MILCDock: Machine Learning Enhanced Consensus Docking for Virtual Screening in Drug Discovery.

Molecular docking tools are regularly used to computationally identify new molecules in virtual screening for drug discovery. However, docking tools suffer from inaccurate scoring functions with widely varying performance on different proteins. To enable more accurate ranking of active over inactive ligands in virtual screening, we created a machine learning consensus docking tool, MILCDock, that uses predictions from five traditional molecular docking tools to predict the probability a ligand binds to a protein. MILCDock was trained and tested on data from both the DUD-E and LIT-PCBA docking datasets and shows improved performance over traditional molecular docking tools and other consensus docking methods on the DUD-E dataset. LIT-PCBA targets proved to be difficult for all methods tested. We also find that DUD-E data, although biased, can be effective in training machine learning tools if care is taken to avoid DUD-E's biases during training.

Modeling SARS-CoV-2 proteins in the CASP-commons experiment.

Critical Assessment of Structure Prediction (CASP) is an organization aimed at advancing the state of the art in computing protein structure from sequence. In the spring of 2020, CASP launched a community project to compute the structures of the most structurally challenging proteins coded for in the SARS-CoV-2 genome. Forty-seven research groups submitted over 3000 three-dimensional models and 700 sets of accuracy estimates on 10 proteins. The resulting models were released to the public. CASP community members also worked together to provide estimates of local and global accuracy and identify structure-based domain boundaries for some proteins. Subsequently, two of these structures (ORF3a and ORF8) have been solved experimentally, allowing assessment of both model quality and the accuracy estimates. Models from the AlphaFold2 group were found to have good agreement with the experimental structures, with main chain GDT_TS accuracy scores ranging from 63 (a correct topology) to 87 (competitive with experiment).

Evaluation of Deep Neural Network ProSPr for Accurate Protein Distance Predictions on CASP14 Targets.

The field of protein structure prediction has recently been revolutionized through the introduction of deep learning. The current state-of-the-art tool AlphaFold2 can predict highly accurate structures; however, it has a prohibitively long inference time for applications that require the folding of hundreds of sequences. The prediction of protein structure annotations, such as amino acid distances, can be achieved at a higher speed with existing tools, such as the ProSPr network. Here, we report on important updates to the ProSPr network, its performance in the recent Critical Assessment of Techniques for Protein Structure Prediction (CASP14) competition, and an evaluation of its accuracy dependency on sequence length and multiple sequence alignment depth. We also provide a detailed description of the architecture and the training process, accompanied by reusable code. This work is anticipated to provide a solid foundation for the further development of protein distance prediction tools.

Context-Dependent Stabilizing Interactions among Solvent-Exposed Residues along the Surface of a Trimeric Helix Bundle.

Here we show that a solvent-exposed f-position (i.e., residue 14) within a well-characterized trimeric helix bundle can facilitate a stabilizing long-range synergistic interaction involving b-position Glu10 (i.e., i - 4 relative to residue 14) and c-position Lys18 (i.e., i + 4), depending the identity of residue 14. The extent of stabilization associated with the Glu10-Lys18 pair depends primarily on the presence of a side-chain hydrogen-bond donor at residue 14; the nonpolar or hydrophobic character of residue 14 plays a smaller but still significant role. Crystal structures and molecular dynamics simulations indicate that Glu10 and Lys18 do not interact directly with each other but suggest the possibility that the proximity of residue 14 with Lys18 allows Glu10 to interact favorably with nearby Lys7. Subsequent thermodynamic experiments confirm the important role of Lys7 in the large synergistic stabilization associated with the Glu10-Lys18 pair. Our results highlight the exquisite complexity and surprising long-range synergistic interactions among b-, c-, and f-position residues within helix bundles, suggesting new possibilities for engineering hyperstable helix bundles and emphasizing the need to consider carefully the impact of substitutions at these positions for application-specific purposes.

Protein structure refinement with adaptively restrained homologous replicas.

A novel protein refinement protocol is presented which utilizes molecular dynamics (MD) simulations of an ensemble of adaptively restrained homologous replicas. This approach adds evolutionary information to the force field and reduces random conformational fluctuations by coupling of several replicas. It is shown that this protocol refines the majority of models from the CASP11 refinement category and that larger conformational changes of the starting structure are possible than with current state of the art methods. The performance of this protocol in the CASP11 experiment is discussed. We found that the quality of the refined model is correlated with the structural variance of the coupled replicas, which therefore provides a good estimator of model quality. Furthermore, some remarkable refinement results are discussed in detail. Proteins 2016; 84(Suppl 1):302-313. © 2015 Wiley Periodicals, Inc.

