Abstract:AlphaFold2 (AF2) and RoseTTaFold (RF) have revolutionized structural biology, serving as highly reliable and effective methods for predicting protein structures. This article explores their impact and limitations, focusing on their integration into experimental pipelines and their application in diverse protein classes, including membrane proteins, intrinsically disordered proteins (IDPs), and oligomers. In experimental pipelines, AF2 models help X-ray crystallography in resolving the phase problem, while complementarity with mass spectrometry and NMR data enhances structure determination and protein flexibility prediction. Predicting the structure of membrane proteins remains challenging for both AF2 and RF due to difficulties in capturing conformational ensembles and interactions with the membrane. Improvements in incorporating membrane-specific features and predicting the structural effect of mutations are crucial. For intrinsically disordered proteins, AF2's confidence score (pLDDT) serves as a competitive disorder predictor, but integrative approaches including molecular dynamics (MD) simulations or hydrophobic cluster analyses are advocated for accurate dynamics representation. AF2 and RF show promising results for oligomeric models, outperforming traditional docking methods, with AlphaFold-Multimer showing improved performance. However, some caveats remain in particular for membrane proteins. Real-life examples demonstrate AF2's predictive capabilities in unknown protein structures, but models should be evaluated for their agreement with experimental data. Furthermore, AF2 models can be used complementarily with MD simulations. In this Perspective, we propose a "wish list" for improving deep-learning-based protein folding prediction models, including using experimental data as constraints and modifying models with binding partners or post-translational modifications. Additionally, a meta-tool for ranking and suggesting composite models is suggested, driving future advancements in this rapidly evolving field.

Effective Molecular Dynamics from Neural Network-Based Structure Prediction Models

High-accuracy protein structures by combining machine-learning with physics-based refinement

A Perspective on the Prospective Use of AI in Protein Structure Prediction

The power and pitfalls of AlphaFold2 for structure prediction beyond rigid globular proteins

Efficient and accurate prediction of protein structure using RoseTTAFold2

Predicting protein conformational motions using energetic frustration analysis and AlphaFold2

Predicting relative populations of protein conformations without a physics engine using AlphaFold2

Accurate prediction of protein folding mechanisms by simple structure-based statistical mechanical models

Protein structure predictions to atomic accuracy with AlphaFold

Protein Design Using Structure-Prediction Networks: AlphaFold and RoseTTAFold as Protein Structure Foundation Models

Dissecting AlphaFolds Capabilities with Limited Sequence Information

Structure prediction of alternative protein conformations

Improved AlphaFold modeling with implicit experimental information

Single-sequence protein structure prediction using a language model and deep learning

Accurate Prediction of Protein Structural Flexibility by Deep Learning Integrating Intricate Atomic Structures and Cryo-EM Density Information

State-of-the-Art Estimation of Protein Model Accuracy using AlphaFold

Highly accurate protein structure prediction with AlphaFold

Comparative studies of AlphaFold, RoseTTAFold and Modeller: a case study involving the use of G-protein-coupled receptors

Do Deep Learning Models for Co-Folding Learn the Physics of Protein-Ligand Interactions?

AlphaFolding: 4D Diffusion for Dynamic Protein Structure Prediction with Reference and Motion Guidance