Predicting small molecules solubility on endpoint devices using deep ensemble neural networks

Aqueous solubility is a valuable yet challenging property to predict. Computing solubility using first-principles methods requires accounting for the competing effects of entropy and enthalpy, resulting in long computations for relatively poor accuracy. Data-driven approaches, such as deep learning, offer improved accuracy and computational efficiency but typically lack uncertainty quantification. Additionally, ease of use remains a concern for any computational technique, resulting in the sustained popularity of group-based contribution methods. In this work, we addressed these problems with a deep learning model with predictive uncertainty that runs on a static website (without a server). This approach moves computing needs onto the website visitor without requiring installation, removing the need to pay for and maintain servers. Our model achieves satisfactory results in solubility prediction. Furthermore, we demonstrate how to create molecular property prediction models that balance uncertainty and ease of use. The code is available at https://github.com/ur-whitelab/mol.dev, and the model is useable at https://mol.dev.

Purpose reflection benefits minoritized students' motivation and well-being in STEM

Students from groups historically excluded from STEM face heightened challenges to thriving and advancing in STEM. Prompting students to reflect on these challenges in light of their purpose can yield benefits by helping students see how their STEM work connects to fundamental motives. We conducted a randomized, controlled trial to test potential benefits of reflecting on purpose-their "why" for pursuing their degrees. This multimethod study included 466 STEM students (232 women; 237 Black/Latinx/Native students). Participants wrote about their challenges in STEM, with half randomly assigned to consider these in light of their purpose. Purpose reflection fostered benefits to beliefs and attitudes about the major, authentic belonging, and stress appraisals. Effects were robust across race and gender identities or larger for minoritized students. Structural and cultural shifts to recognize students' purpose in STEM can provide a clearer pathway for students to advance.

Inferences of Masculinity and Femininity Across Intersections of Social Class and Gender: A Social Structural Perspective

This research employs a social structural perspective to analyze the content of intersectional social class and gender stereotypes. We investigated how the structural positioning of class and gender categories differentially foster inferences of masculinity and femininity. The social structures that organize class and gender differ: Class is marked by access to resources, and gender is marked by a division of labor for care work. Thus, we examined whether masculinity inferences more strongly varied by social class and whether femininity inferences more strongly varied by gender categories. In Study 1, a total 427 undergraduates provided open-ended descriptions of social class and gender groups. In Study 2, a total 758 undergraduates rated the same groups on preselected trait measures. In Study 3, a total 83 adult participants considered a vignette that manipulated a target's structural resources and gender. Across datasets, variation in social class primarily influenced inferences about masculinity while variation in gender primarily influenced inferences about femininity.

IV BCG Vaccination and Aerosol BCG Revaccination Induce Mycobacteria-Responsive γδ T Cells Associated with Protective Efficacy against M. tb Challenge

Intravenously (IV) delivered BCG provides superior tuberculosis (TB) protection compared with the intradermal (ID) route in non-human primates (NHPs). We examined how γδ T cell responses changed in vivo after IV BCG vaccination of NHPs, and whether these correlated with protection against aerosol M. tuberculosis challenge. In the circulation, Vδ2 T cell populations expanded after IV BCG vaccination, from a median of 1.5% (range: 0.8-2.3) of the CD3+ population at baseline, to 5.3% (range: 1.4-29.5) 4 weeks after M. tb, and were associated with TB protection. This protection was related to effector and central memory profiles; homing markers; and production of IFN-γ, TNF-α and granulysin. In comparison, Vδ2 cells did not expand after ID BCG, but underwent phenotypic and functional changes. When Vδ2 responses in bronchoalveolar lavage (BAL) samples were compared between routes, IV BCG vaccination resulted in highly functional mucosal Vδ2 cells, whereas ID BCG did not. We sought to explore whether an aerosol BCG boost following ID BCG vaccination could induce a γδ profile comparable to that induced with IV BCG. We found evidence that the aerosol BCG boost induced significant changes in the Vδ2 phenotype and function in cells isolated from the BAL. These results indicate that Vδ2 population frequency, activation and function are characteristic features of responses induced with IV BCG, and the translation of responses from the circulation to the site of infection could be a limiting factor in the response induced following ID BCG. An aerosol boost was able to localise activated Vδ2 populations at the mucosal surfaces of the lung. This vaccine strategy warrants further investigation to boost the waning human ID BCG response.

Bloom filters for molecules

Ultra-large chemical libraries are reaching 10s to 100s of billions of molecules. A challenge for these libraries is to efficiently check if a proposed molecule is present. Here we propose and study Bloom filters for testing if a molecule is present in a set using either string or fingerprint representations. Bloom filters are small enough to hold billions of molecules in just a few GB of memory and check membership in sub milliseconds. We found string representations can have a false positive rate below 1% and require significantly less storage than using fingerprints. Canonical SMILES with Bloom filters with the simple FNV (Fowler-Noll-Voll) hashing function provide fast and accurate membership tests with small memory requirements. We provide a general implementation and specific filters for detecting if a molecule is purchasable, patented, or a natural product according to existing databases at https://github.com/whitead/molbloom .

Spore-FP1 tuberculosis mucosal vaccine candidate is highly protective in guinea pigs but fails to improve on BCG-conferred protection in non-human primates

Tuberculosis remains a major health threat globally and a more effective vaccine than the current Bacillus Calmette Guerin (BCG) is required, either to replace or boost it. The Spore-FP1 mucosal vaccine candidate is based on the fusion protein of Ag85B-Acr-HBHA/heparin-binding domain, adsorbed on the surface of inactivated Bacillus subtilis spores. The candidate conferred significant protection against Mycobacterium. tuberculosis challenge in naïve guinea pigs and markedly improved protection in the lungs and spleens of animals primed with BCG. We then immunized rhesus macaques with BCG intradermally, and subsequently boosted with one intradermal and one aerosol dose of Spore-FP1, prior to challenge with low dose aerosolized M. tuberculosis Erdman strain. Following vaccination, animals did not show any adverse reactions and displayed higher antigen specific cellular and antibody immune responses compared to BCG alone but this did not translate into significant improvement in disease pathology or bacterial burden in the organs.

