AlphaFold 3
AlphaFold 3 is a biomolecular structure prediction model developed by Google DeepMind and the third generation of the AlphaFold series. Expanding on its predecessors, it predicts not only protein three-dimensional structures but also the structures and interactions of DNA, RNA, small-molecule ligands, and protein complexes. AlphaFold 3 employs diffusion models and other deep learning techniques, providing computational support for drug development, disease mechanism research, and life science exploration.
Core Facts
Timeline (last 90 days)
In comparative tests on 20 human proteins (36 to 857 residues in length), under matched sample sizes, the range of folding coordinates covered by FoldScape is about an order of magnitude greater than that of AlphaFold 3 ensembles with seed variation.
AlphaFold 3 anchors better in the local region immediately adjacent to the native state, whereas FoldScape does not reach that region, making the two complementary.
FoldScape cannot cover the very-near-native region immediately adjacent to the native state, which still needs to rely on models such as AlphaFold 3.
Combining FoldScape with AlphaFold 3 may be a more reasonable practical path.
In February 2026, Isomorphic Labs released the Drug Design Engine (IsoDDE), and a technical report states that its performance on protein-ligand structure prediction is twice that of AlphaFold 3, its binding affinity prediction accuracy is higher than physics-based methods, and its time and cost are far lower than the latter.
AlphaFold3 extends to structure prediction of protein–nucleic acid–small molecule complexes.
Xunming Biosciences open-sourced a biological structure modeling model called OpenDDE in July and claimed it outperforms AlphaFold 3 on certain antigen-antibody benchmarks.
All Facts (8)
AlphaFold3, launched in 2024, further extended predictions to DNA, RNA, and small molecules.
75%VerifiedAlphaFold3 extends to structure prediction of protein–nucleic acid–small molecule complexes.
65%UnverifiedCombining FoldScape with AlphaFold 3 may be a more reasonable practical path.
50%UnverifiedAlphaFold 3 anchors better in the local region immediately adjacent to the native state, whereas FoldScape does not reach that region, making the two complementary.
50%UnverifiedFoldScape cannot cover the very-near-native region immediately adjacent to the native state, which still needs to rely on models such as AlphaFold 3.
50%UnverifiedIn comparative tests on 20 human proteins (36 to 857 residues in length), under matched sample sizes, the range of folding coordinates covered by FoldScape is about an order of magnitude greater than that of AlphaFold 3 ensembles with seed variation.
50%UnverifiedXunming Biosciences open-sourced a biological structure modeling model called OpenDDE in July and claimed it outperforms AlphaFold 3 on certain antigen-antibody benchmarks.
50%UnverifiedIn February 2026, Isomorphic Labs released the Drug Design Engine (IsoDDE), and a technical report states that its performance on protein-ligand structure prediction is twice that of AlphaFold 3, its binding affinity prediction accuracy is higher than physics-based methods, and its time and cost are far lower than the latter.
50%