TrIP─Transformer Interatomic Potential Predicts Realistic Energy Surface Using Physical Bias.

Accurate interatomic energies and forces enable high-quality molecular dynamics simulations, torsion scans, potential energy surface mappings, and geometry optimizations. Machine learning algorithms have enabled rapid estimates of the energies and forces with high accuracy. Further development of machine learning algorithms holds promise for producing universal potentials that support many different atomic species. We present the Transformer Interatomic Potential (TrIP): a chemically sound potential based on the SE(3)-Transformer. TrIP's species-agnostic architecture, which uses continuous atomic representation and homogeneous graph convolutions, encourages parameter sharing between atomic species for more general representations of chemical environments, maintains a reasonable number of parameters, serves as a form of regularization, and is a step toward accurate universal interatomic potentials. TrIP achieves state-of-the-art accuracies on the COMP6 benchmark with an energy prediction of just 1.02 kcal/mol MAE. We introduce physical bias in the form of Ziegler-Biersack-Littmark-screened nuclear repulsion and constrained atomization energies. An energy scan of a water molecule demonstrates that these changes improve long- and near-range interactions compared to other neural network potentials. TrIP also demonstrates stability in molecular dynamics simulations, demonstrating reasonable exploration of Ramachandran space.

Rise of the Allotrope Simple Model: Update from 2023 Fall Allotrope Connect.

The Allotrope Foundation (AF) started as a group of pharmaceutical companies, instrument, and software vendors that set out to simplify the exchange of data in the laboratory. After a decade of work, they released products that have found adoption in various companies. Most recently, the Allotrope Simple Model (ASM) was developed to speed up and widen the adoption. As a result, the Foundation has recently added chemical companies and, importantly, is reworking its business model to lower the entry barrier for smaller companies. Here, we present the proceedings from the Allotrope Connect Fall 2023 conference and summarize the technical and organizational developments at the Foundation since 2020.

Development of a national database for dietary glycemic index and load for nutritional epidemiologic studies in the United States.

BACKGROUND: The quality of carbohydrate intake, as measured by the glycemic index (GI), has not been evaluated nationally over the past 2 decades in the United States. OBJECTIVES: We aimed to develop a comprehensive and nationally representative dietary GI and glycemic load (GL) database from 1999 to 2018 National Health and Nutrition Examination Survey (NHANES) and to examine GI and GL time trends and subpopulation differences. METHODS: We used an artificial intelligence (AI)-enabled model to match GI values from 2 GI databases to food codes from United States Department of Agriculture, which were manually validated. We examined nationally representative distributions of dietary GI and GL from 1999 to 2018 using the multistage, clustered sampling design of NHANES. RESULTS: Assigned GI values covered 99.9% of total carbohydrate intake. The initial AI accuracy was 75.0%, with 31.3% retained after manual curation guided by substantive domain expertise. A total of 7976 unique food codes were matched to GI values, of which soft drinks and white bread were top contributors to dietary GI and GL. Of the 49,205 NHANES adult participants, the mean dietary GI was 55.7 (95% confidence interval [CI]: 55.5, 55.8) and energy-adjusted dietary GL was 133.0 (95% CI: 132.3, 133.8). From 1999 to 2018, dietary GI and GL decreased by 4.6% and 13.8%, respectively. Dietary GL was higher among females (134.6; 95% CI: 133.8, 135.5) than among males (131.3; 95% CI: 130.3, 132.3), those with ≤high school degree (137.7; 95% CI: 136.8, 138.7) than among those with ≥college degree (126.5; 95% CI: 125.3, 127.7), and those living under the poverty level (140.9; 95% CI: 139.6, 142.1) than among those above the poverty level. Differences in race were observed (Black adults, 139.4; 95% CI: 138.2, 140.7; White adults, 131.6; 95% CI: 130.5, 132.6). CONCLUSIONS: The national GI and GL database facilitates large-scale and high-quality surveillance or cohort studies of diet and health outcomes in the United States.

A probabilistic view of protein stability, conformational specificity, and design.

