Putting AlphaFold models to work with phenix.process_predicted_model and ISOLDE

Structural and functional prediction, evaluation, and validation in the post-sequencing era

The surge of genome sequencing data has underlined substantial genetic variants of uncertain significance (VUS). The decryption of VUS discovered by sequencing poses a major challenge in the post-sequencing era. Although experimental assays have progressed in classifying VUS, only a tiny fraction of the human genes have been explored experimentally. Thus, it is urgently needed to generate state-of-the-art functional predictors of VUS in silico. Artificial intelligence (AI) is an invaluable tool to assist in the identification of VUS with high efficiency and accuracy. An increasing number of studies indicate that AI has brought an exciting acceleration in the interpretation of VUS, and our group has already used AI to develop protein structure-based prediction models. In this review, we provide an overview of the previous research on AI-based prediction of missense variants, and elucidate the challenges and opportunities for protein structure-based variant prediction in the post-sequencing era.

DomainFit: Identification of protein domains in cryo-EM maps at intermediate resolution using AlphaFold2-predicted models

Cryoelectron microscopy (cryo-EM) has revolutionized the structural determination of macromolecular complexes. With the paradigm shift to structure determination of highly complex endogenous macromolecular complexes ex vivo and in situ structural biology, there are an increasing number of structures of native complexes. These complexes often contain unidentified proteins, related to different cellular states or processes. Identifying proteins at resolutions lower than 4 Å remains challenging because side chains cannot be visualized reliably. Here, we present DomainFit, a program for semi-automated domain-level protein identification from cryo-EM maps, particularly at resolutions lower than 4 Å. By fitting domains from AlphaFold2-predicted models into cryo-EM maps, the program performs statistical analyses and attempts to identify the domains and protein candidates forming the density. Using DomainFit, we identified two microtubule inner proteins, one of which contains a CCDC81 domain and is exclusively localized in the proximal region of the doublet microtubule in Tetrahymena thermophila.

Structure and dynamics of the cyanobacterial regulator SipA

The small, 78-residue long, regulator SipA interacts with the non-bleaching sensor histidine kinase (NblS). We have solved the solution structure of SipA on the basis of 990 nuclear Overhauser effect- (NOE-) derived distance constraints. The average pairwise root-mean-square deviation (RMSD) for the twenty best structures for the backbone residues, obtained by CYANA, was 1.35 ± 0.21 Å, and 1.90 ± 0.16 Å when all heavy atoms were considered (the target function of CYANA was 0.540 ± 0.08). The structure is that of a β-II class protein, basically formed by a five-stranded β-sheet composed of antiparallel strands following the arrangement: Gly6-Leu11 (β-strand 1), which packs against Leu66-Val69 (β-strand 5) on one side, and against Gly36-Thr42 (β-strand 2) on the other side; Trp50-Phe54 (β-strand 3); and Gly57-Leu60 (β-strand 4). The protein is highly mobile, as shown by measurements of R1, R2, NOE and ηxy relaxation parameters, with an average order parameter () of 0.70; this mobility encompasses movements in different time scales. We hypothesize that this high flexibility allows the interaction with other proteins (among them NblS), and it explains the large conformational stability of SipA.

Automated model building and protein identification in cryo-EM maps

Interpreting electron cryo-microscopy (cryo-EM) maps with atomic models requires high levels of expertise and labour-intensive manual intervention in three-dimensional computer graphics programs1,2. Here we present ModelAngelo, a machine-learning approach for automated atomic model building in cryo-EM maps. By combining information from the cryo-EM map with information from protein sequence and structure in a single graph neural network, ModelAngelo builds atomic models for proteins that are of similar quality to those generated by human experts. For nucleotides, ModelAngelo builds backbones with similar accuracy to those built by humans. By using its predicted amino acid probabilities for each residue in hidden Markov model sequence searches, ModelAngelo outperforms human experts in the identification of proteins with unknown sequences. ModelAngelo will therefore remove bottlenecks and increase objectivity in cryo-EM structure determination.

