Resolving the polycistronic aftermath: Essential role of topoisomerase IA in preventing R-loops in Leishmania.

Kinetoplastid parasites are "living bridges" in the evolution from prokaryotes to higher eukaryotes. The near-intronless genome of the kinetoplastid Leishmania exhibits polycistronic transcription which can facilitate R-loop formation. Therefore, to prevent such DNA-RNA hybrids, Leishmania has retained prokaryotic-like DNA Topoisomerase IA (LdTOPIA) in the course of evolution. LdTOPIA is an essential enzyme that is expressed ubiquitously and is adapted for the compartmentalized eukaryotic form in harboring functional bipartite nuclear localization signals. Although exhibiting greater homology to mycobacterial TOPIA, LdTOPIA could functionally complement the growth lethality of Escherichia coli TOPIA null GyrB ts strain at non-permissive temperatures. Purified LdTOPIA exhibits Mg2+-dependent relaxation of only negatively supercoiled DNA and preference towards single-stranded DNA substrates. LdTOPIA prevents nuclear R-loops as conditional LdTOPIA downregulated parasites exhibit R-loop formation and thereby parasite killing. The clinically used tricyclic antidepressant, norclomipramine could specifically inhibit LdTOPIA and lead to R-loop formation and parasite elimination. This comprehensive study therefore paves an avenue for drug repurposing against Leishmania.

A combination of potently neutralizing monoclonal antibodies isolated from an Indian convalescent donor protects against the SARS-CoV-2 Delta variant.

Although efficacious vaccines have significantly reduced the morbidity and mortality of COVID-19, there remains an unmet medical need for treatment options, which monoclonal antibodies (mAbs) can potentially fill. This unmet need is exacerbated by the emergence and spread of SARS-CoV-2 variants of concern (VOCs) that have shown some resistance to vaccine responses. Here we report the isolation of five neutralizing mAbs from an Indian convalescent donor, out of which two (THSC20.HVTR04 and THSC20.HVTR26) showed potent neutralization of SARS-CoV-2 VOCs at picomolar concentrations, including the Delta variant (B.1.617.2). One of these (THSC20.HVTR26) also retained activity against the Omicron variant. These two mAbs target non-overlapping epitopes on the receptor-binding domain (RBD) of the spike protein and prevent virus attachment to its host receptor, human angiotensin converting enzyme-2 (hACE2). Furthermore, the mAb cocktail demonstrated protection against the Delta variant at low antibody doses when passively administered in the K18 hACE2 transgenic mice model, highlighting their potential as a cocktail for prophylactic and therapeutic applications. Developing the capacity to rapidly discover and develop mAbs effective against highly transmissible pathogens like coronaviruses at a local level, especially in a low- and middle-income country (LMIC) such as India, will enable prompt responses to future pandemics as an important component of global pandemic preparedness.

Acceleration of Graph Neural Network-Based Prediction Models in Chemistry via Co-Design Optimization on Intelligence Processing Units.

Atomic structure prediction and associated property calculations are the bedrock of chemical physics. Since high-fidelity ab initio modeling techniques for computing the structure and properties can be prohibitively expensive, this motivates the development of machine-learning (ML) models that make these predictions more efficiently. Training graph neural networks over large atomistic databases introduces unique computational challenges, such as the need to process millions of small graphs with variable size and support communication patterns that are distinct from learning over large graphs, such as social networks. We demonstrate a novel hardware-software codesign approach to scale up the training of atomistic graph neural networks (GNN) for structure and property prediction. First, to eliminate redundant computation and memory associated with alternative padding techniques and to improve throughput via minimizing communication, we formulate the effective coalescing of the batches of variable-size atomistic graphs as the bin packing problem and introduce a hardware-agnostic algorithm to pack these batches. In addition, we propose hardware-specific optimizations, including a planner and vectorization for the gather-scatter operations targeted for Graphcore's Intelligence Processing Unit (IPU), as well as model-specific optimizations such as merged communication collectives and optimized softplus. Putting these all together, we demonstrate the effectiveness of the proposed codesign approach by providing an implementation of a well-established atomistic GNN on the Graphcore IPUs. We evaluate the training performance on multiple atomistic graph databases with varying degrees of graph counts, sizes, and sparsity. We demonstrate that such a codesign approach can reduce the training time of atomistic GNNs and can improve their performance by up to 1.5× compared to the baseline implementation of the model on the IPUs. Additionally, we compare our IPU implementation with a Nvidia GPU-based implementation and show that our atomistic GNN implementation on the IPUs can run 1.8× faster on average compared to the execution time on the GPUs.

