Modeling conformational states of proteins with AlphaFold

Article Biochemistry & Molecular Biology

GPCRdb in 2023: state-specific structure models using AlphaFold2 and new ligand resources

Gaspar Pandy-Szekeres et al.

Summary: This article introduces GPCRdb as a database supporting interdisciplinary research and provides an overview of its various resources and tools for receptor sequences, structures, and ligands.

NUCLEIC ACIDS RESEARCH (2023)

Add to Collection

Review Biochemistry & Molecular Biology

Frozen in time: analyzing molecular dynamics with time-resolved cryo-EM

Sascha Josef Amann et al.

Summary: Molecular machines achieve complex tasks by restricting random motion through interaction with cofactors. Cryo-EM has been used to elucidate major states of these machines, but the coordination of their individual steps in time is still elusive. Time-resolved cryo-EM (trEM) is a method that allows measurement of short-lived intermediate events by shortening the total reaction time.

STRUCTURE (2023)

Add to Collection

Review Pharmacology & Pharmacy

Targeting in silico GPCR conformations with ultra-large library screening for hit discovery

D. Sala et al.

Summary: The use of deep machine learning in protein structure prediction allows easy access to annotated conformations, which can compensate for missing experimental structures in structure-based drug discovery. However, the accuracy of these predicted conformations for screening chemical compounds that effectively interact with protein targets is still uncertain. This opinion article examines the benefits and limitations of using state-annotated conformations for ultra-large library screening, particularly for common drug targets like G-protein-coupled receptors.

TRENDS IN PHARMACOLOGICAL SCIENCES (2023)

Add to Collection

Article Chemistry, Medicinal

Benchmarking Refined and Unrefined AlphaFold2 Structures for Hit Discovery

Yuqi Zhang et al.

Summary: The AlphaFold2 (AF2) algorithm predicts proteins' 3D structures from amino acid sequences. The open AlphaFold protein structure database covers the complete human proteome. Through virtual screening, AF2 structures show comparable early enrichment of known active compounds to apo structures, and with refinement using an induced-fit protocol, the performance in structure-based virtual screening can be improved. Thus, AF2 structures hold promise for in silico hit identification.

JOURNAL OF CHEMICAL INFORMATION AND MODELING (2023)

Add to Collection

Article Chemistry, Physical

AlphaFold2-RAVE: From Sequence to Boltzmann Ranking

Bodhi P. Vani et al.

Summary: AlphaFold2-RAVE is an efficient protocol to obtain Boltzmann-ranked ensembles from sequence by using structural outputs from AlphaFold2 as initializations for AI-augmented molecular dynamics. The method has demonstrated good results on different proteins and is available as open-source code.

JOURNAL OF CHEMICAL THEORY AND COMPUTATION (2023)

Add to Collection

Article Biotechnology & Applied Microbiology

Protein structure prediction with in-cell photo-crosslinking mass spectrometry and deep learning

Kolja Stahl et al.

Summary: Limitations of Alphafold2 structure prediction are overcome by incorporating experimentally determined distance constraints using AlphaLink. This modified version improves prediction for challenging targets by utilizing sparse experimental contacts and can predict distinct protein conformations. Experimental data on residue-residue contacts obtained through crosslinking mass spectrometry validates the improved performance. The noise-tolerant framework for integrating data allows accurate characterization of protein structures from in-cell data.

NATURE BIOTECHNOLOGY (2023)

Add to Collection

Article Multidisciplinary Sciences

Evolutionary-scale prediction of atomic-level protein structure with a language model

Zeming Lin et al.

Summary: Recent advances in machine learning have allowed for the prediction of protein structure from multiple sequence alignments. By using a large language model, we are able to directly infer full atomic-level protein structure from primary sequence. This has led to a significant acceleration in high-resolution structure prediction, enabling the characterization of a large number of metagenomic proteins. Utilizing this capability, we have constructed the ESM Metagenomic Atlas, which provides insights into the diversity of natural proteins.

SCIENCE (2023)

Add to Collection

Article Biochemistry & Molecular Biology

Biasing AlphaFold2 to predict GPCRs and kinases with user-defined functional or structural properties

Davide Sala et al.