TARGETING SOLUBLE AMYLOID-BETA OLIGOMERS WITH A NOVEL NANOBODY

The classical amyloid cascade hypothesis postulates that the aggregation of amyloid plaques and the accumulation of intracellular hyperphosphorylated Tau tangles, together, lead to profound neuronal death. However, emerging research has demonstrated that soluble amyloid-β oligomers (SAβOs) accumulate early, prior to amyloid plaque formation. SAβOs induce memory impairment and disrupt cognitive function independent of amyloid-β plaques, and even in the absence of plaque formation. This work describes the development and characterization of a novel anti-SAβO (E3) nanobody generated from an alpaca immunized with SAβO. In-vitro assays and in-vivo studies using 5XFAD mice indicate that the fluorescein (FAM)-labeled E3 nanobody recognizes both SAβOs and amyloid-β plaques. The E3 nanobody traverses across the blood-brain barrier and binds to amyloid species in the brain of 5XFAD mice. Imaging of mouse brains reveals that SAβO and amyloid-β plaques are not only different in size, shape, and morphology, but also have a distinct spatial distribution in the brain. SAβOs are associated with neurons, while amyloid plaques reside in the extracellular matrix. The results of this study demonstrate that the SAβO nanobody can serve as a diagnostic agent with potential theragnostic applications in Alzheimer's disease.

Application of Machine Learning in the Quantitative Analysis of the Surface Characteristics of Highly Abundant Cytoplasmic Proteins: Toward AI-Based Biomimetics

Proteins in the crowded environment of human cells have often been studied regarding nonspecific interactions, misfolding, and aggregation, which may cause cellular malfunction and disease. Specifically, proteins with high abundance are more susceptible to these issues due to the law of mass action. Therefore, the surfaces of highly abundant cytoplasmic (HAC) proteins directly exposed to the environment can exhibit specific physicochemical, structural, and geometrical characteristics that reduce nonspecific interactions and adapt to the environment. However, the quantitative relationships between the overall surface descriptors still need clarification. Here, we used machine learning to identify HAC proteins using hydrophobicity, charge, roughness, secondary structures, and B-factor from the protein surfaces and quantified the contribution of each descriptor. First, several supervised learning algorithms were compared to solve binary classification problems for the surfaces of HAC and extracellular proteins. Then, logistic regression was used for the feature importance analysis of descriptors considering model performance (80.2% accuracy and 87.6% AUC) and interpretability. The HAC proteins showed positive correlations with negatively and positively charged areas but negative correlations with hydrophobicity, the B-factor, the proportion of beta structures, roughness, and the proportion of disordered regions. Finally, the details of each descriptor could be explained concerning adaptative surface strategies of HAC proteins to regulate nonspecific interactions, protein folding, flexibility, stability, and adsorption. This study presented a novel approach using various surface descriptors to identify HAC proteins and provided quantitative design rules for the surfaces well-suited to human cellular crowded environments.

Systematic enhancement of protein crystallization efficiency by bulk lysine-to-arginine (KR) substitution

Structural genomics consortia established that protein crystallization is the primary obstacle to structure determination using x-ray crystallography. We previously demonstrated that crystallization propensity is systematically related to primary sequence, and we subsequently performed computational analyses showing that arginine is the most overrepresented amino acid in crystal-packing interfaces in the Protein Data Bank. Given the similar physicochemical characteristics of arginine and lysine, we hypothesized that multiple lysine-to-arginine (KR) substitutions should improve crystallization. To test this hypothesis, we developed software that ranks lysine sites in a target protein based on the redundancy-corrected KR substitution frequency in homologs. This software can be run interactively on the worldwide web at https://www.pxengineering.org/. We demonstrate that three unrelated single-domain proteins can tolerate 5-11 KR substitutions with at most minor destabilization, and, for two of these three proteins, the construct with the largest number of KR substitutions exhibits significantly enhanced crystallization propensity. This approach rapidly produced a 1.9 Å crystal structure of a human protein domain refractory to crystallization with its native sequence. Structures from Bulk KR-substituted domains show the engineered arginine residues frequently make hydrogen-bonds across crystal-packing interfaces. We thus demonstrate that Bulk KR substitution represents a rational and efficient method for probabilistic engineering of protein surface properties to improve crystallization.