A Small Step Toward Generalizability: Training a Machine Learning Scoring Function for Structure-Based Virtual Screening.

Over the past few years, many machine learning-based scoring functions for predicting the binding of small molecules to proteins have been developed. Their objective is to approximate the distribution which takes two molecules as input and outputs the energy of their interaction. Only a scoring function that accounts for the interatomic interactions involved in binding can accurately predict binding affinity on unseen molecules. However, many scoring functions make predictions based on data set biases rather than an understanding of the physics of binding. These scoring functions perform well when tested on similar targets to those in the training set but fail to generalize to dissimilar targets. To test what a machine learning-based scoring function has learned, input attribution, a technique for learning which features are important to a model when making a prediction on a particular data point, can be applied. If a model successfully learns something beyond data set biases, attribution should give insight into the important binding interactions that are taking place. We built a machine learning-based scoring function that aimed to avoid the influence of bias via thorough train and test data set filtering and show that it achieves comparable performance on the Comparative Assessment of Scoring Functions, 2016 (CASF-2016) benchmark to other leading methods. We then use the CASF-2016 test set to perform attribution and find that the bonds identified as important by PointVS, unlike those extracted from other scoring functions, have a high correlation with those found by a distance-based interaction profiler. We then show that attribution can be used to extract important binding pharmacophores from a given protein target when supplied with a number of bound structures. We use this information to perform fragment elaboration and see improvements in docking scores compared to using structural information from a traditional, data-based approach. This not only provides definitive proof that the scoring function has learned to identify some important binding interactions but also constitutes the first deep learning-based method for extracting structural information from a target for molecule design.

Femtosecond Hydration Map of Intrinsically Disordered α-Synuclein.

Protein hydration water plays a fundamentally important role in protein folding, binding, assembly, and function. Little is known about the hydration water in intrinsically disordered proteins that challenge the conventional sequence-structure-function paradigm. Here, by combining experiments and simulations, we show the existence of dynamical heterogeneity of hydration water in an intrinsically disordered presynaptic protein, namely α-synuclein, implicated in Parkinson's disease. We took advantage of nonoccurrence of cysteine in the sequence and incorporated a number of cysteine residues at the N-terminal segment, the central amyloidogenic nonamyloid-ß component (NAC) domain, and the C-terminal end of α-synuclein. We then labeled these cysteine variants using environment-sensitive thiol-active fluorophore and monitored the solvation dynamics using femtosecond time-resolved fluorescence. The site-specific femtosecond time-resolved experiments allowed us to construct the hydration map of α-synuclein. Our results show the presence of three dynamically distinct types of water: bulk, hydration, and confined water. The amyloidogenic NAC domain contains dynamically restrained water molecules that are strikingly different from the water molecules present in the other two domains. Atomistic molecular dynamics simulations revealed longer residence times for water molecules near the NAC domain and supported our experimental observations. Additionally, our simulations allowed us to decipher the molecular origin of the dynamical heterogeneity of water in α-synuclein. These simulations captured the quasi-bound water molecules within the NAC domain originating from a complex interplay between the local chain compaction and the sequence composition. Our findings from this synergistic experimental simulation approach suggest longer trapping of interfacial water molecules near the amyloidogenic hotspot that triggers the pathological conversion into amyloids via chain sequestration, chain desolvation, and entropic liberation of ordered water molecules.

How Hydrophobic Hydration Destabilizes Surfactant Micelles at Low Temperature: A Coarse-Grained Simulation Study.

