Kala Bharath Pilla

Building an accurate atomic structure model of a protein into a cryo-electron microscopy (cryo-EM) map at worse than 3 Å resolution is difficult. To facilitate this task, we devised a method for assigning the amino acid residue sequence to the backbone fragments traced in an input cryo-EM map (EMSequenceFinder). EMSequenceFinder relies on a Bayesian scoring function for ranking 20 standard amino acid residue types at a given backbone position, based on the fit to a density map, map resolution, and secondary structure propensity. The fit to a density is quantified by a convolutional neural network that was trained on ∼5.56 million amino acid residue densities extracted from cryo-EM maps at 3-10 Å resolution and corresponding atomic structure models deposited in the Electron Microscopy Data Bank (EMDB). We benchmarked EMSequenceFinder by predicting the sequences of 58,044 distinct L-helix and β-strand fragments, given the fragment backbone coordinates fitted in their density maps. EMSequenceFinder identifies the correct sequence as the best-scoring sequence in 77.8% of these cases. We also assessed EMSequenceFinder on separate datasets of cryo-EM maps at resolutions from 4 to 6 L. The accuracy of EMSequenceFinder (63.5%) was better than that of three tested state-of-the-art methods, including findMysequence (45%), ModelAngelo (27%), and sequence_from_map in Phenix (12.9%). We further illustrate EMSequenceFinder by threading the SARS-CoV-2 NSP2 sequence into eight cryo-EM maps at resolutions from 3.7 to 7.0 Å. EMSequenceFinder is implemented in our open-source Integrative Modeling Platform (IMP) program. Thus, it is expected to be helpful for integrative structure modeling based on a cryo-EM map and other information, such as models of protein complex components and chemical crosslinks between them. EMSequenceFinder is available as part of our open source IMP distribution at https://integrativemodeling.org/.

Comprehensive modeling of a whole cell requires an integration of vast amounts of information on various aspects of the cell and its parts. To divide-and-conquer this task, we introduce Bayesian metamodeling, a general approach to modeling complex systems by integrating a collection of heterogeneous input models. Each input model can in principle be based on any type of data and can describe a different aspect of the modeled system using any mathematical representation, scale, and level of granularity. These input models are (i) converted to a standardized statistical representation relying on Probabilistic Graphical Models, (ii) coupled by modeling their mutual relations with the physical world, and (iii) finally harmonized with respect to each other. To illustrate Bayesian metamodeling, we provide a proof-of-principle metamodel of glucose-stimulated insulin secretion by human pancreatic ß-cells. The input models include a coarse-grained spatiotemporal simulation of insulin vesicle trafficking, docking, and exocytosis; a molecular network model of glucose-stimulated insulin secretion signaling; a network model of insulin metabolism; a structural model of glucagon-like peptide-1 receptor activation; a linear model of a pancreatic cell population; and ordinary differential equations for systemic postprandial insulin response. Metamodeling benefits from decentralized computing, while often producing a more accurate, precise, and complete model that contextualizes input models as well as resolves conflicting information. We anticipate Bayesian metamodeling will facilitate collaborative science by providing a framework for sharing expertise, resources, data, and models, as exemplified by the Pancreatic ß-Cell Consortium. Significance Statement Cells are the basic units of life, yet their architecture and function remain to be fully characterized. This work describes Bayesian metamodeling, a modeling approach that divides-and-conquers a large problem of modeling numerous aspects of the cell into computing a number of smaller models of different types, followed by assembling these models into a complete map of the cell. Metamodeling enables a facile collaboration of multiple research groups and communities, thus maximizing the sharing of expertise, resources, data, and models. A proof-of-principle is provided by a model of glucose-stimulated insulin secretion produced by the Pancreatic ß-Cell Consortium.

A newly described coronavirus named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which is the causative agent of coronavirus disease 2019 (COVID-19), has infected over 2.3 million people, led to the death of more than 160,000 individuals and caused worldwide social and economic disruption^1,2. There are no antiviral drugs with proven clinical efficacy for the treatment of COVID-19, nor are there any vaccines that prevent infection with SARS-CoV-2, and efforts to develop drugs and vaccines are hampered by the limited knowledge of the molecular details of how SARS-CoV-2 infects cells. Here we cloned, tagged and expressed 26 of the 29 SARS-CoV-2 proteins in human cells and identified the human proteins that physically associated with each of the SARS-CoV-2 proteins using affinity-purification mass spectrometry, identifying 332 high-confidence protein–protein interactions between SARS-CoV-2 and human proteins. Among these, we identify 66 druggable human proteins or host factors targeted by 69 compounds (of which, 29 drugs are approved by the US Food and Drug Administration, 12 are in clinical trials and 28 are preclinical compounds). We screened a subset of these in multiple viral assays and found two sets of pharmacological agents that displayed antiviral activity: inhibitors of mRNA translation and predicted regulators of the sigma-1 and sigma-2 receptors. Further studies of these host-factor-targeting agents, including their combination with drugs that directly target viral enzymes, could lead to a therapeutic regimen to treat COVID-19.

