wenting Bian, Yun Shen, Zhipeng Yang, Miles W. Mee, Mohamed Helmy, Yves Jacob, Irma Lemmens, Thomas Rolland, Gregory G McClain, Atina G. Coté, Marinella Gebbia, Nishka Kishore, Jennifer J. Knapp, Joseph Mellor, Gönen Memişoğlu, Jüri Reimand, Jan Tavernier, Michael E. Cusick, Quan Zhong, Patrick Aloy, Tong Hao, Benoît Charloteaux, Frederick P. Roth, Javier De Las Rivas, Pascal Falter-Braun, David E. Hill, Michael A. Calderwood, Jean‐Claude Twizere, Marc Vidal
Abstract Genotype-phenotype relationships are mediated through intricate networks of physical and functional interactions among macromolecules. Knowledge of the interactome is vital to understand and model genetics and cellular biology. Recent advances in accurately predicting tertiary protein structures using artificial intelligence (AI) approaches such as AlphaFold 1 have revived the vision that the protein-protein interactome might be fully predictable through computational modeling of quaternary structures. Here we present a comprehensive experimental framework to systematically assess the impact of AI-driven interactome predictions for yeast 2 and human 3 . We find that the quality of high-confidence predictions is on par with established experimental approaches. However, in proteome-wide screening, the tested AI approaches underperform in the discovery of strictly novel protein-protein interactions (PPIs) compared to experimental reference interactome maps. In particular, the yeast interactome map described here identifies >40-fold more novel PPIs than its AI counterpart. Strikingly, AlphaFold provides structural models for a substantial number of experimentally identified PPIs missed by the virtual screens. Our results suggest that, at this stage, the main contribution of AI predictions is to provide quaternary structure models for experimentally identified PPIs.