Micelles are self-assembled aggregates of amphiphilic surfactant molecules that are important in a variety of applications, including drug delivery, detergency, and catalysis. It is known that the micellization process is driven by the same physiochemical forces that drive protein folding, aggregation, and biological membrane self-assembly. Nevertheless, the molecular details of how micelle stability changes in water at low temperature are not fully clear. We develop and use a coarse-grained model to investigate how the interplay between nonionic surfactants and the surrounding water at the nanoscale affects the stability of micelles at high and low temperatures. Simulations of preformed C12E5 micelles in explicit water at a range of temperatures reveal the existence of two distinct surfactant conformations within the micelle, a bent structure and an extended structure, the latter being more prevalent at low temperature. Favorable interactions of the surfactant with more ordered solvation water stabilizes the extended configuration, allowing nanoscale wetting of the dry, hydrophobic core of the micelle, leading to the micelle breaking. Taken together, our coarse-grained simulations unravel how energetic and structural changes of the surfactant and the surrounding water destabilize micelles at low temperature, which is a direct consequence of the weakened hydrophobicity. Our approach thus provides an effective mean for extracting the molecular-level changes during hydrophobicity-driven destabilization of molecular self-assembly, which is important in a wide range of fields, including biology, polymer science, and nanotechnology.

Emerging ß-Sheet Rich Conformations in Supercompact Huntingtin Exon-1 Mutant Structures.

There exists strong correlation between the extended polyglutamines (polyQ) within exon-1 of Huntingtin protein (Htt) and age onset of Huntington's disease (HD); however, the underlying molecular mechanism is still poorly understood. Here we apply extensive molecular dynamics simulations to study the folding of Htt-exon-1 across five different polyQ-lengths. We find an increase in secondary structure motifs at longer Q-lengths, including ß-sheet content that seems to contribute to the formation of increasingly compact structures. More strikingly, these longer Q-lengths adopt supercompact structures as evidenced by a surprisingly small power-law scaling exponent (0.22) between the radius-of-gyration and Q-length that is substantially below expected values for compact globule structures (∼0.33) and unstructured proteins (∼0.50). Hydrogen bond analyses further revealed that the supercompact behavior of polyQ is mainly due to the "glue-like" behavior of glutamine's side chains with significantly more side chain-side chain H-bonds than regular proteins in the Protein Data Bank (PDB). The orientation of the glutamine side chains also tend to be "buried" inside, explaining why polyQ domains are insoluble on their own.

A2T and A2V Aß peptides exhibit different aggregation kinetics, primary nucleation, morphology, structure, and LTP inhibition.

The histopathological hallmark of Alzheimer's disease (AD) is the aggregation and accumulation of the amyloid beta peptide (Aß) into misfolded oligomers and fibrils. Here we examine the biophysical properties of a protective Aß variant against AD, A2T, and a causative mutation, A2T, along with the wild type (WT) peptide. The main finding here is that the A2V native monomer is more stable than both A2T and WT, and this manifests itself in different biophysical behaviors: the kinetics of aggregation, the initial monomer conversion to an aggregation prone state (primary nucleation), the abundances of oligomers, and extended conformations. Aggregation reaction modeling of the conversion kinetics from native monomers to fibrils predicts the enhanced stability of the A2V monomer, while ion mobility spectrometry-mass spectrometry measures this directly confirming earlier predictions. Additionally, unique morphologies of the A2T aggregates are observed using atomic force microscopy, providing a basis for the reduction in long term potentiation inhibition of hippocampal cells for A2T compared with A2V and the wild type (WT) peptide. The stability difference of the A2V monomer and the difference in aggregate morphology for A2T (both compared with WT) are offered as alternate explanations for their pathological effects.

Alzheimer's protective A2T mutation changes the conformational landscape of the Aß1₋42 monomer differently than does the A2V mutation.