Summary: Determining the native functional structure of proteins has long been a challenge in structural biology. Integrative structural biology and deep machine-learning algorithms, such as AlphaFold2, have greatly advanced computational predictions of protein structures. In this study, we expanded AlphaFold2 to model user-defined functional or structural features, specifically focusing on G-protein-coupled receptors (GPCRs) and kinases for drug discovery. Our approach automatically selects templates that satisfy specified features and incorporates genetic information, providing accurate and biased models. This protocol allows for automatic modeling of user-defined conformational states.

FRONTIERS IN MOLECULAR BIOSCIENCES (2023)

Add to Collection

Letter Biochemistry & Molecular Biology

Can AlphaFold2 predict the impact of missense mutations on structure?

Gwen R. Buel et al.

NATURE STRUCTURAL & MOLECULAR BIOLOGY (2022)

Add to Collection

Article Biochemistry & Molecular Biology

AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models

Mihaly Varadi et al.

Summary: AlphaFold DB is an openly accessible database with high-accuracy protein-structure predictions, powered by DeepMind's AlphaFold v2.0. It provides programmatic access to a vast number of predicted structures and is expanding to cover more sequences.

NUCLEIC ACIDS RESEARCH (2022)

Add to Collection

Review Chemistry, Multidisciplinary

Subsecond Time-Resolved Mass Spectrometry in Dynamic Structural Biology

Cristina Lento et al.

Summary: Life at the molecular level is a dynamic world, and subsecond time-resolved mass spectrometry provides a unique window into the dynamic behavior of biological macromolecules. This technique has made important contributions to understanding the functional and pathological processes of biomolecules.

CHEMICAL REVIEWS (2022)

Add to Collection

Article Biochemical Research Methods

Impact of protein conformational diversity on AlphaFold predictions

Tadeo Saldano et al.

Summary: By studying the performance of AlphaFold2 under the ensemble paradigm, it was found that AlphaFold2 performs well in predicting the holo form of proteins, but its performance deteriorates with increasing conformational diversity. There is a negative correlation between the main-chain flexibility associated with apo-holo pairs of conformers and the predicted local model quality score.

BIOINFORMATICS (2022)

Add to Collection

Article Biochemistry & Molecular Biology

AlphaFold2: A Role for Disordered Protein/Region Prediction?

Carter J. Wilson et al.

Summary: The development of AlphaFold2 has brought a paradigm shift in structural biology and has shown excellent performance in predicting disordered regions. The accuracy of the constructed disorder predictor from AlphaFold2 structures varies, with the naive assumption of assigning residues to ordered or disordered based on helices, strands, and H-bond stabilized turns leading to a significant overestimation of disorder. The predicted local distance difference test (pLDDT) provides a reliable measure of residue-wise disorder and shows an interesting relationship with secondary structure.

INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES (2022)

Add to Collection

Article Biochemical Research Methods

ColabFold: making protein folding accessible to all

Milot Mirdita et al.

Summary: ColabFold combines fast homology search and optimized model utilization to offer accelerated prediction of protein structures and complexes, with a processing speed that is 40-60 times faster. It serves as a free and accessible platform for protein folding, capable of predicting close to 1,000 structures per day.

NATURE METHODS (2022)

Add to Collection

Article Biochemistry & Molecular Biology

Multi-state modeling of G-protein coupled receptors at experimental accuracy

Lim Heo et al.

Summary: The paper introduces a protocol that uses AlphaFold2 and GPCR databases to predict the active or inactive states of GPCRs with high accuracy. The predicted models capture the structural changes upon activation of GPCRs and are more suitable for protein-ligand docking.

PROTEINS-STRUCTURE FUNCTION AND BIOINFORMATICS (2022)

Add to Collection

Article Biology

Sampling alternative conformational states of transporters and receptors with AlphaFold2

Diego del Alamo et al.

Summary: This article presents an approach to drive AlphaFold2 to sample multiple conformations of membrane proteins and demonstrates that this approach can generate accurate models. The results suggest the need for the next generation of deep learning algorithms to be designed to predict ensembles of biophysically relevant states.

ELIFE (2022)

Add to Collection

Article Physics, Multidisciplinary

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

James P. Roney et al.

