Protein complex prediction with AlphaFold-Multimer

TLDR

AlphaFold accurately predicts single protein chains, but multi‑chain complex prediction remains challenging. The study introduces AlphaFold‑Multimer, a multimer‑specific AlphaFold model that improves interface accuracy while preserving intra‑chain predictions. AlphaFold‑Multimer was trained on multimeric inputs of known stoichiometry and applied to 4,446 recent protein complexes, scoring all non‑redundant interfaces with low template identity. On 17 template‑free heterodimers, AlphaFold‑Multimer achieved medium accuracy on 13 targets and high accuracy on 7, surpassing the prior state of the art; across 4,446 recent complexes, it correctly predicted heteromeric interfaces in 70% (high accuracy 26%) and homomeric interfaces in 72% (high accuracy 36%), improving performance by 27–14% for heteromers and 8–7% for homomers over the flexible‑linker AlphaFold.

Abstract

While the vast majority of well-structured single protein chains can now be predicted to high accuracy due to the recent AlphaFold [1] model, the prediction of multi-chain protein complexes remains a challenge in many cases. In this work, we demonstrate that an AlphaFold model trained specifically for multimeric inputs of known stoichiometry, which we call AlphaFold-Multimer, significantly increases accuracy of predicted multimeric interfaces over input-adapted single-chain AlphaFold while maintaining high intra-chain accuracy. On a benchmark dataset of 17 heterodimer proteins without templates (introduced in [2]) we achieve at least medium accuracy (DockQ [3] ≥ 0.49) on 13 targets and high accuracy (DockQ ≥ 0.8) on 7 targets, compared to 9 targets of at least medium accuracy and 4 of high accuracy for the previous state of the art system (an AlphaFold-based system from [2]). We also predict structures for a large dataset of 4,446 recent protein complexes, from which we score all non-redundant interfaces with low template identity. For heteromeric interfaces we successfully predict the interface (DockQ ≥ 0.23) in 70% of cases, and produce high accuracy predictions (DockQ ≥ 0.8) in 26% of cases, an improvement of +27 and +14 percentage points over the flexible linker modification of AlphaFold [4] respectively. For homomeric inter-faces we successfully predict the interface in 72% of cases, and produce high accuracy predictions in 36% of cases, an improvement of +8 and +7 percentage points respectively.