The aggregation of amyloid-ß (Aß) peptides plays a crucial role in the etiology of Alzheimer's disease (AD). Recently, it has been reported that an A2T mutation in Aß can protect against AD. Interestingly, a nonpolar A2V mutation also has been found to offer protection against AD in the heterozygous state, although it causes early-onset AD in homozygous carriers. Since the conformational landscape of the Aß monomer is known to directly contribute to the early-stage aggregation mechanism, it is important to characterize the effects of the A2T and A2V mutations on Aß1₋42 monomer structure. Here, we have performed extensive atomistic replica-exchange molecular dynamics simulations of the solvated wild-type (WT), A2V, and A2T Aß1₋42 monomers. Our simulations reveal that although all three variants remain as collapsed coils in solution, there exist significant structural differences among them at shorter timescales. A2V exhibits an enhanced double-hairpin population in comparison to the WT, similar to those reported in toxic WT Aß1₋42 oligomers. Such double-hairpin formation is caused by hydrophobic clustering between the N-terminus and the central and C-terminal hydrophobic patches. In contrast, the A2T mutation causes the N-terminus to engage in unusual electrostatic interactions with distant residues, such as K16 and E22, resulting in a unique population comprising only the C-terminal hairpin. These findings imply that a single A2X (where X = V or T) mutation in the primarily disordered N-terminus of the Aß1₋42 monomer can dramatically alter the ß-hairpin population and switch the equilibrium toward alternative structures. The atomistically detailed, comparative view of the structural landscapes of A2V and A2T variant monomers obtained in this study can enhance our understanding of the mechanistic differences in their early-stage aggregation.

Aggregation of γ-crystallins associated with human cataracts via domain swapping at the C-terminal ß-strands.

The prevalent eye disease age-onset cataract is associated with aggregation of human γD-crystallins, one of the longest-lived proteins. Identification of the γ-crystallin precursors to aggregates is crucial for developing strategies to prevent and reverse cataract. Our microseconds of atomistic molecular dynamics simulations uncover the molecular structure of the experimentally detected aggregation-prone folding intermediate species of monomeric native γD-crystallin with a largely folded C-terminal domain and a mostly unfolded N-terminal domain. About 30 residues including a, b, and c strands from the Greek Key motif 4 of the C-terminal domain experience strong solvent exposure of hydrophobic residues as well as partial unstructuring upon N-terminal domain unfolding. Those strands comprise the domain-domain interface crucial for unusually high stability of γD-crystallin. We further simulate the intermolecular linkage of these monomeric aggregation precursors, which reveals domain-swapped dimeric structures. In the simulated dimeric structures, the N-terminal domain of one monomer is frequently found in contact with residues 135-164 encompassing the a, b, and c strands of the Greek Key motif 4 of the second molecule. The present results suggest that γD-crystallin may polymerize through successive domain swapping of those three C-terminal ß-strands leading to age-onset cataract, as an evolutionary cost of its very high stability. Alanine substitutions of the hydrophobic residues in those aggregation-prone ß-strands, such as L145 and M147, hinder domain swapping as a pathway toward dimerization. These findings thus provide critical molecular insights onto the initial stages of age-onset cataract, which is important for understanding protein aggregation diseases.

ChatGPT: Literate or intelligent about UN sustainable development goals?

Generative AI tools, such as ChatGPT, are progressively transforming numerous sectors, demonstrating a capacity to impact human life dramatically. This research seeks to evaluate the UN Sustainable Development Goals (SDGs) literacy of ChatGPT, which is crucial for diverse stakeholders involved in SDG-related policies. Experimental outcomes from two widely used Sustainability Assessment tests-the UN SDG Fitness Test and Sustainability Literacy Test (SULITEST) - suggest that ChatGPT exhibits high SDG literacy, yet its comprehensive SDG intelligence needs further exploration. The Fitness Test gauges eight vital competencies across introductory, intermediate, and advanced levels. Accurate mapping of these to the test questions is essential for partial evaluation of SDG intelligence. To assess SDG intelligence, the questions from both tests were mapped to 17 SDGs and eight cross-cutting SDG core competencies, but both test questionnaires were found to be insufficient. SULITEST could satisfactorily map only 5 out of 8 competencies, whereas the Fitness Test managed to map 6 out of 8. Regarding the coverage of the Fitness Test and SULITEST, their mapping to the 17 SDGs, both tests fell short. Most SDGs were underrepresented in both instruments, with certain SDGs not represented at all. Consequently, both tools proved ineffective in assessing SDG intelligence through SDG coverage. The study recommends future versions of ChatGPT to enhance competencies such as collaboration, critical thinking, systems thinking, and others to achieve the SDGs. It concludes that while AI models like ChatGPT hold considerable potential in sustainable development, their usage must be approached carefully, considering current limitations and ethical implications.

