Toward universal cell embeddings: integrating single-cell RNA-seq datasets across species with SATURN.

Analysis of single-cell datasets generated from diverse organisms offers unprecedented opportunities to unravel fundamental evolutionary processes of conservation and diversification of cell types. However, interspecies genomic differences limit the joint analysis of cross-species datasets to homologous genes. Here we present SATURN, a deep learning method for learning universal cell embeddings that encodes genes' biological properties using protein language models. By coupling protein embeddings from language models with RNA expression, SATURN integrates datasets profiled from different species regardless of their genomic similarity. SATURN can detect functionally related genes coexpressed across species, redefining differential expression for cross-species analysis. Applying SATURN to three species whole-organism atlases and frog and zebrafish embryogenesis datasets, we show that SATURN can effectively transfer annotations across species, even when they are evolutionarily remote. We also demonstrate that SATURN can be used to find potentially divergent gene functions between glaucoma-associated genes in humans and four other species.

Microbial communities are indicators of parasite infection status.

Growing evidence suggests that microbiomes have been shaping the evolutionary pathways of macroorganisms for millennia and that these tiny symbionts can influence, and possibly even control, species interactions like host-parasite relationships. Yet, while studies have investigated host-parasites and microbiomes separately, little has been done to understand all three groups synergistically. Here, we collected infected and uninfected Eurypanopeus depressus crab hosts from a coastal North Carolina oyster reef three times over 4 months. Infected crabs demonstrated an external stage of the rhizocephalan parasite, Loxothylacus panopaei. Community analyses revealed that microbial richness and diversity were significantly different among tissue types (uninfected crab, infected crab, parasite externae and parasite larvae) and over time (summer and fall). Specifically, the microbial communities from parasite externae and larvae had similar microbiomes that were consistent through time. Infected crabs demonstrated microbial communities spanning those of their host and parasite, while uninfected crabs showed more distinctive communities with greater variability over time. Microbial communities were also found to be indicators of early-stage infections. Resolving the microbial community composition of a host and its parasite is an important step in understanding the microbiome's role in the host-parasite relationship and determining how this tripartite relationship impacts coevolutionary processes.

7-UP: Generating in silico CODEX from a small set of immunofluorescence markers.

Multiplex immunofluorescence (mIF) assays multiple protein biomarkers on a single tissue section. Recently, high-plex CODEX (co-detection by indexing) systems enable simultaneous imaging of 40+ protein biomarkers, unlocking more detailed molecular phenotyping, leading to richer insights into cellular interactions and disease. However, high-plex data can be slower and more costly to collect, limiting its applications, especially in clinical settings. We propose a machine learning framework, 7-UP, that can computationally generate in silico 40-plex CODEX at single-cell resolution from a standard 7-plex mIF panel by leveraging cellular morphology. We demonstrate the usefulness of the imputed biomarkers in accurately classifying cell types and predicting patient survival outcomes. Furthermore, 7-UP's imputations generalize well across samples from different clinical sites and cancer types. 7-UP opens the possibility of in silico CODEX, making insights from high-plex mIF more widely available.

Deep learning-guided discovery of an antibiotic targeting Acinetobacter baumannii.

Acinetobacter baumannii is a nosocomial Gram-negative pathogen that often displays multidrug resistance. Discovering new antibiotics against A. baumannii has proven challenging through conventional screening approaches. Fortunately, machine learning methods allow for the rapid exploration of chemical space, increasing the probability of discovering new antibacterial molecules. Here we screened ~7,500 molecules for those that inhibited the growth of A. baumannii in vitro. We trained a neural network with this growth inhibition dataset and performed in silico predictions for structurally new molecules with activity against A. baumannii. Through this approach, we discovered abaucin, an antibacterial compound with narrow-spectrum activity against A. baumannii. Further investigations revealed that abaucin perturbs lipoprotein trafficking through a mechanism involving LolE. Moreover, abaucin could control an A. baumannii infection in a mouse wound model. This work highlights the utility of machine learning in antibiotic discovery and describes a promising lead with targeted activity against a challenging Gram-negative pathogen.

From patterns to patients: Advances in clinical machine learning for cancer diagnosis, prognosis, and treatment.

Machine learning (ML) is increasingly used in clinical oncology to diagnose cancers, predict patient outcomes, and inform treatment planning. Here, we review recent applications of ML across the clinical oncology workflow. We review how these techniques are applied to medical imaging and to molecular data obtained from liquid and solid tumor biopsies for cancer diagnosis, prognosis, and treatment design. We discuss key considerations in developing ML for the distinct challenges posed by imaging and molecular data. Finally, we examine ML models approved for cancer-related patient usage by regulatory agencies and discuss approaches to improve the clinical usefulness of ML.

Towards Universal Cell Embeddings: Integrating Single-cell RNA-seq Datasets across Species with SATURN.