14 examples of how LLMs can transform materials science and chemistry: a reflection on a large language model hackathon

Large-language models (LLMs) such as GPT-4 caught the interest of many scientists. Recent studies suggested that these models could be useful in chemistry and materials science. To explore these possibilities, we organized a hackathon. This article chronicles the projects built as part of this hackathon. Participants employed LLMs for various applications, including predicting properties of molecules and materials, designing novel interfaces for tools, extracting knowledge from unstructured data, and developing new educational applications. The diverse topics and the fact that working prototypes could be generated in less than two days highlight that LLMs will profoundly impact the future of our fields. The rich collection of ideas and projects also indicates that the applications of LLMs are not limited to materials science and chemistry but offer potential benefits to a wide range of scientific disciplines.

Neural potentials of proteins extrapolate beyond training data

We evaluate neural network (NN) coarse-grained (CG) force fields compared to traditional CG molecular mechanics force fields. We conclude that NN force fields are able to extrapolate and sample from unseen regions of the free energy surface when trained with limited data. Our results come from 88 NN force fields trained on different combinations of clustered free energy surfaces from four protein mapped trajectories. We used a statistical measure named total variation similarity to assess the agreement between reference free energy surfaces from mapped atomistic simulations and CG simulations from trained NN force fields. Our conclusions support the hypothesis that NN CG force fields trained with samples from one region of the proteins' free energy surface can, indeed, extrapolate to unseen regions. Additionally, the force matching error was found to only be weakly correlated with a force field's ability to reconstruct the correct free energy surface.

Recent advances in the self-referencing embedded strings (SELFIES) library

String-based molecular representations play a crucial role in cheminformatics applications, and with the growing success of deep learning in chemistry, have been readily adopted into machine learning pipelines. However, traditional string-based representations such as SMILES are often prone to syntactic and semantic errors when produced by generative models. To address these problems, a novel representation, SELF-referencing embedded strings (SELFIES), was proposed that is inherently 100% robust, alongside an accompanying open-source implementation called selfies. Since then, we have generalized SELFIES to support a wider range of molecules and semantic constraints, and streamlined its underlying grammar. We have implemented this updated representation in subsequent versions of selfies, where we have also made major advances with respect to design, efficiency, and supported features. Hence, we present the current status of selfies (version 2.1.1) in this manuscript. Our library, selfies, is available at GitHub (https://github.com/aspuru-guzik-group/selfies).

Learning Peptide Properties with Positive Examples Only

Deep learning can create accurate predictive models by exploiting existing large-scale experimental data, and guide the design of molecules. However, a major barrier is the requirement of both positive and negative examples in the classical supervised learning frameworks. Notably, most peptide databases come with missing information and low number of observations on negative examples, as such sequences are hard to obtain using high-throughput screening methods. To address this challenge, we solely exploit the limited known positive examples in a semi-supervised setting, and discover peptide sequences that are likely to map to certain antimicrobial properties via positive-unlabeled learning (PU). In particular, we use the two learning strategies of adapting base classifier and reliable negative identification to build deep learning models for inferring solubility, hemolysis, binding against SHP-2, and non-fouling activity of peptides, given their sequence. We evaluate the predictive performance of our PU learning method and show that by only using the positive data, it can achieve competitive performance when compared with the classical positive-negative (PN) classification approach, where there is access to both positive and negative examples.

Simultaneous determination of electrical conductivity and thickness of nonmagnetic metallic foils using two-sided multifrequency eddy current techniques

In this paper, we describe a simple method for performing multifrequency eddy current characterization of free-standing uniform-thickness metallic foils using a forked inductive coil arrangement. The method involves measuring the mutual inductance between two coils when a foil is present between the coils, and when it is not present; the ratio of these mutual inductances is compared with an analytical solution, and foil conductivity, thickness, and sheet resistance are simultaneously estimated using numerical inversion and least-squares fitting. This method was used to characterize 34 non-ferrous metallic samples with thicknesses between 50 and 640 μm and with conductivities between 0.8 × 107 and 5.8 × 107 S/m. The estimated thicknesses from eddy current characterization agreed well with those measured using confocal optical techniques; the two approaches agreed to within 1 μm for samples that were thinner than 200 μm, and to within 0.5% for samples that had a thickness of 200 μm or greater. The estimated conductivities from eddy current characterization were in close agreement with expected values, given knowledge of the materials used. A particular strength of this approach is that the instrumentation needed is broadly available in research and development laboratories and the associated fixturing is easy to manufacture and assemble. A calibration procedure is described that can be used to reduce errors from geometric uncertainties. This calibration requires a sample that has only a known conductivity or thickness; both do not need to be known. The method described herein is likely extensible to conductivities and thickness well outside the ranges measured as part of this work.

Assessment of chemistry knowledge in large language models that generate code

In this work, we investigate the question: do code-generating large language models know chemistry? Our results indicate, mostly yes. To evaluate this, we introduce an expandable framework for evaluating chemistry knowledge in these models, through prompting models to solve chemistry problems posed as coding tasks. To do so, we produce a benchmark set of problems, and evaluate these models based on correctness of code by automated testing and evaluation by experts. We find that recent LLMs are able to write correct code across a variety of topics in chemistry and their accuracy can be increased by 30 percentage points via prompt engineering strategies, like putting copyright notices at the top of files. Our dataset and evaluation tools are open source which can be contributed to or built upon by future researchers, and will serve as a community resource for evaluating the performance of new models as they emerge. We also describe some good practices for employing LLMs in chemistry. The general success of these models demonstrates that their impact on chemistry teaching and research is poised to be enormous.

Serverless Prediction of Peptide Properties with Recurrent Neural Networks

We present three deep learning sequence-based prediction models for peptide properties including hemolysis, solubility, and resistance to nonspecific interactions that achieve comparable results to the state-of-the-art models. Our sequence-based solubility predictor, MahLooL, outperforms the current state-of-the-art methods for short peptides. These models are implemented as a static website without the use of a dedicated server or cloud computing. Web-based models like this allow for accessible and effective reproducibility. Most existing approaches rely on third-party servers that typically require upkeep and maintenance. Our predictive models do not require servers, require no installation of dependencies, and work across a range of devices. The specific architecture is bidirectional recurrent neural networks. This serverless approach is a demonstration of edge machine learning that removes the dependence on cloud providers. The code and models are accessible at https://github.com/ur-whitelab/peptide-dashboard.