Various approaches have used neural networks as probabilistic models for the design of protein sequences. These "inverse folding" models employ different objective functions, which come with trade-offs that have not been assessed in detail before. This study introduces probabilistic definitions of protein stability and conformational specificity and demonstrates the relationship between these chemical properties and the [Formula: see text] Boltzmann probability objective. This links the Boltzmann probability objective function to experimentally verifiable outcomes. We propose a novel sequence decoding algorithm, referred to as "BayesDesign", that leverages Bayes' Rule to maximize the [Formula: see text] objective instead of the [Formula: see text] objective common in inverse folding models. The efficacy of BayesDesign is evaluated in the context of two protein model systems, the NanoLuc enzyme and the WW structural motif. Both BayesDesign and the baseline ProteinMPNN algorithm increase the thermostability of NanoLuc and increase the conformational specificity of WW. The possible sources of error in the model are analyzed.

Computational methods to simulate molten salt thermophysical properties.

Molten salts are important thermal conductors used in molten salt reactors and solar applications. To use molten salts safely, accurate knowledge of their thermophysical properties is necessary. However, it is experimentally challenging to measure these properties and a comprehensive evaluation of the full chemical space is unfeasible. Computational methods provide an alternative route to access these properties. Here, we summarize the developments in methods over the last 70 years and cluster them into three relevant eras. We review the main advances and limitations of each era and conclude with an optimistic perspective for the next decade, which will likely be dominated by emerging machine learning techniques. This article is aimed to help researchers in peripheral scientific domains understand the current challenges of molten salt simulation and identify opportunities to contribute.

Self-reporting data assets and their representation in the pharmaceutical industry.

Standardizing data is crucial for preserving and exchanging scientific information. In particular, recording the context in which data were created ensures that information remains findable, accessible, interoperable, and reusable. Here, we introduce the concept of self-reporting data assets (SRDAs), which preserve data and contextual information. SRDAs are an abstract concept, which requires a suitable data format for implementation. Four promising data formats or languages are popularly used to represent data in pharma: JCAMP-DX, JSON, AnIML, and, more recently, the Allotrope Data Format (ADF). Here, we evaluate these four options in common use cases within the pharmaceutical industry using multiple criteria. The evaluation shows that ADF is the most suitable format for the implementation of SRDAs.

Training undergraduate research assistants with an outcome-oriented and skill-based mentoring strategy.

Effective mentoring of undergraduate students is a growing requirement for the promotion of faculty at many universities. It is often challenging for young investigators to define a successful mentoring strategy, partially due to the absence of a broadly accepted definition of what mentoring should entail. To overcome this, an outcome-oriented mentoring framework was developed and used with more than 25 students over three years. It was found that a systematic mentoring approach can help students quickly realize their scientific potential and result in meaningful contributions to science. This report especially shows how the Critical Assessment of Protein Structure Prediction (CASP14) challenge was used to amplify student research efforts. As a result of this challenge, multiple publications, presentations and scholarships were awarded to the participating students. The mentoring framework continues to see much success in allowing undergraduate students, including students from underrepresented groups, to foster scientific talent and make meaningful contributions to the scientific community.

The whole is greater than its parts: ensembling improves protein contact prediction.

The prediction of amino acid contacts from protein sequence is an important problem, as protein contacts are a vital step towards the prediction of folded protein structures. We propose that a powerful concept from deep learning, called ensembling, can increase the accuracy of protein contact predictions by combining the outputs of different neural network models. We show that ensembling the predictions made by different groups at the recent Critical Assessment of Protein Structure Prediction (CASP13) outperforms all individual groups. Further, we show that contacts derived from the distance predictions of three additional deep neural networks-AlphaFold, trRosetta, and ProSPr-can be substantially improved by ensembling all three networks. We also show that ensembling these recent deep neural networks with the best CASP13 group creates a superior contact prediction tool. Finally, we demonstrate that two ensembled networks can successfully differentiate between the folds of two highly homologous sequences. In order to build further on these findings, we propose the creation of a better protein contact benchmark set and additional open-source contact prediction methods.

Coming of age of Allotrope: Proceedings from the Fall 2020 Allotrope Connect.

The Allotrope Foundation (AF) is a group of pharmaceutical, device vendor, and software companies that develops and releases technologies [the Allotrope Data Format (ADF), the Allotrope Foundation Ontology (AFO), and the Allotrope Data Models (ADM)] to simplify the exchange of electronic data. We present here the first comprehensive history of the AF, its structure, a list of members and partners, and an introduction to the technologies. Finally, we provide current insights into the adoption and development of the technologies by summarizing the Fall 2020 Allotrope Connect virtual conference. This overview provides an easy access to the AF and highlights opportunities for collaboration.

World trends in sugar-sweetened beverage and dietary sugar intakes in children and adolescents: a systematic review.