AMPK activation reduces cancer cell aggressiveness via inhibition of monoamine oxidase A (MAO-A) expression/activity.

AIM: AMPK can be considered as an important target molecule for cancer for its unique ability to directly recognize cellular energy status. The main aim of this study is to explore the role of different AMPK activators in managing cancer cell aggressiveness and to understand the mechanistic details behind the process. MAIN METHODS: First, we explored the AMPK expression pattern and its significance in different subtypes of lung cancer by accessing the TCGA data sets for LUNG, LUAD and LUSC patients and then established the correlation between AMPK expression pattern and overall survival of lung cancer patients using Kaplan-Meire plot. We further carried out several cell-based assays by employing different wet lab techniques including RT-PCR, Western Blot, proliferation, migration and invasion assays to fulfil the aim of the study. KEY FINDINGS: SIGNIFICANCE: This study identifies the importance of AMPK activators as a repurposing agent for combating lung and colon cancer cell aggressiveness. It also suggests SRT-1720 as a potent repurposing agent for cancer treatment especially in NSCLC patients where a point mutation is present in LKB1.

Interplay between drying and stability of a TIM barrel protein: a combined simulation-experimental study.

Recent molecular dynamics simulations have suggested important roles for nanoscale dewetting in the stability, function, and folding dynamics of proteins. Using a synergistic simulation-experimental approach on the αTS TIM barrel protein, we validated this hypothesis by revealing the occurrence of drying inside hydrophobic amino acid clusters and its manifestation in experimental measures of protein stability and structure. Cavities created within three clusters of branched aliphatic amino acids [isoleucine, leucine, and valine (ILV) clusters] were found to experience strong water density fluctuations or intermittent dewetting transitions in simulations. Individually substituting 10 residues in the large ILV cluster at the N-terminus with less hydrophobic alanines showed a weakening or diminishing effect on dewetting that depended on the site of the mutation. Our simulations also demonstrated that replacement of buried leucines with isosteric, polar asparagines enhanced the wetting of the N- and C-terminal clusters. The experimental results on the stability, secondary structure, and compactness of the native and intermediate states for the asparagine variants are consistent with the preferential drying of the large N-terminal cluster in the intermediate. By contrast, the region encompassing the small C-terminal cluster experiences only partial drying in the intermediate, and its structure and stability are unaffected by the asparagine substitution. Surprisingly, the structural distortions required to accommodate the replacement of leucine by asparagine in the N-terminal cluster revealed the existence of alternative stable folds in the native basin. This combined simulation-experimental study demonstrates the critical role of drying within hydrophobic ILV clusters in the folding and stability of the αTS TIM barrel.

Effects of sequence and solvation on the temperature-pressure conformational landscape of proteinlike heteropolymers.

We study the role of sequence and solvation in shaping the temperature-pressure (T, P) conformational landscape of model heteropolymers with a coarse-grained model. We design foldable primarily hydrophobic sequences with fixed polar content in water at physiological conditions, which demonstrate (T, P) dependence of conformational stability similar to biological proteins. Inherent helicity emerges as a result of local polar-polar interactions in the sequences that mimic biological α helices. The helical propensity is reduced upon solvation and remains unaltered at cold T and high P, which is driven by the T-P induced changes of the hydration shell. Consequently, at nonphysiological conditions the weakening of hydrophobic interactions facilitates population of non-native, helical, compact conformations stabilized through direct nonlocal interactions between polar residues.

Collapse of a hydrophobic polymer in a mixture of denaturants.