A Perspective on Explanations of Molecular Prediction Models

Chemists can be skeptical in using deep learning (DL) in decision making, due to the lack of interpretability in "black-box" models. Explainable artificial intelligence (XAI) is a branch of artificial intelligence (AI) which addresses this drawback by providing tools to interpret DL models and their predictions. We review the principles of XAI in the domain of chemistry and emerging methods for creating and evaluating explanations. Then, we focus on methods developed by our group and their applications in predicting solubility, blood-brain barrier permeability, and the scent of molecules. We show that XAI methods like chemical counterfactuals and descriptor explanations can explain DL predictions while giving insight into structure-property relationships. Finally, we discuss how a two-step process of developing a black-box model and explaining predictions can uncover structure-property relationships.

Science for Others or the Self? Presumed Motives for Science Shape Public Trust in Science

Science can improve life around the world, but public trust in science is at risk. Understanding the presumed motives of scientists and science can inform the social psychological underpinnings of public trust in science. Across five independent datasets, perceiving the motives of science and scientists as prosocial promoted public trust in science. In Studies 1 and 2, perceptions that science was more prosocially oriented were associated with greater trust in science. Studies 3 and 4a & 4b employed experimental methods to establish that perceiving other-oriented motives, versus self-oriented motives, enhanced public trust in science. Respondents recommend greater funding allocations for science subdomains described as prosocially oriented versus power-oriented. Emphasizing the prosocial aspects of science can build stronger foundations of public trust in science.

NPP-669, a Novel Broad-Spectrum Antiviral Therapeutic with Excellent Cellular Uptake, Antiviral Potency, Oral Bioavailability, Preclinical Efficacy, and a Promising Safety Margin

DNA viruses are responsible for many diseases in humans. Current treatments are often limited by toxicity, as in the case of cidofovir (CDV, Vistide), a compound used against cytomegalovirus (CMV) and adenovirus (AdV) infections. CDV is a polar molecule with poor bioavailability, and its overall clinical utility is limited by the high occurrence of acute nephrotoxicity. To circumvent these disadvantages, we designed nine CDV prodrug analogues. The prodrugs modulate the polarity of CDV with a long sulfonyl alkyl chain attached to one of the phosphono oxygens. We added capping groups to the end of the alkyl chain to minimize β-oxidation and focus the metabolism on the phosphoester hydrolysis, thereby tuning the rate of this reaction by altering the alkyl chain length. With these modifications, the prodrugs have excellent aqueous solubility, optimized metabolic stability, increased cellular permeability, and rapid intracellular conversion to the pharmacologically active diphosphate form (CDV-PP). The prodrugs exhibited significantly enhanced antiviral potency against a wide range of DNA viruses in infected human foreskin fibroblasts. Single-dose intravenous and oral pharmacokinetic experiments showed that the compounds maintained plasma and target tissue levels of CDV well above the EC₅₀ for 24 h. These experiments identified a novel lead candidate, NPP-669. NPP-669 demonstrated efficacy against CMV infections in mice and AdV infections in hamsters following oral (p.o.) dosing at a dose of 1 mg/kg BID and 0.1 mg/kg QD, respectively. We further showed that NPP-669 at 30 mg/kg QD did not exhibit histological signs of toxicity in mice or hamsters. These data suggest that NPP-669 is a promising lead candidate for a broad-spectrum antiviral compound.

Deep Learning for Molecules and Materials

Deep learning is becoming a standard tool in chemistry and materials science. Although there are learning materials available for deep learning, none cover the applications in chemistry and materials science or the peculiarities of working with molecules. The textbook described here provides a systematic and applied introduction to the latest research in deep learning in chemistry and materials science. It covers the math fundamentals, the requisite machine learning, the common neural network architectures used today, and the details necessary to be a practitioner of deep learning. The textbook is a living document and will be updated as the rapidly changing deep learning field evolves.

Identification of HLA-E Binding Mycobacterium tuberculosis-Derived Epitopes through Improved Prediction Models

Tuberculosis (TB) remains one of the deadliest infectious diseases worldwide, posing great social and economic burden to affected countries. Novel vaccine approaches are needed to increase protective immunity against the causative agent Mycobacterium tuberculosis (Mtb) and to reduce the development of active TB disease in latently infected individuals. Donor-unrestricted T cell responses represent such novel potential vaccine targets. HLA-E-restricted T cell responses have been shown to play an important role in protection against TB and other infections, and recent studies have demonstrated that these cells can be primed in vitro. However, the identification of novel pathogen-derived HLA-E binding peptides presented by infected target cells has been limited by the lack of accurate prediction algorithms for HLA-E binding. In this study, we developed an improved HLA-E binding peptide prediction algorithm and implemented it to identify (to our knowledge) novel Mtb-derived peptides with capacity to induce CD8+ T cell activation and that were recognized by specific HLA-E-restricted T cells in Mycobacterium-exposed humans. Altogether, we present a novel algorithm for the identification of pathogen- or self-derived HLA-E-presented peptides.

SELFIES and the future of molecular string representations

Artificial intelligence (AI) and machine learning (ML) are expanding in popularity for broad applications to challenging tasks in chemistry and materials science. Examples include the prediction of properties, the discovery of new reaction pathways, or the design of new molecules. The machine needs to read and write fluently in a chemical language for each of these tasks. Strings are a common tool to represent molecular graphs, and the most popular molecular string representation, Smiles, has powered cheminformatics since the late 1980s. However, in the context of AI and ML in chemistry, Smiles has several shortcomings-most pertinently, most combinations of symbols lead to invalid results with no valid chemical interpretation. To overcome this issue, a new language for molecules was introduced in 2020 that guarantees 100% robustness: SELF-referencing embedded string (Selfies). Selfies has since simplified and enabled numerous new applications in chemistry. In this perspective, we look to the future and discuss molecular string representations, along with their respective opportunities and challenges. We propose 16 concrete future projects for robust molecular representations. These involve the extension toward new chemical domains, exciting questions at the interface of AI and robust languages, and interpretability for both humans and machines. We hope that these proposals will inspire several follow-up works exploiting the full potential of molecular string representations for the future of AI in chemistry and materials science.