OBJECTIVE: To provide a systematic overview of world dietary sugar and sugar-sweetened beverage (SSB) intake trends in children and adolescents. DATA SOURCES: Medline, Embase, and the Cochrane Central Register of Controlled Trials in the Cochrane Library were searched through January 2019 to identify longitudinal follow-up studies with time-trend data and repeated cross-sectional studies. DATA EXTRACTION: Data from studies reporting ≥ 2 measurements (sugars, SSB, or sweets/candy) over ≥ 2 years and included ≥ 20 healthy, normal- or overweight children or adolescents aged 1-19 years. DATA ANALYSIS: Data from 43 articles (n = 4 prospective cohort studies; n = 39 repeated cross-sectional studies) from 15 countries (n = 8 European countries plus Australia, Canada, China, South Korea, Mexico, Russia, and the United States) are presented narratively. According to the risk of bias in nonrandomized studies of interventions tool, 34 studies were judged to have a moderate risk of bias, and 5 to have a serious risk of bias. CONCLUSIONS: Consumption among US children and adolescents increased substantially in the decades preceding 2000, followed by a faster and continued decline. As a whole, other international intake trends did not reveal drastic increases and decreases in SSB and dietary sugars; they tended to change only slightly across 3 decades.

Engineering and application of a biosensor with focused ligand specificity.

Cell factories converting bio-based precursors to chemicals present an attractive avenue to a sustainable economy, yet screening of genetically diverse strain libraries to identify the best-performing whole-cell biocatalysts is a low-throughput endeavor. For this reason, transcriptional biosensors attract attention as they allow the screening of vast libraries when used in combination with fluorescence-activated cell sorting (FACS). However, broad ligand specificity of transcriptional regulators (TRs) often prohibits the development of such ultra-high-throughput screens. Here, we solve the structure of the TR LysG of Corynebacterium glutamicum, which detects all three basic amino acids. Based on this information, we follow a semi-rational engineering approach using a FACS-based screening/counterscreening strategy to generate an L-lysine insensitive LysG-based biosensor. This biosensor can be used to isolate L-histidine-producing strains by FACS, showing that TR engineering towards a more focused ligand spectrum can expand the scope of application of such metabolite sensors.

Integrated NMR, Fluorescence, and Molecular Dynamics Benchmark Study of Protein Mechanics and Hydrodynamics.

Understanding the function of a protein requires not only knowledge of its tertiary structure but also an understanding of its conformational dynamics. Nuclear magnetic resonance (NMR) spectroscopy, polarization-resolved fluorescence spectroscopy and molecular dynamics (MD) simulations are powerful methods to provide detailed insight into protein dynamics on multiple time scales by monitoring global rotational diffusion and local flexibility (order parameters) that are sensitive to inter- and intramolecular interactions, respectively. We present an integrated approach where data from these techniques are analyzed and interpreted within a joint theoretical description of depolarization and diffusion, demonstrating their conceptual similarities. This integrated approach is then applied to the autophagy-related protein GABARAP in its cytosolic form, elucidating its dynamics on the pico- to nanosecond time scale and its rotational and translational diffusion for protein concentrations spanning 9 orders of magnitude. We compare the dynamics of GABARAP as monitored by 15N spin relaxation of the backbone amide groups, fluorescence anisotropy decays and fluorescence correlation spectroscopy of side chains labeled with BODIPY FL, and molecular movies of the protein from MD simulations. The recovered parameters agree very well between the distinct techniques if the different measurement conditions (probe localization, sample concentration) are taken into account. Moreover, we propose a method that compares the order parameters of the backbone and side chains to identify potential hinges for large-scale, functionally relevant intradomain motions, such as residues 27/28 at the interface between the two subdomains of GABARAP. In conclusion, the integrated concept of cross-fertilizing techniques presented here is fundamental to obtaining a comprehensive quantitative picture of multiscale protein dynamics and solvation. The possibility to employ these validated techniques under cellular conditions and combine them with fluorescence imaging opens up the perspective of studying the functional dynamics of GABARAP or other proteins in live cells.

Coupling an Ensemble of Homologues Improves Refinement of Protein Homology Models.

Atomic models of proteins built by homology modeling or from low-resolution experimental data may contain considerable local errors. The refinement success of molecular dynamics simulations is usually limited by both force field accuracy and by the substantial width of the conformational distribution at physiological temperatures. We propose a method to overcome both these problems by coupling homologous replicas during a molecular dynamics simulation, which narrows the conformational distribution, and smoothens and even improves the energy landscape by adding evolutionary information.