Predictive modeling and cryo-EM: A synergistic approach to modeling macromolecular structure

Over the last 15 years, structural biology has seen unprecedented development and improvement in two areas: electron cryo-microscopy (cryo-EM) and predictive modeling. Once relegated to low resolutions, single-particle cryo-EM is now capable of achieving near-atomic resolutions of a wide variety of macromolecular complexes. Ushered in by AlphaFold, machine learning has powered the current generation of predictive modeling tools, which can accurately and reliably predict models for proteins and some complexes directly from the sequence alone. Although they offer new opportunities individually, there is an inherent synergy between these techniques, allowing for the construction of large, complex macromolecular models. Here, we give a brief overview of these approaches in addition to illustrating works that combine these techniques for model building. These examples provide insight into model building, assessment, and limitations when integrating predictive modeling with cryo-EM density maps. Together, these approaches offer the potential to greatly accelerate the generation of macromolecular structural insights, particularly when coupled with experimental data.

Integrating cellular electron microscopy with multimodal data to explore biology across space and time

Biology spans a continuum of length and time scales. Individual experimental methods only glimpse discrete pieces of this spectrum but can be combined to construct a more holistic view. In this Review, we detail the latest advancements in volume electron microscopy (vEM) and cryo-electron tomography (cryo-ET), which together can visualize biological complexity across scales from the organization of cells in large tissues to the molecular details inside native cellular environments. In addition, we discuss emerging methodologies for integrating three-dimensional electron microscopy (3DEM) imaging with multimodal data, including fluorescence microscopy, mass spectrometry, single-particle analysis, and AI-based structure prediction. This multifaceted approach fills gaps in the biological continuum, providing functional context, spatial organization, molecular identity, and native interactions. We conclude with a perspective on incorporating diverse data into computational simulations that further bridge and extend length scales while integrating the dimension of time.

AlphaFold predictions are valuable hypotheses and accelerate but do not replace experimental structure determination

Artificial intelligence-based protein structure prediction methods such as AlphaFold have revolutionized structural biology. The accuracies of these predictions vary, however, and they do not take into account ligands, covalent modifications or other environmental factors. Here, we evaluate how well AlphaFold predictions can be expected to describe the structure of a protein by comparing predictions directly with experimental crystallographic maps. In many cases, AlphaFold predictions matched experimental maps remarkably closely. In other cases, even very high-confidence predictions differed from experimental maps on a global scale through distortion and domain orientation, and on a local scale in backbone and side-chain conformation. We suggest considering AlphaFold predictions as exceptionally useful hypotheses. We further suggest that it is important to consider the confidence in prediction when interpreting AlphaFold predictions and to carry out experimental structure determination to verify structural details, particularly those that involve interactions not included in the prediction.

Artificial intelligence in small molecule drug discovery from 2018 to 2023: Does it really work?

Utilizing artificial intelligence (AI) in drug design represents an advanced approach for identifying targets and developing new drugs. Integrating AI techniques significantly reduces the workload involved in drug development and enhances the efficiency of early-stage drug discovery. This review aims to present a comprehensive overview of the utilization of AI methods in the field of small drug design, with a specific focus on four key areas: protein structure prediction, molecular virtual screening, molecular design, and absorption, distribution, metabolism, excretion, and toxicity (ADMET) prediction. Additionally, the role and limitations of AI in drug development are explored, and the impact of AI on decision-making processes is studied. It is important to note that while AI can bring numerous benefits to the early stage of drug development, the direction and quality of decision-making should still be emphasized, as AI should be considered as a tool rather than a decisive factor.