Inferring spatial source of disease outbreaks using maximum entropy

Mathematical modeling of disease outbreaks can infer the future trajectory of an epidemic, allowing for making more informed policy decisions. Another task is inferring the origin of a disease, which is relatively difficult with current mathematical models. Such frameworks, across varying levels of complexity, are typically sensitive to input data on epidemic parameters, case counts, and mortality rates, which are generally noisy and incomplete. To alleviate these limitations, we propose a maximum entropy framework that fits epidemiological models, provides calibrated infection origin probabilities, and is robust to noise due to a prior belief model. Maximum entropy is agnostic to the parameters or model structure used and allows for flexible use when faced with sparse data conditions and incomplete knowledge in the dynamical phase of disease-spread, providing for more reliable modeling at early stages of outbreaks. We evaluate the performance of our model by predicting future disease trajectories based on simulated epidemiological data in synthetic graph networks and the real mobility network of New York State. In addition, unlike existing approaches, we demonstrate that the method can be used to infer the origin of the outbreak with accurate confidence. Indeed, despite the prevalent belief on the feasibility of contact-tracing being limited to the initial stages of an outbreak, we report the possibility of reconstructing early disease dynamics, including the epidemic seed, at advanced stages.

Abstract 4003: Identification of a targeted anti-mitotic agent that degrades Myc and specifically induces cancer cell death

For decades, chemotherapy has remained the standard of care for ovarian cancer patients. Although 70% of patients initially respond to platinum-based therapy, >90% succumb to chemoresistance. There is a need to uncover targeted drugs for ovarian cancer. A quicker, more cost-effective, and lower-risk route to identify new treatments is through repurposing FDA approved drugs. Using a computational drug repositioning platform, DrugPredict, we previously uncovered that the antiarrhythmic drug amiodarone potently decreases cell viability and triggers apoptosis in numerous patient-derived ovarian cancer cell lines through the degradation of c-Myc. However, given the dose-limiting toxicity of amiodarone, we applied structure-activity relationship to identify DL78, which lacks hERG activity but retains the anti-cancer properties and ability to regulate Myc. DL78 is significantly more potent and tumor specific than amiodarone. It rapidly induces G2/M arrest which ultimately leads to loss of Myc, mitotic catastrophe, and apoptosis in several types of cancer cells. Furthermore, pharmacokinetics studies show that the compound is retained in the tumor upon intraperitoneal injection and has reasonable solubility and permeability, yielding good absorption, distinct from amiodarone. Given its structural and molecular differences, we expect that DL78 works through a different mechanism that amiodarone and could represent an effective treatment for ovarian cancer and other MYC-driven tumors. Additional studies will establish the foundation for further development of this novel compound, as well as reveal targeted ovarian cancer vulnerabilities.

Citation Format: Jessica McAnulty, Michele Dziubinski, Agharnan Gandhi, Margaret Farah, Pil Lee, Andrew D. White, Analisa DiFeo. Identification of a targeted anti-mitotic agent that degrades Myc and specifically induces cancer cell death [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 4003.

Symmetric Molecular Dynamics

We derive a formulation of molecular dynamics that generates only symmetric configurations. We implement it for all 2D planar and 3D space groups. An atlas of 2D Lennard-Jones crystals under all planar groups is created with symmetric molecular dynamics.

Federated learning of molecular properties with graph neural networks in a heterogeneous setting

Chemistry research has both high material and computational costs to conduct experiments. Intuitions are interested in differing classes of molecules, creating heterogeneous data that cannot be easily joined by conventional methods. This work introduces federated heterogeneous molecular learning. Federated learning allows end users to build a global model collaboratively while keeping their training data isolated. We first simulate a heterogeneous federated-learning benchmark (FedChem) by jointly performing scaffold splitting and latent Dirichlet allocation on existing datasets. Our results on FedChem show that significant learning challenges arise when working with heterogeneous molecules across clients. We then propose a method to alleviate the problem: Federated Learning by Instance reweighTing (FLIT(+)). FLIT(+) can align local training across clients. Experiments conducted on FedChem validate the advantages of this method. This work should enable a new type of collaboration for improving artificial intelligence (AI) in chemistry that mitigates concerns about sharing valuable chemical data.

Simulation-based inference with approximately correct parameters via maximum entropy

Inferring the input parameters of simulators from observations is a crucial challenge with applications from epidemiology to molecular dynamics. Here we show a simple approach in the regime of sparse data and approximately correct models, which is common when trying to use an existing model to infer latent variables with observed data. This approach is based on the principle of maximum entropy (MaxEnt) and provably makes the smallest change in the latent joint distribution to fit new data. This method requires no likelihood or model derivatives and its fit is insensitive to prior strength, removing the need to balance observed data fit with prior belief. The method requires the ansatz that data is fit in expectation, which is true in some settings and may be reasonable in all settings with few data points. The method is based on sample reweighting, so its asymptotic run time is independent of prior distribution dimension. We demonstrate this MaxEnt approach and compare with other likelihood-free inference methods across three systems: a point particle moving in a gravitational field, a compartmental model of epidemic spread and molecular dynamics simulation of a protein.

Influence of Aerosol Delivered BCG Vaccination on Immunological and Disease Parameters Following SARS-CoV-2 Challenge in Rhesus Macaques