The solvent quality of an aqueous mixture of two good solvents, urea and guanidinium chloride (GdmCl), for a hydrophobic polymer was investigated using atomistic molecular dynamics simulations. A counterintuitive collapse of the polymer was found, suggesting that mixing the two denaturants reduces the solvent quality. This cononsolvency of the polymer in the urea + GdmCl mixture is found to be caused by the preferential adsorption of urea on the polymer. The polymer collapses as a result of indirect long-range interactions between monomers resulting from the presence of urea clouds surrounding them. Surprisingly, urea behaves as the better solvent in the mixture not because there exists a stronger affinity of the polymer for urea. Instead, attractive interactions between two unlike denaturant molecules combined with the direct dispersion interactions of the polymer with both denaturants determine the solvent quality of the mixture.

An end-to-end deep learning framework for translating mass spectra to de-novo molecules.

Elucidating the structure of a chemical compound is a fundamental task in chemistry with applications in multiple domains including drug discovery, precision medicine, and biomarker discovery. The common practice for elucidating the structure of a compound is to obtain a mass spectrum and subsequently retrieve its structure from spectral databases. However, these methods fail for novel molecules that are not present in the reference database. We propose Spec2Mol, a deep learning architecture for molecular structure recommendation given mass spectra alone. Spec2Mol is inspired by the Speech2Text deep learning architectures for translating audio signals into text. Our approach is based on an encoder-decoder architecture. The encoder learns the spectra embeddings, while the decoder, pre-trained on a massive dataset of chemical structures for translating between different molecular representations, reconstructs SMILES sequences of the recommended chemical structures. We have evaluated Spec2Mol by assessing the molecular similarity between the recommended structures and the original structure. Our analysis showed that Spec2Mol is able to identify the presence of key molecular substructures from its mass spectrum, and shows on par performance, when compared to existing fragmentation tree methods particularly when test structure information is not available during training or present in the reference database.

Accurate clinical toxicity prediction using multi-task deep neural nets and contrastive molecular explanations.

Explainable machine learning for molecular toxicity prediction is a promising approach for efficient drug development and chemical safety. A predictive ML model of toxicity can reduce experimental cost and time while mitigating ethical concerns by significantly reducing animal and clinical testing. Herein, we use a deep learning framework for simultaneously modeling in vitro, in vivo, and clinical toxicity data. Two different molecular input representations are used; Morgan fingerprints and pre-trained SMILES embeddings. A multi-task deep learning model accurately predicts toxicity for all endpoints, including clinical, as indicated by the area under the Receiver Operator Characteristic curve and balanced accuracy. In particular, pre-trained molecular SMILES embeddings as input to the multi-task model improved clinical toxicity predictions compared to existing models in MoleculeNet benchmark. Additionally, our multitask approach is comprehensive in the sense that it is comparable to state-of-the-art approaches for specific endpoints in in vitro, in vivo and clinical platforms. Through both the multi-task model and transfer learning, we were able to indicate the minimal need of in vivo data for clinical toxicity predictions. To provide confidence and explain the model's predictions, we adapt a post-hoc contrastive explanation method that returns pertinent positive and negative features, which correspond well to known mutagenic and reactive toxicophores, such as unsubstituted bonded heteroatoms, aromatic amines, and Michael receptors. Furthermore, toxicophore recovery by pertinent feature analysis captures more of the in vitro (53%) and in vivo (56%), rather than of the clinical (8%), endpoints, and indeed uncovers a preference in known toxicophore data towards in vitro and in vivo experimental data. To our knowledge, this is the first contrastive explanation, using both present and absent substructures, for predictions of clinical and in vivo molecular toxicity.

IL-13 and the hydroperoxy fatty acid 13(S)HpODE play crucial role in inducing an apoptotic pathway in cancer cells involving MAO-A/ROS/p53/p21 signaling axis.