DomainFit: Identification of Protein Domains in cryo-EM maps at Intermediate Resolution using AlphaFold2-predicted Models

Cryo-electron microscopy (cryo-EM) has revolutionized our understanding of macromolecular complexes, enabling high-resolution structure determination. With the paradigm shift to in situ structural biology recently driven by the ground-breaking development of cryo-focused ion beam milling and cryo-electron tomography, there are an increasing number of structures at sub-nanometer resolution of complexes solved directly within their cellular environment. These cellular complexes often contain unidentified proteins, related to different cellular states or processes. Identifying proteins at resolutions lower than 4 Å remains challenging because the side chains cannot be visualized reliably. Here, we present DomainFit, a program for automated domain-level protein identification from cryo-EM maps at resolutions lower than 4 Å. By fitting domains from artificial intelligence-predicted models such as AlphaFold2-predicted models into cryo-EM maps, the program performs statistical analyses and attempts to identify the proteins forming the density. Using DomainFit, we identified two microtubule inner proteins, one of them, a CCDC81 domain-containing protein, is exclusively localized in the proximal region of the doublet microtubule from the ciliate Tetrahymena thermophila. The flexibility and capability of DomainFit makes it a valuable tool for analyzing in situ structures.

Automated model building and protein identification in cryo-EM maps

Interpreting electron cryo-microscopy (cryo-EM) maps with atomic models requires high levels of expertise and labour-intensive manual intervention. We present ModelAngelo, a machine-learning approach for automated atomic model building in cryo-EM maps. By combining information from the cryo-EM map with information from protein sequence and structure in a single graph neural network, ModelAngelo builds atomic models for proteins that are of similar quality as those generated by human experts. For nucleotides, ModelAngelo builds backbones with similar accuracy as humans. By using its predicted amino acid probabilities for each residue in hidden Markov model sequence searches, ModelAngelo outperforms human experts in the identification of proteins with unknown sequences. ModelAngelo will thus remove bottlenecks and increase objectivity in cryo-EM structure determination.

Inter-species association mapping links splice site evolution to METTL16 and SNRNP27K

Eukaryotic genes are interrupted by introns that are removed from transcribed RNAs by splicing. Patterns of splicing complexity differ between species, but it is unclear how these differences arise. We used inter-species association mapping with Saccharomycotina species to correlate splicing signal phenotypes with the presence or absence of splicing factors. Here, we show that variation in 5' splice site sequence preferences correlate with the presence of the U6 snRNA N6-methyladenosine methyltransferase METTL16 and the splicing factor SNRNP27K. The greatest variation in 5' splice site sequence occurred at the +4 position and involved a preference switch between adenosine and uridine. Loss of METTL16 and SNRNP27K orthologs, or a single SNRNP27K methionine residue, was associated with a preference for +4 U. These findings are consistent with splicing analyses of mutants defective in either METTL16 or SNRNP27K orthologs and models derived from spliceosome structures, demonstrating that inter-species association mapping is a powerful orthogonal approach to molecular studies. We identified variation between species in the occurrence of two major classes of 5' splice sites, defined by distinct interaction potentials with U5 and U6 snRNAs, that correlates with intron number. We conclude that variation in concerted processes of 5' splice site selection by U6 snRNA is associated with evolutionary changes in splicing signal phenotypes.

Predicted models and CCP4

In late 2020, the results of CASP14, the 14th event in a series of competitions to assess the latest developments in computational protein structure-prediction methodology, revealed the giant leap forward that had been made by Google's Deepmind in tackling the prediction problem. The level of accuracy in their predictions was the first instance of a competitor achieving a global distance test score of better than 90 across all categories of difficulty. This achievement represents both a challenge and an opportunity for the field of experimental structural biology. For structure determination by macromolecular X-ray crystallography, access to highly accurate structure predictions is of great benefit, particularly when it comes to solving the phase problem. Here, details of new utilities and enhanced applications in the CCP4 suite, designed to allow users to exploit predicted models in determining macromolecular structures from X-ray diffraction data, are presented. The focus is mainly on applications that can be used to solve the phase problem through molecular replacement.

