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.

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

[This corrects the article DOI: 10.1039/D3DD00217A.].

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.

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).

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.

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.

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 .

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.

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.

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.

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.

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.

Surveillance of Daughter Micronodule Formation Is a Key Factor for Vaccine Evaluation Using Experimental Infection Models of Tuberculosis in Macaques.

Tuberculosis (TB) is still a major worldwide health problem and models using non-human primates (NHP) provide the most relevant approach for vaccine testing. In this study, we analysed CT images collected from cynomolgus and rhesus macaques following exposure to ultra-low dose Mycobacterium tuberculosis (Mtb) aerosols, and monitored them for 16 weeks to evaluate the impact of prior intradermal or inhaled BCG vaccination on the progression of lung disease. All lesions found (2553) were classified according to their size and we subclassified small micronodules (<4.4 mm) as 'isolated', or as 'daughter', when they were in contact with consolidation (described as lesions ≥ 4.5 mm). Our data link the higher capacity to contain Mtb infection in cynomolgus with the reduced incidence of daughter micronodules, thus avoiding the development of consolidated lesions and their consequent enlargement and evolution to cavitation. In the case of rhesus, intradermal vaccination has a higher capacity to reduce the formation of daughter micronodules. This study supports the 'Bubble Model' defined with the C3HBe/FeJ mice and proposes a new method to evaluate outcomes in experimental models of TB in NHP based on CT images, which would fit a future machine learning approach to evaluate new vaccines.

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 ß-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.

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.

Determination of the Pharmacokinetics and Pharmacodynamics of Isoniazid, Rifampicin, Pyrazinamide and Ethambutol in a Cross-Over Cynomolgus Macaque Model of Mycobacterium tuberculosis Infection.

Innovative cross-over study designs were explored in non-human primate (NHP) studies to determine the value of this approach for the evaluation of drug efficacy against tuberculosis (TB). Firstly, the pharmacokinetics (PK) of each of the drugs Isoniazid (H), Rifampicin (R), Pyrazinamide (Z) and Ethambutol (E), that are standardly used for the treatment of tuberculosis, was established in the blood of macaques after oral dosing as a monotherapy or in combination. Two studies were conducted to evaluate the pharmacokinetics and pharmacodynamics of different drug combinations using cross-over designs. The first employed a balanced, three-period Pigeon design with an extra period; this ensured that treatment by period interactions and carry-over could be detected comparing the treatments HR, HZ and HRZ using H37Rv as the challenge strain of Mycobacterium tuberculosis (M. tb). Although the design accounted for considerable variability between animals, the three regimens evaluated could not be distinguished using any of the alternative endpoints assessed. However, the degree of pathology achieved using H37Rv in the model during this study was less than expected. Based on these findings, a second experiment using a classical AB/BA design comparing HE with HRZ was conducted using the M. tb Erdman strain. More extensive pathology was observed, and differences in computerized tomography (CT) scores and bacteriology counts in the lungs were detected, although due to the small group sizes, clearer differences were not distinguished. Type 1 T helper (Th1) cell response profiles were characterized using the IFN-γ ELISPOT assay and revealed differences between drug treatments that corresponded to decreases in disease burden. Therefore, the studies performed support the utility of the NHP model for the determination of PK/PD of TB drugs, although further work is required to optimize the use of cross-over study designs.

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.