Augmenting large language models with chemistry tools.

Large language models (LLMs) have shown strong performance in tasks across domains but struggle with chemistry-related problems. These models also lack access to external knowledge sources, limiting their usefulness in scientific applications. We introduce ChemCrow, an LLM chemistry agent designed to accomplish tasks across organic synthesis, drug discovery and materials design. By integrating 18 expert-designed tools and using GPT-4 as the LLM, ChemCrow augments the LLM performance in chemistry, and new capabilities emerge. Our agent autonomously planned and executed the syntheses of an insect repellent and three organocatalysts and guided the discovery of a novel chromophore. Our evaluation, including both LLM and expert assessments, demonstrates ChemCrow's effectiveness in automating a diverse set of chemical tasks. Our work not only aids expert chemists and lowers barriers for non-experts but also fosters scientific advancement by bridging the gap between experimental and computational chemistry.

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.

Animal behaviour on the move: the use of auxiliary information and semi-supervision to improve behavioural inferences from Hidden Markov Models applied to GPS tracking datasets.

BACKGROUND: State-space models, such as Hidden Markov Models (HMMs), are increasingly used to classify animal tracks into behavioural states. Typically, step length and turning angles of successive locations are used to infer where and when an animal is resting, foraging, or travelling. However, the accuracy of behavioural classifications is seldom validated, which may badly contaminate posterior analyses. In general, models appear to efficiently infer behaviour in species with discrete foraging and travelling areas, but classification is challenging for species foraging opportunistically across homogenous environments, such as tropical seas. Here, we use a subset of GPS loggers deployed simultaneously with wet-dry data from geolocators, activity measurements from accelerometers, and dive events from Time Depth Recorders (TDR), to improve the classification of HMMs of a large GPS tracking dataset (478 deployments) of red-billed tropicbirds (Phaethon aethereus), a poorly studied pantropical seabird. METHODS: We classified a subset of fixes as either resting, foraging or travelling based on the three auxiliary sensors and evaluated the increase in overall accuracy, sensitivity (true positive rate), specificity (true negative rate) and precision (positive predictive value) of the models in relation to the increasing inclusion of fixes with known behaviours. RESULTS: We demonstrate that even with a small informed sub-dataset (representing only 9% of the full dataset), we can significantly improve the overall behavioural classification of these models, increasing model accuracy from 0.77 ± 0.01 to 0.85 ± 0.01 (mean ± sd). Despite overall improvements, the sensitivity and precision of foraging behaviour remained low (reaching 0.37 ± 0.06, and 0.06 ± 0.01, respectively). CONCLUSIONS: This study demonstrates that the use of a small subset of auxiliary data with known behaviours can both validate and notably improve behavioural classifications of state space models of opportunistic foragers. However, the improvement is state-dependant and caution should be taken when interpreting inferences of foraging behaviour from GPS data in species foraging on the go across homogenous environments.

Cost-effectiveness analysis of a digital Diabetes Prevention Program (dDPP) in prediabetic patients.

OBJECTIVES: To assess the cost-effectiveness of a digital Diabetes Prevention Program (dDPP) in preventing type 2 diabetes mellitus among prediabetic patients from a health system perspective over a 10-year time horizon. METHODS: A Markov cohort model was constructed to assess the cost-effectiveness of dDPP compared to a small group education (SGE) intervention. Transition probabilities for the first year of the model were derived from two clinical trials on dDPP. Transition probabilities for longer-term effects were derived from meta-analyses on lifestyle and Diabetes Prevention Program interventions. Cost and health utilities were derived from published literature. Partial completion of interventions was incorporated to provide a robust prediction of a real-world deployment. Parameter uncertainties were assessed using univariate and probabilistic sensitivity analyses. Cost-effectiveness was measured by an incremental cost-effectiveness ratio (ICER) between dDPP and SGE from a health system perspective over a 10-year time horizon. RESULTS: The dDPP dominated the SGE at the $50,000, $100,000, and $150,000 willingness-to-pay thresholds per quality-adjusted life years (QALYs). The base case analysis at the $100,000 willingness-to-pay threshold (WTP) revealed a dominated ICER, with the SGE costing $1332 more and accruing an average of 0.04 fewer QALYs. Probabilistic sensitivity analysis showed that the dDPP was preferred in 64.4% of simulations across the $100,000 WTP thresholds. CONCLUSIONS: The findings comparing a dDPP to an SGE suggest that a dDPP can be cost-effective for patients with a high risk of developing type 2 diabetes.

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.

Differential remodeling in small and large murine airways revealed by novel whole lung airway analysis.

