Abramson and colleagues described AlphaFold 3, a diffusion-based model that predicts joint structures of proteins with nucleic acids, small molecules, ions and modified residues, with far better protein–ligand accuracy than docking tools.
Publisher update notice: addendum
As reported in: Abramson, Adler, Dunger et al. (2024). Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature. doi:10.1038/s41586-024-07487-w
Merchant and colleagues used graph networks trained at scale to discover 2.2 million structures below the known stability hull; 736 of the stable structures had already been made independently in experiments.
As reported in: Merchant, Batzner, Schoenholz et al. (2023). Scaling deep learning for materials discovery. Nature. doi:10.1038/s41586-023-06735-9
Szymanski and colleagues ran an autonomous lab that, over 17 days, made 41 of 58 targeted inorganic compounds using computation, literature data, machine learning and robotics.
Disputed: concerns were raised about structure identification and novelty; a 2026 author correction says the compounds were new to the prediction platform, not necessarily new to science (doi:10.1038/s41586-025-09992-y).
Publisher update notice: correction
As reported in: Szymanski, Rendy, Fei et al. (2023). An autonomous laboratory for the accelerated synthesis of inorganic materials. Nature. doi:10.1038/s41586-023-06734-w
Smirnova and colleagues coined "organoid intelligence" and laid out a research program for biocomputing with brain organoids, with an embedded ethics approach.
As reported in: Smirnova, Caffo, Gracias et al. (2023). Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. Frontiers in Science. doi:10.3389/fsci.2023.1017235
Lin and colleagues scaled protein language models to 15 billion parameters, predicting atomic-level structure up to 60× faster, and predicted over 617 million metagenomic protein structures.
As reported in: Lin, Akin, Rao et al. (2023). Evolutionary-scale prediction of atomic-level protein structure with a language model. Science. doi:10.1126/science.ade2574
Kagan and colleagues connected human or rodent neuron cultures to a multielectrode array in a simulated Pong game and saw apparent learning within five minutes, which depended on closed-loop feedback.
As reported in: Kagan, Kitchen, Tran et al. (2022). In vitro neurons learn and exhibit sentience when embodied in a simulated game-world. Neuron. doi:10.1016/j.neuron.2022.09.001
Jumper and colleagues presented AlphaFold, the first computational method to regularly predict protein structures with atomic accuracy even when no similar structure is known, with accuracy competitive with experiments in most CASP14 cases.
As reported in: Jumper, Evans, Pritzel et al. (2021). Highly accurate protein structure prediction with AlphaFold. Nature. doi:10.1038/s41586-021-03819-2
Baek and colleagues built a three-track neural network whose predictions approach AlphaFold's CASP14 accuracy and that can model protein–protein complexes from sequence alone.
As reported in: Baek, DiMaio, Anishchenko et al. (2021). Accurate prediction of protein structures and interactions using a three-track neural network. Science. doi:10.1126/science.abj8754
Rives and colleagues trained a language model on 86 billion amino acids from 250 million proteins; its learned representations encoded structure and enabled state-of-the-art prediction of mutation effects.
As reported in: Rives, Meier, Sercu et al. (2021). Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences. Proceedings of the National Academy of Sciences. doi:10.1073/pnas.2016239118
Burger and colleagues used a free-roaming robot that ran 688 experiments over eight days and found photocatalyst mixtures six times more active than the starting ones.
As reported in: Burger, Maffettone, Gusev et al. (2020). A mobile robotic chemist. Nature. doi:10.1038/s41586-020-2442-2
Dorigo and colleagues proposed the "ant system", a computing method based on how ant colonies work, and tested it on the travelling salesman problem.
As reported in: Dorigo, Maniezzo, Colorni (1996). Ant system: optimization by a colony of cooperating agents. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics). doi:10.1109/3477.484436
Abstract only