Summary: Predicting a protein's 3D structure has been a challenge in structural biology, but recent approaches like AlphaFold have made significant progress by combining deep learning and coevolutionary data. We provide evidence that AlphaFold has learned a biophysical energy function and uses coevolution data to solve the global search problem of finding a low-energy conformation. AlphaFold's learned energy function accurately ranks the quality of candidate protein structures without using coevolution data.

PHYSICAL REVIEW LETTERS (2022)

Add to Collection

Review Chemistry, Physical

AlphaFold, Artificial Intelligence (AI), and Allostery

Ruth Nussinov et al.

Summary: AlphaFold is a powerful algorithm that leverages biological sequence data and artificial intelligence. It has the potential for vast impacts in the life sciences, but also faces challenges in certain areas such as protein folding mechanisms.

JOURNAL OF PHYSICAL CHEMISTRY B (2022)

Add to Collection

Article Multidisciplinary Sciences

Integrated AlphaFold2 and DEER investigation of the conformational dynamics of a pH-dependent APC antiporter

Diego del Alamo et al.

Summary: The APC transporter GadC plays a role in the survival of pathogenic bacteria under extreme acid stress. Through studying its conformational dynamics using DEER spectroscopy, researchers discovered acid-induced conformational changes in GadC, enabling isomerization between inward- and outward-facing states. The substrate glutamate modulates the dynamics of an extracellular gate. This study provides insights into the conformational cycle of GadC and highlights the divergence in dynamics among different families in the LeuT-fold.

PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA (2022)

Add to Collection

Article Biochemistry & Molecular Biology

Modeling of protein conformational changes with Rosetta guided by limited experimental data

Davide Sala et al.

Summary: Conformational changes are crucial for the functional cycles of proteins, and their study often requires an integrative approach. In this study, a new method called ConfChangeMover (CCM) was introduced and tested for modeling protein conformational changes using sparse experimental data. CCM outperformed state-of-the-art methods in two benchmarks, demonstrating its ability to model diverse conformational changes. Moreover, the integration of CCM with other experimental data using the Rosetta framework further expands its capability.

STRUCTURE (2022)

Add to Collection

Review Biochemical Research Methods

Structural dynamics: review of time-resolved cryo-EM

Mart-Erik Maeots et al.

Summary: Determining the structures of biological macromolecules has greatly advanced our understanding of biochemical mechanisms and therapeutic developments, but characterizing structural dynamics underlying biochemical processes has been challenging. Single-particle cryo-EM presents a powerful platform to address this challenge by allowing samples to be frozen faster than the timescales of most biochemical reactions, bringing us closer to the goal of directly observing protein dynamics in milliseconds to seconds range. Innovations in handling and preparing cryo-EM samples have enabled time-resolved analysis, reviewed in this article alongside discussion of promising future research directions.

ACTA CRYSTALLOGRAPHICA SECTION D-STRUCTURAL BIOLOGY (2022)

Add to Collection

Article Biotechnology & Applied Microbiology

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

Ratul Chowdhury et al.

Summary: This article reports the development of an end-to-end differentiable recurrent geometric network (RGN) called RGN2 for predicting protein structure using the protein sequence. Compared to AlphaFold2 and RoseTTAFold, RGN2 performs better on orphan proteins and designed protein classes, while achieving a computational speedup of up to 10^6-fold. These findings demonstrate the practical and theoretical advantages of protein language models over multiple sequence alignments (MSAs) in protein structure prediction.

NATURE BIOTECHNOLOGY (2022)

Add to Collection

Article Biochemical Research Methods

SPEACH_AF: Sampling protein ensembles and conformational heterogeneity with Alphafold2

Richard A. Stein et al.

Summary: This study presents a general approach to drive Alphafold2 to model alternate protein conformations through manipulation of the multiple sequence alignment. The applicability of this approach is demonstrated through the investigation of membrane proteins, expanding the potential uses of Alphafold2.

PLOS COMPUTATIONAL BIOLOGY (2022)

Add to Collection

Review Biochemistry & Molecular Biology

Protein structure prediction in the deep learning era

Zhenling Peng et al.

