Search

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.

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. 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 \selfieslib, where we have also made major advances with respect to design, efficiency, and supported features. Hence, we present the current status of \selfieslib (version 2.1.1) in this manuscript., Comment: 11 pages, 2 figures

Evolutionary symbolic regression (SR) fits a symbolic equation to data, which gives a concise interpretable model. We explore using SR as a method to propose which data to gather in an active learning setting with physical constraints. SR with active learning proposes which experiments to do next. Active learning is done with query by committee, where the Pareto frontier of equations is the committee. The physical constraints improve proposed equations in very low data settings. These approaches reduce the data required for SR and achieves state of the art results in data required to rediscover known equations.

The dual use of machine learning applications, where models can be used for both beneficial and malicious purposes, presents a significant challenge. This has recently become a particular concern in chemistry, where chemical datasets containing sensitive labels (e.g. toxicological information) could be used to develop predictive models that identify novel toxins or chemical warfare agents. To mitigate dual use risks, we propose a model-agnostic method of selectively noising datasets while preserving the utility of the data for training deep neural networks in a beneficial region. We evaluate the effectiveness of the proposed method across least squares, a multilayer perceptron, and a graph neural network. Our findings show selectively noised datasets can induce model variance and bias in predictions for sensitive labels with control, suggesting the safe sharing of datasets containing sensitive information is feasible. We also find omitting sensitive data often increases model variance sufficiently to mitigate dual use. This work is proposed as a foundation for future research on enabling more secure and collaborative data sharing practices and safer machine learning applications in chemistry.

Large language models (LLMs) are able to do accurate classification with zero or only a few examples (in-context learning). We show a prompting system that enables regression with uncertainty for in-context learning with frozen LLM (GPT-3, GPT-3.5, and GPT-4) models, allowing predictions without features or architecture tuning. By incorporating uncertainty, our approach enables Bayesian optimization for catalyst or molecule optimization using natural language, eliminating the need for training or simulation. Here, we performed the optimization using the synthesis procedure of catalysts to predict properties. Working with natural language mitigates difficulty synthesizability since the literal synthesis procedure is the model's input. We showed that in-context learning could improve past a model context window (maximum number of tokens the model can process at once) as data is gathered via example selection, allowing the model to scale better. Although our method does not outperform all baselines, it requires zero training, feature selection, and minimal computing while maintaining satisfactory performance. We also find Gaussian Process Regression on text embeddings is strong at Bayesian optimization. The code is available in our GitHub repository: https://github.com/ur-whitelab/BO-LIFT

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., This research was funded by ‘La Caixa’ Foundation (ID 100010434), under agreement LCF/PR/GN16/10290002 and AGAUR 2017 SGR 00500, 2021 SGR 00920 and 2021 SGR 00934., Objectius de Desenvolupament Sostenible::3 - Salut i Benestar, Postprint (published version)

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.

Large Language Models (LLMs) generalize well across language tasks, but suffer from hallucinations and uninterpretability, making it difficult to assess their accuracy without ground-truth. Retrieval-Augmented Generation (RAG) models have been proposed to reduce hallucinations and provide provenance for how an answer was generated. Applying such models to the scientific literature may enable large-scale, systematic processing of scientific knowledge. We present PaperQA, a RAG agent for answering questions over the scientific literature. PaperQA is an agent that performs information retrieval across full-text scientific articles, assesses the relevance of sources and passages, and uses RAG to provide answers. Viewing this agent as a question answering model, we find it exceeds performance of existing LLMs and LLM agents on current science QA benchmarks. To push the field closer to how humans perform research on scientific literature, we also introduce LitQA, a more complex benchmark that requires retrieval and synthesis of information from full-text scientific papers across the literature. Finally, we demonstrate PaperQA's matches expert human researchers on LitQA.

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 usable at https://mol.dev.

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.

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.

Vehicle-to-grid (V2G) is a promising approach to solve the problem of grid-level intermittent supply and demand mismatch, caused due to renewable energy resources, because it uses the existing resource of electric vehicle (EV) batteries as the energy storage medium. EV battery design together with an impetus on profitability for participating EV owners is pivotal for V2G success. To better understand what battery device parameters are most important for V2G adoption, we model the economics of V2G process under realistic conditions. Most previous studies that perform V2G economic analysis, assume ideal driving conditions, use linear battery degradation models, or only consider V2G for ancillary services. Our model accounts realistic battery degradation, empirical charging efficiencies, for randomness in commute behavior, and historic hourly electricity prices in six cities in the United States. We model user behavior with Bayesian optimization to provide a best-case scenario for V2G. Across all cities, we find that charging rate and efficiency are the most important factors that determine EV users' profits. Surprisingly, EV battery cost and thus degradation due to cycling has little effect. These findings should help focus research on figures of merit that better reflect real usage of batteries in a V2G economy., Comment: 17 Pages, 10 Figures, 1 Table

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 1,206 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.