Airway remodeling occurs in chronic asthma leading to increased airway smooth muscle (ASM) mass and extracellular matrix (ECM) deposition. Although extensively studied in murine airways, studies report only selected larger airways at one time-point meaning the spatial distribution and resolution of remodeling are poorly understood. Here we use a new method allowing comprehensive assessment of the spatial and temporal changes in ASM, ECM, and epithelium in large numbers of murine airways after allergen challenge. Using image processing to analyze 20-50 airways per mouse from a whole lung section revealed increases in ASM and ECM after allergen challenge were greater in small and large rather than intermediate airways. ASM predominantly accumulated adjacent to the basement membrane, whereas ECM was distributed across the airway wall. Epithelial hyperplasia was most marked in small and intermediate airways. After challenge, ASM changes resolved over 7 days, whereas ECM and epithelial changes persisted. The new method suggests large and small airways remodel differently, and the long-term consequences of airway inflammation may depend more on ECM and epithelial changes than ASM. The improved quantity and quality of unbiased data provided by the method reveals important spatial differences in remodeling and could set new analysis standards for murine asthma models.

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.

Feces DNA analyses track the rehabilitation of a free-ranging beluga whale.

Following the sudden appearance, and subsequent efforts to support the survival of a beluga whale (Delphinapterus leucas) speculated to have been previously trained off the coast of Norway, we investigate the animal's ability to readapt to life in the wild. Dietary DNA (dDNA) analysis was used to assess diet throughout this rehabilitation process, and during a return to unassisted foraging and self-feeding. Metabarcoding of feces collected throughout this process, confirmed the diversification of the beluga whale's diet to local prey. These findings are indicative of improved foraging behavior, and the ability of this individual to resume wild foraging following a period of dependency in managed care. New insight of digestion rates, and the time window during which prey detection through dDNA analysis is appropriate was also obtained. Beyond the case study presented here, we demonstrate the power of dDNA analysis as a non-intrusive tool to assess the diet of large mammals and track progress adapting to life in the wild following release from captivity and rehabilitation programs.

High mortality rates in a juvenile free-ranging marine predator and links to dive and forage ability.

High juvenile mortality rates are typical of many long-lived marine vertebrate predators. Insufficient development in dive and forage ability is considered a key driver of this. However, direct links to survival outcome are sparse, particularly in free-ranging marine animals that may not return to land.In this study, we conduct exploratory investigations toward early mortality in juvenile southern elephant seals Mirounga leonina. Twenty postweaning pups were equipped with (a) a new-generation satellite relay data tag, capable of remotely transmitting fine-scale behavioral movements from accelerometers, and (b) a location transmitting only tag (so that mortality events could be distinguished from device failures). Individuals were followed during their first trip at sea (until mortality or return to land). Two analyses were conducted. First, the behavioral movements and encountered environmental conditions of nonsurviving pups were individually compared to temporally concurrent observations from grouped survivors. Second, common causes of mortality were investigated using Cox's proportional hazard regression and penalized shrinkage techniques.Nine individuals died (two females and seven males) and 11 survived (eight females and three males). All but one individual died before the return phase of their first trip at sea, and all but one were negatively buoyant. Causes of death were variable, although common factors included increased horizontal travel speeds and distances, decreased development in dive and forage ability, and habitat type visited (lower sea surface temperatures and decreased total [eddy] kinetic energy).For long-lived marine vertebrate predators, such as the southern elephant seal, the first few months of life following independence represent a critical period, when small deviations in behavior from the norm appear sufficient to increase mortality risk. Survival rates may subsequently be particularly vulnerable to changes in climate and environment, which will have concomitant consequences on the demography and dynamics of populations.

Body condition influences ontogeny of foraging behavior in juvenile southern elephant seals.

Ontogeny of diving and foraging behavior in marine top predators is poorly understood despite its importance in population recruitment. This lack of knowledge is partly due to the difficulties of monitoring juveniles in the wild, which is linked to high mortality early in life. Pinnipeds are good models for studying the development of foraging behaviors because juveniles are large enough to robustly carry tracking devices for many months. Moreover, parental assistance is absent after a juvenile departs for its first foraging trip, minimizing confounding effects of parental input on the development of foraging skills. In this study, we tracked 20 newly weaned juvenile southern elephant seals from Kerguelen Islands for up to 338 days during their first trip at sea following weaning. We used a new generation of satellite relay tags, which allow for the transmission of dive, accelerometer, and location data. We also monitored, at the same time, nine adult females from the colony during their post-breeding trips, in order to compare diving and foraging behaviors. Juveniles showed a gradual improvement through time in their foraging skills. Like adults females, they remarkably adjusted their swimming effort according to temporal changes in buoyancy (i.e., a proxy of their body condition). They also did not appear to exceed their aerobic physiological diving limits, although dives were constrained by their smaller size compared to adults. Changes in buoyancy appeared to also influence their decision to either keep foraging or return to land, alongside the duration of their haul outs and choice of foraging habitat (oceanic vs. plateau). Further studies are thus needed to better understand how patterns in juveniles survival, and therefore elephant seal populations, might be affected by their changes in foraging skills and changes in their environmental conditions.

Processing of acceleration and dive data on-board satellite relay tags to investigate diving and foraging behaviour in free-ranging marine predators.