An outbreak of the novel coronavirus SARS-CoV-2, the causative agent of COVID-19 respiratory disease, has infected over 290,000 people since the end of 2019, killed over 12,000, and caused worldwide social and economic disruption ^1,2 . There are currently no antiviral drugs with proven efficacy nor are there vaccines for its prevention. Unfortunately, the scientific community has little knowledge of the molecular details of SARS-CoV-2 infection. To illuminate this, we cloned, tagged and expressed 26 of the 29 viral proteins in human cells and identified the human proteins physically associated with each using affinity-purification mass spectrometry (AP-MS), which identified 332 high confidence SARS-CoV-2-human protein-protein interactions (PPIs). Among these, we identify 67 druggable human proteins or host factors targeted by 69 existing FDA-approved drugs, drugs in clinical trials and/or preclinical compounds, that we are currently evaluating for efficacy in live SARS-CoV-2 infection assays. The identification of host dependency factors mediating virus infection may provide key insights into effective molecular targets for developing broadly acting antiviral therapeutics against SARS-CoV-2 and other deadly coronavirus strains.

The 3D fold determination of proteins by computational algorithms guided by experimental restraints is a reliable and efficient approach. However, the current algorithms struggle with sampling the conformational space and scaling in performance with the increase in the size of the proteins. This paper demonstrates a new data-driven, time-efficient, heuristics algorithm that assembles the 3D structure of a protein from its elemental super-secondary structure motifs (Smotifs) using a limited number of nuclear magnetic resonance (NMR) derived restraints. The DINGO-NOE-RDC algorithm (3D assembly of Individual smotifs to Near-native Geometry as Orchestrated by limited nuclear Overhauser effects (NOE) and residual dipolar couplings (RDC)) leverages on the distance restraints recorded on methyl-methyl (CH ₃ -CH ₃ ), methyl-amide (CH ₃ -H ^N ), and amide-amide (H ^N –H ^N ) NOE contacts, and orientation restraints recorded via RDC on the backbone amide protons, to assemble the target’s Smotifs. Two conceptual advancements were made to bootstrap the structure determination from limited NMR restraints; first, expand the basic definition of a ‘Smotif’ and, second, employ a data driven approach for selection, scoring, ranking and clustering of Smotif assemblies. In contrast to existing methods, the DINGO-NOE-RDC algorithm does not use any force-fields, physical/empirical scoring functions or the target’s amino-acid sequence makeup. Additionally, the algorithm employs a universal Smotif library that applies to any target protein and, can generate numerically reproducible results. For a benchmark set of ten different targets with different topologies, ranging from 100-200 residues, the algorithm identified near-native Smotifs in all of them.

Computational modeling of proteins using evolutionary or de novo approaches offers rapid structural characterization, but often suffers from low success rates in generating high quality models comparable to the accuracy of structures observed in X-ray crystallography or nuclear magnetic resonance (NMR) spectroscopy. A computational/experimental hybrid approach incorporating sparse experimental restraints in computational modeling algorithms drastically improves reliability and accuracy of 3D models. This chapter discusses the use of structural information obtained from various paramagnetic NMR measurements and demonstrates computational algorithms implementing pseudocontact shifts as restraints to determine the structure of proteins at atomic resolution.

Structure-guided drug design relies on detailed structural knowledge of protein-ligand complexes, but crystallization of co-complexes is not always possible. Here we present a sensitive nuclear magnetic resonance (NMR) approach to determine the binding mode of tightly binding lead compounds in complex with difficult target proteins. In contrast to established NMR methods, it does not depend on rapid exchange between bound and free ligand or on stable isotope labeling, relying instead on a tert-butyl group as a chemical label. Tert-butyl groups are found in numerous protein ligands and deliver an exceptionally narrow and tall 1H NMR signal. We show that a tert-butyl group also pro-duces outstandingly intense intra- and intermolecular NOESY cross-peaks. These enable measurements of pseu-docontact shifts generated by lanthanide tags attached to the protein, which in turn allows positioning of the ligand on the protein. Once the ligand has been located, assignments of intermolecular NOEs become possible even without prior resonance assignments of protein side chains. The approach is demonstrated with the dengue virus NS2B-NS3 protease in complex with a high-affinity ligand containing a tert-butyl group.