This study depicted the effect of IL-13 and 13(S)HpODE (the endogenous product during IL-13 activation) in the process of cancer cell apoptosis. We examined the role of both IL-13 and 13(S)HpODE in mediating apoptotic pathway in three different in vitro cellular models namely A549 lung cancer, HCT116 colorectal cancer and CCF52 GBM cells. Our data showed that IL-13 promotes apoptosis of A549 lung carcinoma cells through the involvement of 15-LO, PPARγ and MAO-A. Our observations demonstrated that IL-13/13(S)HpODE stimulate MAO-A-mediated intracellular ROS production and p53 as well as p21 induction which play a crucial role in IL-13-stimulated A549 cell apoptosis. We further showed that 13(S)HpODE promotes apoptosis of HCT116 and CCF52 cells through the up-regulation of p53 and p21 expression. Our data delineated that IL-13 stimulates p53 and p21 induction which is mediated through 15-LO and MAO-A in A549 cells. In addition, we observed that PPARγ plays a vital role in apoptosis as well as in p53 and p21 expression in A549 cells in the presence of IL-13. We validated our observations in case of an in vivo colon cancer tumorigenic study using syngeneic mice model and demonstrated that 13(S)HpODE significantly reduces solid tumor growth through the activation of apoptosis. These data thus confirmed that IL-13 > 15-LO>13(S)HpODE > PPARγ>MAO-A > ROS > p53>p21 axis has a major contribution in regulating cancer cell apoptosis and further identified 13(S)HpODE as a potential chemo-preventive agent which can improve the efficacy of cancer treatment as a combination compound.

Accelerating drug target inhibitor discovery with a deep generative foundation model.

Inhibitor discovery for emerging drug-target proteins is challenging, especially when target structure or active molecules are unknown. Here, we experimentally validate the broad utility of a deep generative framework trained at-scale on protein sequences, small molecules, and their mutual interactions-unbiased toward any specific target. We performed a protein sequence-conditioned sampling on the generative foundation model to design small-molecule inhibitors for two dissimilar targets: the spike protein receptor-binding domain (RBD) and the main protease from SARS-CoV-2. Despite using only the target sequence information during the model inference, micromolar-level inhibition was observed in vitro for two candidates out of four synthesized for each target. The most potent spike RBD inhibitor exhibited activity against several variants in live virus neutralization assays. These results establish that a single, broadly deployable generative foundation model for accelerated inhibitor discovery is effective and efficient, even in the absence of target structure or binder information.

Ancestral SARS-CoV-2-Driven Antibody Repertoire Diversity in an Unvaccinated Individual Correlates with Expanded Neutralization Breadth.

Understanding the quality of immune repertoire triggered during natural infection can provide vital clues that form the basis for development of a humoral immune response in some individuals capable of broadly neutralizing pan-SARS-CoV-2 variants. In the present study, we report variations in neutralization potential against Omicron variants of two novel neutralizing monoclonal antibodies (MAbs), THSC20.HVTR11 and THSC20.HVTR55, isolated from an unvaccinated convalescent individual that represent distinct B cell lineage origins and epitope specificity compared to five MAbs we previously reported that were isolated from the same individual. In addition, we observed neutralization of Omicron variants by plasma antibodies obtained from this particular individual postvaccination with increased magnitude. Interestingly, this observation was found to be comparable with six additional individuals who initially were also infected with ancestral SARS-CoV-2 and then received vaccines, indicating that hybrid immunity can provide robust humoral immunity likely by antibody affinity maturation. Development of a distinct antigen-specific B cell repertoire capable of producing polyclonal antibodies with distinct affinity and specificities offers the highest probability of protecting against evolving SARS-CoV-2 variants. IMPORTANCE Development of robust neutralizing antibodies in SARS-CoV-2 convalescent individuals is known; however, it varies at the population level. We isolated monoclonal antibodies from an individual infected with ancestral SARS-CoV-2 in early 2020 that not only varied in their B cell lineage origin but also varied in their capability and potency to neutralize all the known variants of concern (VOCs) and currently circulating Omicron variants. This indicated establishment of unique lineages that contributed in forming a B cell repertoire in this particular individual immediately following infection, giving rise to diverse antibody responses that could complement each other in providing a broadly neutralizing polyclonal antibody response. Individuals who were able to produce polyclonal antibody responses with higher magnitude have a higher chance of being protected from evolving SARS-CoV-2 variants.