AlphaFold-predicted protein structures and small-angle X-ray scattering: insights from an extended examination of selected data in the Small-Angle Scattering Biological Data Bank

By providing predicted protein structures from nearly all known protein sequences, the artificial intelligence program AlphaFold (AF) is having a major impact on structural biology. While a stunning accuracy has been achieved for many folding units, predicted unstructured regions and the arrangement of potentially flexible linkers connecting structured domains present challenges. Focusing on single-chain structures without prosthetic groups, an earlier comparison of features derived from small-angle X-ray scattering (SAXS) data taken from the Small-Angle Scattering Biological Data Bank (SASBDB) is extended to those calculated using the corresponding AF-predicted structures. Selected SASBDB entries were carefully examined to ensure that they represented data from monodisperse protein solutions and had sufficient statistical precision and q resolution for reliable structural evaluation. Three examples were identified where there is clear evidence that the single AF-predicted structure cannot account for the experimental SAXS data. Instead, excellent agreement is found with ensemble models generated by allowing for flexible linkers between high-confidence predicted structured domains. A pool of representative structures was generated using a Monte Carlo method that adjusts backbone dihedral allowed angles along potentially flexible regions. A fast ensemble modelling method was employed that optimizes the fit of pair distance distribution functions [P(r) versus r] and intensity profiles [I(q) versus q] computed from the pool to their experimental counterparts. These results highlight the complementarity between AF prediction, solution SAXS and molecular dynamics/conformational sampling for structural modelling of proteins having both structured and flexible regions.

Beyond the US-SOMO-AF database: a new website for hydrodynamic, structural, and circular dichroism calculations on user-supplied structures

At the 25th International Analytical Ultracentrifugation Workshop and Symposium, we described the recent implementation of the UltraScan SOlution MOdeler AlphaFold (US-SOMO-AF) database, containing hydrodynamic, structural, CD calculations, and other ancillary information, performed on the entire AF v2 database of predicted protein structures, containing more than 1,000,000 entries. The scope of the US-SOMO-AF database was that of providing direct access to pre-calculated physicochemical parameters for rapid assessment against their experimentally determined counterparts to test the compatibility in solution of predicted AlphaFold structures. In the meantime, the AlphaFold consortium has extended its database of predicted structures to an astonishing > 200 million entries, making it quite impractical for their coverage in the US-SOMO-AF database. Therefore, we have created the US-SOMO-Web site, allowing the rapid calculations of all the properties, as present in the US-SOMO-AF database, on user-supplied PDB and mmCIF structures, as well as allowing direct processing of the latest AlphaFold models. Major features on the website are described, along with current limitations and potential future developments.

Structural insights into perilipin 3 membrane association in response to diacylglycerol accumulation

Lipid droplets (LDs) are dynamic organelles that contain an oil core mainly composed of triglycerides (TAG) that is surrounded by a phospholipid monolayer and LD-associated proteins called perilipins (PLINs). During LD biogenesis, perilipin 3 (PLIN3) is recruited to nascent LDs as they emerge from the endoplasmic reticulum. Here, we analyze how lipid composition affects PLIN3 recruitment to membrane bilayers and LDs, and the structural changes that occur upon membrane binding. We find that the TAG precursors phosphatidic acid and diacylglycerol (DAG) recruit PLIN3 to membrane bilayers and define an expanded Perilipin-ADRP-Tip47 (PAT) domain that preferentially binds DAG-enriched membranes. Membrane binding induces a disorder to order transition of alpha helices within the PAT domain and 11-mer repeats, with intramolecular distance measurements consistent with the expanded PAT domain adopting a folded but dynamic structure upon membrane binding. In cells, PLIN3 is recruited to DAG-enriched ER membranes, and this requires both the PAT domain and 11-mer repeats. This provides molecular details of PLIN3 recruitment to nascent LDs and identifies a function of the PAT domain of PLIN3 in DAG binding.