Summary: This article reviews the progress of deep learning-based protein structure prediction methods in the past two years, analyzes the advantages and disadvantages of the two-step and end-to-end approaches, emphasizes the value of developing both approaches, and points out the challenges in function-oriented protein structure prediction.

CURRENT OPINION IN STRUCTURAL BIOLOGY (2022)

Add to Collection

Article Mathematical & Computational Biology

Nearest neighbor search on embeddings rapidly identifies distant protein relations

Konstantin Schuetze et al.

Summary: This article explores the use of embeddings for nearest neighbor searches to identify the relationships between protein pairs with diverged sequences. While the approach performs well for proteins with single domains, it faces challenges with multi-domain proteins. The authors present ideas to overcome these limitations.

FRONTIERS IN BIOINFORMATICS (2022)

Add to Collection

Review Chemistry, Physical

A review of advancements in coarse-grained molecular dynamics simulations

Soumil Y. Joshi et al.

Summary: Coarse-grained molecular dynamics has become an efficient way to model large systems, but developing models and accurate interaction potentials remains challenging. Traditional parameterisation techniques, while tedious, are still widely used, and advanced optimization methods and machine learning can help develop models with higher transferability and accuracy more quickly.

MOLECULAR SIMULATION (2021)

Add to Collection

Article Biochemistry & Molecular Biology

KLIFS: an overhaul after the first 5 years of supporting kinase research

Georgi K. Kanev et al.

Summary: Kinases are a major target for drug development, with KLIFS being a database focusing on kinase structures and interactions to support research. The latest version of KLIFS has integrated approved drugs, extended coverage to atypical kinases, and provided a RESTful API for programmatic access.

NUCLEIC ACIDS RESEARCH (2021)

Add to Collection

Review Biochemistry & Molecular Biology

Machine learning in protein structure prediction

Mohammed AlQuraishi

Summary: Prediction of protein structure from sequence has made significant progress in the past two years, driven by the increasing use of neural networks in structure prediction pipelines. These neural networks have optimized the previous energy models and sampling procedures, resulting in algorithms that can now predict protein structures with a median accuracy of 2.1 angstroms.

CURRENT OPINION IN CHEMICAL BIOLOGY (2021)

Add to Collection

Article Biochemistry & Molecular Biology

GPCRdb in 2021: integrating GPCR sequence, structure and function

Albert J. Kooistra et al.

Summary: GPCRdb is a database that serves over 4,000 researchers, providing a wide range of information and tools related to G protein-coupled receptors, including structural models, drug and ligand data. It continuously updates and improves its tools and resources for research and analysis.

NUCLEIC ACIDS RESEARCH (2021)

Add to Collection

Article Biochemistry & Molecular Biology

Applying and improving AlphaFold at CASP14

John Jumper et al.

Summary: The AlphaFold system made significant improvements in CASP14, achieving a high level of accuracy in protein structure prediction and performing remarkably well on Free Modeling targets.

PROTEINS-STRUCTURE FUNCTION AND BIOINFORMATICS (2021)

Add to Collection

Review Biochemistry & Molecular Biology

AlphaFold and Implications for Intrinsically Disordered Proteins

Kiersten M. Ruff et al.

Summary: Accurate predictions of protein structures using AlphaFold have made significant progress. Most protein sequences in the human proteome have been structurally annotated, with over 30% of regions being disordered.

JOURNAL OF MOLECULAR BIOLOGY (2021)

Add to Collection

Article Multidisciplinary Sciences

Highly accurate protein structure prediction with AlphaFold

John Jumper et al.

Summary: Proteins are essential for life, and accurate prediction of their structures is a crucial research problem. Current experimental methods are time-consuming, highlighting the need for accurate computational approaches to address the gap in structural coverage. Despite recent progress, existing methods fall short of atomic accuracy in protein structure prediction.

NATURE (2021)

Add to Collection

Article Multidisciplinary Sciences

Accurate prediction of protein structures and interactions using a three-track neural network

Minkyung Baek et al.

Summary: Through the three-track network, we achieved accuracies approaching those of DeepMind in CASP14, enabling rapid solution of challenging x-ray crystallography and cryo-electron microscopy structure modeling problems, and providing insights into the functions of proteins with currently unknown structure.

SCIENCE (2021)

Add to Collection

Article Biochemical Research Methods