Biologging technologies are changing the way in which the marine environment is observed and monitored. However, because device retrieval is typically required to access the high-resolution data they collect, their use is generally restricted to those animals that predictably return to land. Data abstraction and transmission techniques aim to address this, although currently these are limited in scope and do not incorporate, for example, acceleration measurements which can quantify animal behaviours and movement patterns over fine-scales.In this study, we present a new method for the collection, abstraction and transmission of accelerometer data from free-ranging marine predators via the Argos satellite system. We test run the technique on 20 juvenile southern elephant seals Mirounga leonina from the Kerguelen Islands during their first months at sea following weaning. Using retrieved archival data from nine individuals that returned to the colony, we compare and validate abstracted transmissions against outputs from established accelerometer processing procedures.Abstracted transmissions included estimates, across five segments of a dive profile, of time spent in prey catch attempt (PrCA) behaviours, swimming effort and pitch. These were then summarised and compared to archival outputs across three dive phases: descent, bottom and ascent. Correlations between the two datasets were variable but generally good (dependent on dive phase, marginal R2 values of between .45 and .6 to >.9) and consistent between individuals. Transmitted estimates of PrCA behaviours and swimming effort were positively biased to those from archival processing.Data from this study represent some of the first remotely transmitted quantifications from accelerometers. The methods presented and analysed can be used to provide novel insight towards the behaviours and movements of free-ranging marine predators, such as juvenile southern elephant seals, from whom logger retrieval is challenging. Future applications could however benefit from some adaption, particularly to reduce positive bias in transmitted PrCA behaviours and swimming effort, for which this study provides useful insight.

Early-life foraging: Behavioral responses of newly fledged albatrosses to environmental conditions.

In order to survive and later recruit into a population, juvenile animals need to acquire resources through the use of innate and/or learnt behaviors in an environment new to them. For far-ranging marine species, such as the wandering albatross Diomedea exulans, this is particularly challenging as individuals need to be able to rapidly adapt and optimize their movement strategies in response to the highly dynamic and heterogeneous nature of their open-ocean pelagic habitats. Critical to this is the development and flexibility of dispersal and exploratory behaviors. Here, we examine the movements of eight juvenile wandering albatrosses, tracked using GPS/Argos satellite transmitters for eight months following fledging, and compare these to the trajectories of 17 adults to assess differences and similarities in behavioral strategies through time. Behavioral clustering algorithms (Expectation Maximization binary Clustering) were combined with multinomial regression analyses to investigate changes in behavioral mode probabilities over time, and how these may be influenced by variations in day duration and in biophysical oceanographic conditions. We found that juveniles appeared to quickly acquire the same large-scale behavioral strategies as those employed by adults, although generally more time was spent resting at night. Moreover, individuals were able to detect and exploit specific oceanographic features in a manner similar to that observed in adults. Together, the results of this study suggest that while shortly after fledging juvenile wandering albatrosses are able to employ similar foraging strategies to those observed in adults, additional skills need to be acquired during the immature period before the efficiency of these behaviors matches that of adults.

Sexually dimorphic expression of secreted frizzled-related (SFRP) genes in the developing mouse Müllerian duct.

In developing male embryos, the female reproductive tract primordia (Müllerian ducts) regress due to the production of testicular anti-Müllerian hormone (AMH). Because of the association between secreted frizzled-related proteins (SFRPs) and apoptosis, their reported developmental expression patterns and the role of WNT signaling in female reproductive tract development, we examined expression of Sfrp2 and Sfrp5 during development of the Müllerian duct in male (XY) and female (XX) mouse embryos. We show that expression of both Sfrp2 and Sfrp5 is dynamic and sexually dimorphic. In addition, the male-specific expression observed for both genes prior to the onset of regression is absent in mutant male embryos that fail to undergo Müllerian duct regression. We identified ENU-induced point mutations in Sfrp5 and Sfrp2 that are predicted to severely disrupt the function of these genes. Male embryos and adults homozygous for these mutations, both individually and in combination, are viable and apparently fertile with no overt abnormalities of reproductive tract development.

A gene-driven ENU-based approach to generating an allelic series in any gene.

N-ethyl-N-nitrosourea (ENU) introduces mutations throughout the mouse genome at relatively high efficiency. Successful high-throughput phenotype screens have been reported and alternative screens using sequence-based approaches have been proposed. For the purpose of generating an allelic series in selected genes by a sequence-based approach, we have constructed an archive of over 4000 DNA samples from individual F1 ENU-mutagenized mice paralleled by frozen sperm samples. Together with our previously reported archive, the total size now exceeds 6000 individuals. A gene-based screen of 27.4 Mbp of DNA, carried out using denaturing high-performance liquid chromatography (DHPLC), found a mutation rate of 1 in 1.01 Mbp of which 1 in 1.82 Mbp were potentially functional. Screening of whole or selected regions of genes on subsets of the archive has allowed us to identify 15 new alleles from 9 genes out of 15 tested. This is a powerful adjunct to conventional mutagenesis strategies and has the advantage of generating a variety of alleles with potentially different phenotypic outcomes that facilitate the investigation of gene function. It is now available to academic collaborators as a community resource.