The CCP4 suite: integrative software for macromolecular crystallography

The Collaborative Computational Project No. 4 (CCP4) is a UK-led international collective with a mission to develop, test, distribute and promote software for macromolecular crystallography. The CCP4 suite is a multiplatform collection of programs brought together by familiar execution routines, a set of common libraries and graphical interfaces. The CCP4 suite has experienced several considerable changes since its last reference article, involving new infrastructure, original programs and graphical interfaces. This article, which is intended as a general literature citation for the use of the CCP4 software suite in structure determination, will guide the reader through such transformations, offering a general overview of the new features and outlining future developments. As such, it aims to highlight the individual programs that comprise the suite and to provide the latest references to them for perusal by crystallographers around the world.

Recent breakthroughs in computational structural biology harnessing the power of sequences and structures

Recent advances in computational approaches and their integration into structural biology enable tackling increasingly complex questions. Here, we discuss several key areas, highlighting breakthroughs and remaining challenges. Theoretical modeling has provided tools to accurately predict and design protein structures on a scale currently difficult to achieve using experimental approaches. Molecular Dynamics simulations have become faster and more precise, delivering actionable information inaccessible by current experimental methods. Virtual screening workflows allow a high-throughput approach to discover ligands that bind and modulate protein function, while Machine Learning methods enable the design of proteins with new functionalities. Integrative structural biology combines several of these approaches, pushing the frontiers of structural and functional characterization to ever larger systems, advancing towards a complete understanding of the living cell. These breakthroughs will accelerate and significantly impact diverse areas of science.

Protein structure-based evaluation of missense variants: Resources, challenges and future directions

We provide an overview of the methods that can be used for protein structure-based evaluation of missense variants. The algorithms can be broadly divided into those that calculate the difference in free energy (ΔΔG) between the wild type and variant structures and those that use structural features to predict the damaging effect of a variant without providing a ΔΔG. A wide range of machine learning approaches have been employed to develop those algorithms. We also discuss challenges and opportunities for variant interpretation in view of the recent breakthrough in three-dimensional structural modelling using deep learning.

Likelihood-based docking of models into cryo-EM maps

Optimized docking of models into cryo-EM maps requires exploiting an understanding of the signal expected in the data to minimize the calculation time while maintaining sufficient signal. The likelihood-based rotation function used in crystallography can be employed to establish plausible orientations in a docking search. A phased likelihood translation function yields scores for the placement and rigid-body refinement of oriented models. Optimized strategies for choices of the resolution of data from the cryo-EM maps to use in the calculations and the size of search volumes are based on expected log-likelihood-gain scores computed in advance of the search calculation. Tests demonstrate that the new procedure is fast, robust and effective at placing models into even challenging cryo-EM maps.

Accelerating crystal structure determination with iterative AlphaFold prediction

Experimental structure determination can be accelerated with artificial intelligence (AI)-based structure-prediction methods such as AlphaFold. Here, an automatic procedure requiring only sequence information and crystallographic data is presented that uses AlphaFold predictions to produce an electron-density map and a structural model. Iterating through cycles of structure prediction is a key element of this procedure: a predicted model rebuilt in one cycle is used as a template for prediction in the next cycle. This procedure was applied to X-ray data for 215 structures released by the Protein Data Bank in a recent six-month period. In 87% of cases our procedure yielded a model with at least 50% of C^α atoms matching those in the deposited models within 2 Å. Predictions from the iterative template-guided prediction procedure were more accurate than those obtained without templates. It is concluded that AlphaFold predictions obtained based on sequence information alone are usually accurate enough to solve the crystallographic phase problem with molecular replacement, and a general strategy for macromolecular structure determination that includes AI-based prediction both as a starting point and as a method of model optimization is suggested.