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-¿ spot forming units (SFU) measured in peripheral blood mononuclear cells (PBMCs) by ELISpot, although the induction of IFN-¿ 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, Peer Reviewed, Postprint (published version)

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 multi-task benchmark database of peptides designed to advance these methods for experimental design. Each task is 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 average accuracy across datasets. Combining meta-learning with active learning offers inconsistent benefits., Comment: 19 pages, 8 figures, 9 tables

PurposeIn conventional neurons, Ca2+ enters presynaptic terminals during an action potential and its increased local concentration triggers transient exocytosis. In contrast, vertebrate photoreceptors are nonspiking neurons that maintain sustained depolarization and neurotransmitter release from ribbon synapses in darkness and produce light-dependent graded hyperpolarizing responses. Rods transmit single photon responses with high fidelity, whereas cones are less sensitive and exhibit faster response kinetics. These differences are likely due to variations in presynaptic Ca2+ dynamics. Metabolic coupling and cross-talk between mitochondria, endoplasmic reticulum (ER), plasma membrane Ca2+ ATPase (PMCA), and Na+-Ca2+ exchanger (NCX) coordinately control presynaptic ATP production and Ca2+ dynamics. The goal of our structural and functional studies was to determine the spatiotemporal regulation of ATP and Ca2+ dynamics in rod spherules and cone pedicles.MethodsCentral retina tissue from C57BL/6 mice was used. Laser scanning confocal microscopy (LSCM) experiments were conducted on fixed-frozen vertical sections. Primary antibodies were selected for their tissue/cellular specificity and ability to recognize single, multiple or all splice variants of selected isoforms. Electron microscopy (EM) and 3-D electron tomography (ET) studies used our standard procedures on thin- and thick-sectioned retinas, respectively. Calibrated fluo-3-Ca2+ imaging experiments of dark- and light-adapted rod and cone terminals in retinal slices were conducted.ResultsConfocal microscopy showed that mitochondria, ER, PMCA, and NCX1 exhibited distinct retinal lamination patterns and differential distribution in photoreceptor synapses. Antibodies for three distinct mitochondrial compartments differentially labeled retinal areas with high metabolic demand: rod and cone inner segments, previously undescribed cone juxtanuclear mitochondria and the two plexiform layers. Rod spherule membranes uniforml

PurposeIn conventional neurons, Ca2+ enters presynaptic terminals during an action potential and its increased local concentration triggers transient exocytosis. In contrast, vertebrate photoreceptors are nonspiking neurons that maintain sustained depolarization and neurotransmitter release from ribbon synapses in darkness and produce light-dependent graded hyperpolarizing responses. Rods transmit single photon responses with high fidelity, whereas cones are less sensitive and exhibit faster response kinetics. These differences are likely due to variations in presynaptic Ca2+ dynamics. Metabolic coupling and cross-talk between mitochondria, endoplasmic reticulum (ER), plasma membrane Ca2+ ATPase (PMCA), and Na+-Ca2+ exchanger (NCX) coordinately control presynaptic ATP production and Ca2+ dynamics. The goal of our structural and functional studies was to determine the spatiotemporal regulation of ATP and Ca2+ dynamics in rod spherules and cone pedicles.MethodsCentral retina tissue from C57BL/6 mice was used. Laser scanning confocal microscopy (LSCM) experiments were conducted on fixed-frozen vertical sections. Primary antibodies were selected for their tissue/cellular specificity and ability to recognize single, multiple or all splice variants of selected isoforms. Electron microscopy (EM) and 3-D electron tomography (ET) studies used our standard procedures on thin- and thick-sectioned retinas, respectively. Calibrated fluo-3-Ca2+ imaging experiments of dark- and light-adapted rod and cone terminals in retinal slices were conducted.ResultsConfocal microscopy showed that mitochondria, ER, PMCA, and NCX1 exhibited distinct retinal lamination patterns and differential distribution in photoreceptor synapses. Antibodies for three distinct mitochondrial compartments differentially labeled retinal areas with high metabolic demand: rod and cone inner segments, previously undescribed cone juxtanuclear mitochondria and the two plexiform layers. Rod spherule membranes uniforml

Thesis (Ph. D.)--University of Rochester. Department of Chemical Engineering, 2020., Explaining and predicting experimental results are the goals of molecular simulations. Molecular simulations that reproduce experimental results give a molecular explanation of phenomena that are fundamentally inaccessible to experiments. Such important phenomena include fast time-dependent processes, like the binding of a ligand to an active site of an enzyme or protein misfolding. These processes are hard to measure experimentally and even if they are time-independent they may be challenging to observe experimentally. These molecular processes are nevertheless important and have wide applications from drug design to discovering novel materials with specific properties. Simulations excel at elucidating these phenomena. Simulations allow one to see the nanoscale details that lead to bulk properties and macroscopic observables. Understanding molecule mechanisms from simulations' perspective can lead to innovations by providing insight, giving guidance to design of materials that are nonfouling, antimicrobial, etc. Because the goal is understanding an experiment, utilizing the experimental results to improve the accuracy of simulations is an obvious advantage. Experiment Directed Simulations (EDS) is a method of minimally biasing a simulation such that the simulation matches experimental results. This research focuses on implementing EDS on biomolecules, such as tripeptides and amyloid peptides, to analyze the structure and dynamics of these systems.

Thesis (Ph. D.)--University of Rochester. Department of Chemical Engineering, 2020., Machine learning (ML) is a broad, flexible suite of applied statistics tools combined with optimization algorithms, made possible by the advent of modern computing power. The strength of ML methods is limited by data availability and complexity of the task to learn, but nonetheless, ML has seen extensive use in the fields of medicine, technology, and engineering. It also has applications in chemical engineering and computational chemistry. In this thesis, I present various such applications. Using ML for computer vision, I helped make a mixed-reality educational tool for chemical engineering undergraduates, which led to further work on a virtual-reality app for high school chemistry students. I have also used ML methods for functional design of antimicrobial peptides, and investigated ways to address the problem of dataset sizes in this setting, to increase the chances of its use in experimental laboratory settings, where data is scarce. I also helped develop a framework for ML in molecular dynamics simulations to reduce the barrier of entry for non-experts in ML. Finally, my most recent work has been toward applying biasing methods from statistical mechanics to COVID-19 disease modeling to improve model prediction without parameter fitting. The broad range of applications presented here only scratches the surface of what can be done with the powerful combination of traditional numerical models augmented by machine learning methods, serving to illustrate how much can be done with its wide adoption in the field of chemical engineering, and the sciences as a whole.