The tuberculosis vaccine, Bacille Calmette-Guerin (BCG), also affords protection against non-tuberculous diseases attributable to heterologous immune mechanisms such as trained innate immunity, activation of non-conventional T-cells, and cross-reactive adaptive immunity. Aerosol vaccine delivery can target immune responses toward the primary site of infection for a respiratory pathogen. Therefore, we hypothesised that aerosol delivery of BCG would enhance cross-protective action against severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infection and be a deployable intervention against coronavirus disease 2019 (COVID-19). Immune parameters were monitored in vaccinated and unvaccinated rhesus macaques for 28 days following aerosol BCG vaccination. High-dose SARS-CoV-2 challenge was applied by intranasal and intrabronchial instillation and animals culled 6–8 days later for assessment of viral, disease, and immunological parameters. Mycobacteria-specific cell-mediated immune responses were detected following aerosol BCG vaccination, but SARS-CoV-2-specific cellular- and antibody-mediated immunity was only measured following challenge. Early secretion of cytokine and chemokine markers associated with the innate cellular and adaptive antiviral immune response was detected following SARS-CoV-2 challenge in vaccinated animals, at concentrations that exceeded titres measured in unvaccinated macaques. Classical CD14+ monocytes and Vδ2 γδ T-cells quantified by whole-blood immunophenotyping increased rapidly in vaccinated animals following SARS-CoV-2 challenge, indicating a priming of innate immune cells and non-conventional T-cell populations. However, viral RNA quantified in nasal and pharyngeal swabs, bronchoalveolar lavage (BAL), and tissue samples collected at necropsy was equivalent in vaccinated and unvaccinated animals, and in-life CT imaging and histopathology scoring applied to pulmonary tissue sections indicated that the disease induced by SARS-CoV-2 challenge was comparable between vaccinated and unvaccinated groups. Hence, aerosol BCG vaccination did not induce, or enhance the induction of, SARS-CoV-2 cross-reactive adaptive cellular or humoral immunity, although an influence of BCG vaccination on the subsequent immune response to SARS-CoV-2 challenge was apparent in immune signatures indicative of trained innate immune mechanisms and primed unconventional T-cell populations. Nevertheless, aerosol BCG vaccination did not enhance the initial clearance of virus, nor reduce the occurrence of early disease pathology after high dose SARS-CoV-2 challenge. However, the heterologous immune mechanisms primed by BCG vaccination could contribute to the moderation of COVID-19 disease severity in more susceptible species following natural infection.

Peptide framework for screening the effects of amino acids on assembly

Discovery of peptide domains with unique intermolecular interactions is essential for engineering peptide-based materials. Rather than attempting a brute-force approach, we instead identify a previously unexplored strategy for discovery and study of intermolecular interactions: “co-assembly of oppositely charged peptide” (CoOP), a framework that “encourages” peptide assembly by mixing two oppositely charged hexapeptides. We used an integrated computational and experimental approach, probed the free energy of association and probability of amino acid contacts during co-assembly with atomic-resolution simulations, and correlated them to the physical properties of the aggregates. We introduce CoOP with three examples: dialanine, ditryptophan, and diisoleucine. Our results indicated that the opposite charges initiate the assembly, and the subsequent stability is enhanced by the presence of an undisturbed hydrophobic core. CoOP represents a unique, simple, and elegant framework that can be used to identify the structure-property relationships of self-assembling peptide-based materials.

Induction of Functional Specific Antibodies, IgG-Secreting Plasmablasts and Memory B Cells Following BCG Vaccination

Tuberculosis (TB) is a major global health problem and the only currently-licensed vaccine, BCG, is inadequate. Many TB vaccine candidates are designed to be given as a boost to BCG; an understanding of the BCG-induced immune response is therefore critical, and the opportunity to relate this to circumstances where BCG does confer protection may direct the design of more efficacious vaccines. While the T cell response to BCG vaccination has been well-characterized, there is a paucity of literature on the humoral response. We demonstrate BCG vaccine-mediated induction of specific antibodies in different human populations and macaque species which represent important preclinical models for TB vaccine development. We observe a strong correlation between antibody titers in serum versus plasma with modestly higher titers in serum. We also report for the first time the rapid and transient induction of antibody-secreting plasmablasts following BCG vaccination, together with a robust and durable memory B cell response in humans. Finally, we demonstrate a functional role for BCG vaccine-induced specific antibodies in opsonizing mycobacteria and enhancing macrophage phagocytosis in vitro, which may contribute to the BCG vaccine-mediated control of mycobacterial growth observed. Taken together, our findings indicate that the humoral immune response in the context of BCG vaccination merits further attention to determine whether TB vaccine candidates could benefit from the induction of humoral as well as cellular immunity.

Natural language processing models that automate programming will transform chemistry research and teaching

Natural language processing models have emerged that can generate useable software and automate a number of programming tasks with high fidelity. These tools have yet to have an impact on the chemistry community. Yet, our initial testing demonstrates that this form of artificial intelligence is poised to transform chemistry and chemical engineering research. Here, we review developments that brought us to this point, examine applications in chemistry, and give our perspective on how this may fundamentally alter research and teaching.

Natural language processing models have emerged that can generate useable software and automate a number of programming tasks with high fidelity.

Model agnostic generation of counterfactual explanations for molecules

An outstanding challenge in deep learning in chemistry is its lack of interpretability. The inability of explaining why a neural network makes a prediction is a major barrier to deployment of AI models. This not only dissuades chemists from using deep learning predictions, but also has led to neural networks learning spurious correlations that are difficult to notice. Counterfactuals are a category of explanations that provide a rationale behind a model prediction with satisfying properties like providing chemical structure insights. Yet, counterfactuals have been previously limited to specific model architectures or required reinforcement learning as a separate process. In this work, we show a universal model-agnostic approach that can explain any black-box model prediction. We demonstrate this method on random forest models, sequence models, and graph neural networks in both classification and regression.

Generating model agnostic molecular counterfactual explanations to explain model predictions.

Characterization of the Infant Immune System and the Influence and Immunogenicity of BCG Vaccination in Infant and Adult Rhesus Macaques

In many countries where tuberculosis (TB) is endemic, the Bacillus Calmette–Guérin (BCG) vaccine is given as close to birth as possible to protect infants and children from severe forms of TB. However, BCG has variable efficacy and is not as effective against adult pulmonary TB. At present, most animal models used to study novel TB vaccine candidates rely on the use of adult animals. Human studies show that the infant immune system is different to that of an adult. Understanding how the phenotypic profile and functional ability of the immature host immune system compares to that of a mature adult, together with the subsequent BCG immune response, is critical to ensuring that new TB vaccines are tested in the most appropriate models. BCG-specific immune responses were detected in macaques vaccinated within a week of birth from six weeks after immunization indicating that neonatal macaques are able to generate a functional cellular response to the vaccine. However, the responses measured were significantly lower than those typically observed following BCG vaccination in adult rhesus macaques and infant profiles were skewed towards the activation and attraction of macrophages and monocytes and the synthesis in addition to release of pro-inflammatory cytokines such as IL-1, IL-6 and TNF-α. The frequency of specific immune cell populations changed significantly through the first three years of life as the infants developed into young adult macaques. Notably, the CD4:CD8 ratio significantly declined as the macaques aged due to a significant decrease in the proportion of CD4⁺ T-cells relative to a significant increase in CD8⁺ T-cells. Also, the frequency of both CD4⁺ and CD8⁺ T-cells expressing the memory marker CD95, and memory subset populations including effector memory, central memory and stem cell memory, increased significantly as animals matured. Infant macaques, vaccinated with BCG within a week of birth, possessed a significantly higher frequency of CD14⁺ classical monocytes and granulocytes which remained different throughout the first three years of life compared to unvaccinated age matched animals. These findings, along with the increase in monokines following vaccination in infants, may provide an insight into the mechanism by which vaccination with BCG is able to provide non-specific immunity against non-mycobacterial organisms.

Correction: Graph neural network based coarse-grained mapping prediction

Correction for ‘Graph neural network based coarse-grained mapping prediction’ by Zhiheng Li et al., Chem. Sci., 2020, 11, 9524–9531, DOI: 10.1039/D0SC02458A.

Cheap and Commonplace: Making the Case for BCG and γδ T Cells in COVID-19

Antigen-specific vaccines developed for the COVID-19 pandemic demonstrate a remarkable achievement and are currently being used in high income countries with much success. However, new SARS-CoV-2 variants are threatening this success via mutations that lessen the efficacy of antigen-specific antibodies. One simple approach to assisting with this issue is focusing on strategies that build on the non-specific protection afforded by the innate immune response. The BCG vaccine has been shown to provide broad protection beyond tuberculosis disease, including against respiratory viruses, and ongoing studies are investigating its efficacy as a tool against SARS-CoV-2. Gamma delta (γδ) T cells, particularly the Vδ2 subtype, undergo rapid expansion after BCG vaccination due to MHC-independent mechanisms. Consequently, γδ T cells can produce diverse defenses against virally infected cells, including direct cytotoxicity, death receptor ligands, and pro-inflammatory cytokines. They can also assist in stimulating the adaptive immune system. BCG is affordable, commonplace and non-specific, and therefore could be a useful tool to initiate innate protection against new SARS-CoV-2 variants. However, considerations must also be made to BCG vaccine supply and the prioritization of countries where it is most needed to combat tuberculosis first and foremost.

TB and SIV Coinfection; a Model for Evaluating Vaccine Strategies against TB Reactivation in Asian Origin Cynomolgus Macaques: A Pilot Study Using BCG Vaccination

This pilot study aimed to determine the utility of a cynomolgus macaque model of coinfection with simian immunodeficiency virus (SIV) for the assessment of vaccines designed to prevent reactivation of TB. Following infection caused by aerosol exposure to an ultralow dose of Mycobacterium tuberculosis (M. tb), data trends indicated that subsequent coinfection with SIVmac32H perturbed control of M. tb infection as evidenced by the increased occurrence of progressive disease in this group, higher levels of pathology and increased frequency of progressive tuberculous granulomas in the lung. BCG vaccination led to improved control of TB-induced disease and lower viral load in comparison to unvaccinated coinfected animals. The M. tb-specific IFNγ response after exposure to M. tb, previously shown to be associated with bacterial burden, was lower in the BCG-vaccinated group than in the unvaccinated groups. Levels of CD4+ and CD8+ T cells decreased in coinfected animals, with counts recovering more quickly in the BCG-vaccinated group. This pilot study provides proof of concept to support the use of the model for evaluation of interventions against reactivated/exacerbated TB caused by human immunodeficiency virus (HIV) infection.

Predicting chemical shifts with graph neural networks

Inferring molecular structure from Nuclear Magnetic Resonance (NMR) measurements requires an accurate forward model that can predict chemical shifts from 3D structure. Current forward models are limited to specific molecules like proteins and state-of-the-art models are not differentiable. Thus they cannot be used with gradient methods like biased molecular dynamics. Here we use graph neural networks (GNNs) for NMR chemical shift prediction. Our GNN can model chemical shifts accurately and capture important phenomena like hydrogen bonding induced downfield shift between multiple proteins, secondary structure effects, and predict shifts of organic molecules. Previous empirical NMR models of protein NMR have relied on careful feature engineering with domain expertise. These GNNs are trained from data alone with no feature engineering yet are as accurate and can work on arbitrary molecular structures. The models are also efficient, able to compute one million chemical shifts in about 5 seconds. This work enables a new category of NMR models that have multiple interacting types of macromolecules.

ChAdOx1 nCoV-19 protection against SARS-CoV-2 in rhesus macaque and ferret challenge models

Vaccines against SARS-CoV-2 are urgently required, but early development of vaccines against SARS-CoV-1 resulted in enhanced disease after vaccination. Careful assessment of this phenomena is warranted for vaccine development against SARS CoV-2. Here we report detailed immune profiling after ChAdOx1 nCoV-19 (AZD1222) and subsequent high dose challenge in two animal models of SARS-CoV-2 mediated disease. We demonstrate in rhesus macaques the lung pathology caused by SARS-CoV-2 mediated pneumonia is reduced by prior vaccination with ChAdOx1 nCoV-19 which induced neutralising antibody responses after a single intramuscular administration. In a second animal model, ferrets, ChAdOx1 nCoV-19 reduced both virus shedding and lung pathology. Antibody titre were boosted by a second dose. Data from these challenge models on the absence of enhanced disease and the detailed immune profiling, support the continued clinical evaluation of ChAdOx1 nCoV-19.

Differences in host immune populations between rhesus macaques and cynomolgus macaque subspecies in relation to susceptibility to Mycobacterium tuberculosis infection

Rhesus (Macaca mulatta) and cynomolgus (Macaca fasicularis) macaques of distinct genetic origin are understood to vary in susceptibility to Mycobacterium tuberculosis, and therefore differences in their immune systems may account for the differences in disease control. Monocyte:lymphocyte (M:L) ratio has been identified as a risk factor for M. tuberculosis infection and is known to vary between macaque species. We aimed to characterise the constituent monocyte and lymphocyte populations between macaque species, and profile other major immune cell subsets including: CD4⁺ and CD8⁺ T-cells, NK-cells, B-cells, monocyte subsets and myeloid dendritic cells. We found immune cell subsets to vary significantly between macaque species. Frequencies of CD4⁺ and CD8⁺ T-cells and the CD4:CD8 ratio showed significant separation between species, while myeloid dendritic cells best associated macaque populations by M. tuberculosis susceptibility. A more comprehensive understanding of the immune parameters between macaque species may contribute to the identification of new biomarkers and correlates of protection.

Discovery of Acyl-sulfonamide Nav1.7 Inhibitors GDC-0276 and GDC-0310

Nav1.7 is an extensively investigated target for pain with a strong genetic link in humans, yet in spite of this effort, it remains challenging to identify efficacious, selective, and safe inhibitors. Here, we disclose the discovery and preclinical profile of GDC-0276 (1) and GDC-0310 (2), selective Nav1.7 inhibitors that have completed Phase 1 trials. Our initial search focused on close-in analogues to early compound 3. This resulted in the discovery of GDC-0276 (1), which possessed improved metabolic stability and an acceptable overall pharmacokinetics profile. To further derisk the predicted human pharmacokinetics and enable QD dosing, additional optimization of the scaffold was conducted, resulting in the discovery of a novel series of N-benzyl piperidine Nav1.7 inhibitors. Improvement of the metabolic stability by blocking the labile benzylic position led to the discovery of GDC-0310 (2), which possesses improved Nav selectivity and pharmacokinetic profile over 1.

Comparison of rhesus and cynomolgus macaques as an infection model for COVID-19

A novel coronavirus, SARS-CoV-2, has been identified as the causative agent of the current COVID-19 pandemic. Animal models, and in particular non-human primates, are essential to understand the pathogenesis of emerging diseases and to assess the safety and efficacy of novel vaccines and therapeutics. Here, we show that SARS-CoV-2 replicates in the upper and lower respiratory tract and causes pulmonary lesions in both rhesus and cynomolgus macaques. Immune responses against SARS-CoV-2 are also similar in both species and equivalent to those reported in milder infections and convalescent human patients. This finding is reiterated by our transcriptional analysis of respiratory samples revealing the global response to infection. We describe a new method for lung histopathology scoring that will provide a metric to enable clearer decision making for this key endpoint. In contrast to prior publications, in which rhesus are accepted to be the preferred study species, we provide convincing evidence that both macaque species authentically represent mild to moderate forms of COVID-19 observed in the majority of the human population and both species should be used to evaluate the safety and efficacy of interventions against SARS-CoV-2. Importantly, accessing cynomolgus macaques will greatly alleviate the pressures on current rhesus stocks.

Investigating Active Learning and Meta-Learning for Iterative Peptide Design

Often the development of novel functional peptides is not amenable to high throughput or purely computational screening methods. Peptides must be synthesized one at a time in a process that does not generate large amounts of data. One way this method can be improved is by ensuring that each experiment provides the best improvement in both peptide properties and predictive modeling accuracy. Here, we study the effectiveness of active learning, optimizing experiment order, and meta-learning, transferring knowledge between contexts, to reduce the number of experiments necessary to build a predictive model. We present a multitask benchmark database of peptides designed to advance these methods for experimental design. Each task is a binary classification of peptides represented as a sequence string. We find neither active learning method tested to be better than random choice. The meta-learning method Reptile was found to improve the average accuracy across data sets. Combining meta-learning with active learning offers inconsistent benefits.

Generation of Highly Selective, Potent, and Covalent G Protein-Coupled Receptor Kinase 5 Inhibitors

The ability of G protein-coupled receptor (GPCR) kinases (GRKs) to regulate the desensitization of GPCRs has made GRK2 and GRK5 attractive targets for treating diseases such as heart failure and cancer. Previously, our work showed that Cys474, a GRK5 subfamily-specific residue located on a flexible loop adjacent to the active site, can be used as a covalent handle to achieve selective inhibition of GRK5 over GRK2 subfamily members. However, the potency of the most selective inhibitors remained modest. Herein, we describe a successful campaign to adapt an indolinone scaffold with covalent warheads, resulting in a series of 2-haloacetyl-containing compounds that react quickly and exhibit three orders of magnitude selectivity for GRK5 over GRK2 and low nanomolar potency. They however retain a similar selectivity profile across the kinome as the core scaffold, which was based on Sunitinib.

Experimentally Consistent Simulation of Aβ21-30 Peptides with a Minimal NMR Bias

Misfolded amyloid peptides are neurotoxic molecules associated with Alzheimer's disease. The Aβ_21-30 peptide fragment is a decapeptide fragment of the complete Aβ42 peptide which is a hypothesized cause of Alzheimer's disease via amyloid fibrillogenesis. Aβ_21-30 is investigated here with a combination of NMR (nuclear magnetic resonance) spectroscopy experiments and molecular dynamics simulations with experiment directed simulation (EDS). EDS is a maximum entropy biasing method that augments a molecular dynamics simulation with experimental data (NMR chemical shifts) to improve agreement with experiments and thus accuracy. EDS molecular dynamics shows that the Aβ_21-30 monomer has a β turn stabilized by the following interactions: S26-K28, D23-S26, and D23-K28. NMR, total correlation spectroscopy, and rotating frame Overhauser effect spectroscopy experiments provide independent agreement. Subsequent two- and four-monomer EDS simulations show aggregation. Diffusion coefficients calculated from molecular simulation also agreed with experimentally measured values only after using EDS, providing independent assessment of accuracy. This work demonstrates how accuracy can be improved by directly using experimental data in molecular dynamics of complex processes like self-assembly.

Graph neural network based coarse-grained mapping prediction

The selection of coarse-grained (CG) mapping operators is a critical step for CG molecular dynamics (MD) simulation. It is still an open question about what is optimal for this choice and there is a need for theory. The current state-of-the art method is mapping operators manually selected by experts. In this work, we demonstrate an automated approach by viewing this problem as supervised learning where we seek to reproduce the mapping operators produced by experts. We present a graph neural network based CG mapping predictor called Deep Supervised Graph Partitioning Model (DSGPM) that treats mapping operators as a graph segmentation problem. DSGPM is trained on a novel dataset, Human-annotated Mappings (HAM), consisting of 1180 molecules with expert annotated mapping operators. HAM can be used to facilitate further research in this area. Our model uses a novel metric learning objective to produce high-quality atomic features that are used in spectral clustering. The results show that the DSGPM outperforms state-of-the-art methods in the field of graph segmentation. Finally, we find that predicted CG mapping operators indeed result in good CG MD models when used in simulation.

We propose a scalable graph neural network-based method for automating coarse-grained mapping prediction for molecules.

Protective Efficacy of Inhaled BCG Vaccination Against Ultra-Low Dose Aerosol M. tuberculosis Challenge in Rhesus Macaques

Ten million cases of tuberculosis (TB) were reported in 2018 with a further 1.5 million deaths attributed to the disease. Improved vaccination strategies are urgently required to tackle the ongoing global TB epidemic. In the absence of a validated correlate of protection, highly characterised pre-clinical models are required to assess the protective efficacy of new vaccination strategies. In this study, we demonstrate the application of a rhesus macaque ultra-low dose (ULD) aerosol M. tuberculosis challenge model for the evaluation of TB vaccination strategies by directly comparing the immunogenicity and efficacy of intradermal (ID) and aerosol BCG vaccination delivered using a portable vibrating mesh nebulizer (VMN). Aerosol- and ID-delivered Bacille Calmette-Guérin (BCG) induced comparable frequencies of IFN-γ spot forming units (SFU) measured in peripheral blood mononuclear cells (PBMCs) by ELISpot, although the induction of IFN-γ SFU was significantly delayed following aerosol immunisation. This delayed response was also apparent in an array of secreted pro-inflammatory and chemokine markers, as well as in the frequency of antigen-specific cytokine producing CD4 and CD8 T-cells measured by multi-parameter flow cytometry. Interrogation of antigen-specific memory T-cell phenotypes revealed that vaccination-induced CD4 and CD8 T-cell populations primarily occupied the central memory (TCM) and transitional effector memory (TransEM) phenotype, and that the frequency of CD8 TCM and TransEM populations was significantly higher in aerosol BCG-vaccinated animals in the week prior to M. tuberculosis infection. The total and lung pathology measured following M. tuberculosis challenge was significantly lower in vaccinated animals relative to the unvaccinated control group and pathology measured in extra-pulmonary tissues was significantly reduced in aerosol BCG-vaccinated animals, relative to the ID-immunised group. Similarly, significantly fewer viable M. tuberculosis CFU were recovered from the extra-pulmonary tissues of aerosol BCG-vaccinated macaques relative to unvaccinated animals. In this study, a rhesus macaque ULD M. tuberculosis aerosol challenge model was applied as a refined and sensitive system for the evaluation of TB vaccine efficacy and to confirm that aerosol BCG vaccination delivered by portable VMN can confer a significant level of protection that is equivalent, and by some measures superior, to intradermal BCG vaccination.

Synthesis and structure-activity relationships of thieno[2,3-d]pyrimidines as atypical protein kinase C inhibitors to control retinal vascular permeability and cytokine-induced edema

Studies demonstrate that small molecule targeting of atypical protein kinase C (aPKC) may provide an effective means to control vascular permeability, prevent edema, and reduce inflammation providing novel and important alternatives to anti-VEGF therapies for certain blinding eye diseases. Based on a literature tricyclic thieno[2,3-d]pyrimidine lead (1), an ATP-competitive inhibitor of the aPKC iota (ι) and aPKC zeta (ζ) isoforms, we have synthesized a small series of compounds in 1-2 steps from a readily available chloro intermediate. A single pyridine congener was also made using 2D NMR to assign regiochemistry. Within the parent pyrimidine series, a range of potencies was observed against aPKCζ whereas the pyridine congener was inactive. Selected compounds were also tested for their effect toward VEGF-induced permeability in BREC cells. The most potent of these (7l) was further assayed against the aPKCι isoform and showed a favorable selectivity profile against a panel of 31 kinases, including kinases from the AGC superfamily, with a focus on PKC isoforms and kinases previously shown to affect permeability. Further testing of 7l in a luciferase assay in HEK293 cells showed an ability to prevent TNF-α induced NFκB activation while not having any effect on cell survival. Intravitreal administration of 7l to the eye yielded a complete reduction in permeability in a test to determine whether the compound could block VEGF- and TNFα-induced permeability across the retinal vasculature in a rat model. The compound in mice displayed good microsomal stability and in plasma moderate exposure (AUC and C_max), low clearance, a long half-life and high oral bioavailability. With IV dosing, higher levels were observed in the brain and eye relative to plasma, with highest levels in the eye by either IV or PO dosing. With a slow oral absorption profile, 7l accumulates in the eye to maintain a high concentration after dosing with higher levels than in plasma. Compound 7l may represent a class of aPKC inhibitors for further investigation.

Is preservation of symmetry necessary for coarse-graining?

This work investigates if preserving the symmetry of the underlying molecular graph of a given molecule when choosing a coarse-grained (CG) mapping significantly affects the CG model accuracy.

Structure-Based Design of Selective, Covalent G Protein-Coupled Receptor Kinase 5 Inhibitors

The ability of G protein-coupled receptor (GPCR) kinases (GRKs) to regulate desensitization of GPCRs has made GRK2 and GRK5 attractive targets for treating heart failure and other diseases such as cancer. Although advances have been made toward developing inhibitors that are selective for GRK2, there have been far fewer reports of GRK5 selective compounds. Herein, we describe the development of GRK5 subfamily selective inhibitors, 5 and 16d that covalently interact with a nonconserved cysteine (Cys474) unique to this subfamily. Compounds 5 and 16d feature a highly amenable pyrrolopyrimidine scaffold that affords high nanomolar to low micromolar activity that can be easily modified with Michael acceptors with various reactivities and geometries. Our work thereby establishes a new pathway toward further development of subfamily selective GRK inhibitors and establishes Cys474 as a new and useful covalent handle in GRK5 drug discovery.