Analysis of single-cell datasets generated from diverse organisms offers unprecedented opportunities to unravel fundamental evolutionary processes of conservation and diversification of cell types. However, inter-species genomic differences limit the joint analysis of cross-species datasets to homologous genes. Here, we present SATURN, a deep learning method for learning universal cell embeddings that encodes genes' biological properties using protein language models. By coupling protein embeddings from language models with RNA expression, SATURN integrates datasets profiled from different species regardless of their genomic similarity. SATURN has a unique ability to detect functionally related genes co-expressed across species, redefining differential expression for cross-species analysis. We apply SATURN to three species whole-organism atlases and frog and zebrafish embryogenesis datasets. We show that cell embeddings learnt in SATURN can be effectively used to transfer annotations across species and identify both homologous and species-specific cell types, even across evolutionarily remote species. Finally, we use SATURN to reannotate the five species Cell Atlas of Human Trabecular Meshwork and Aqueous Outflow Structures and find evidence of potentially divergent functions between glaucoma associated genes in humans and other species.

The timing versus resource problem in nonnative sentence processing: Evidence from a time-frequency analysis of anaphora resolution in successive wh-movement in native and nonnative speakers of French.

Nonnative processing has been argued to reflect either reduced processing capacity or delayed timing of structural analysis compared to the extraction of lexical/semantic information. The current study simultaneously investigates timing and resource allocation through a time-frequency analysis of the intrinsic neural activity during syntactic processing in native and English-speaking nonnative speakers of French. It involved structurally constrained anaphora resolution in bi-clausal wh-filler-gap dependencies such as Quelle décision à propos de lui est-ce que Paul a dit que Lydie avait rejetée sans hésitation? 'Which decision about him did Paul say that Lydie rejected without hesitation?'. We tested the hypothesis that nonnative speakers may allocate greater resources than native speakers to the computation of syntactic representations based on the grammatical specifications encoded in lexical entries, though both native and nonnative processing involves the immediate application of structural constraints. This distinct resource allocation is likely to arise in response to higher activation thresholds for nonnative knowledge acquired after the first language grammar has been fully acquired. To examine this bias in nonnative neurocognitive processing, we manipulated the wh-filler to contain either a lexically specified noun complement such as à propos de lui 'about him' or a non-lexcially specified noun phrase modifier such as le concernant 'concerning him'. We focused on processing at the intermediate gap site, that is, the point of information exchange between the matrix and the embedded clauses by adopting a measurement window corresponding to the bridge verb dit 'said' and subordinator que 'that' introducing the embedded clause. Our results showed that structural constraints on anaphora produced event-related spectral perturbations at 13-14Hz early into the presentation of the bridge verb across groups. An interaction of structural constraints on anaphora with group was found at 18-19Hz early into the presentation of the bridge verb. In this interaction, the nonnative-speaker activity at 18-19Hz echoed the concurrent general patterns at 13-14Hz, whereas the native-speaker activity revealed distinct power at 18-19Hz and at 13-14Hz. There was no evidence of delay of structural constraints on intermediate gaps with respect to lexical access to the bridge verb and subordinator. However, nonnative speakers' allocation of power in cell assembly synchronizations of fillers and gaps at the intermediate gap site reflected the grammatical specifications lexically encoded in the fillers.

ADMET-AI: A machine learning ADMET platform for evaluation of large-scale chemical libraries.

Summary: The emergence of large chemical repositories and combinatorial chemical spaces, coupled with high-throughput docking and generative AI, have greatly expanded the chemical diversity of small molecules for drug discovery. Selecting compounds for experimental validation requires filtering these molecules based on favourable druglike properties, such as Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET). We developed ADMET-AI, a machine learning platform that provides fast and accurate ADMET predictions both as a website and as a Python package. ADMET-AI has the highest average rank on the TDC ADMET Benchmark Group leaderboard, and it is currently the fastest web-based ADMET predictor, with a 45% reduction in time compared to the next fastest ADMET web server. ADMET-AI can also be run locally with predictions for one million molecules taking just 3.1 hours. Availability and Implementation: The ADMET-AI platform is freely available both as a web server at admet.ai.greenstonebio.com and as an open-source Python package for local batch prediction at github.com/swansonk14/admet_ai (also archived on Zenodo at doi.org/10.5281/zenodo.10372930 ). All data and models are archived on Zenodo at doi.org/10.5281/zenodo.10372418 .

Graph deep learning for the characterization of tumour microenvironments from spatial protein profiles in tissue specimens.

Multiplexed immunofluorescence imaging allows the multidimensional molecular profiling of cellular environments at subcellular resolution. However, identifying and characterizing disease-relevant microenvironments from these rich datasets is challenging. Here we show that a graph neural network that leverages spatial protein profiles in tissue specimens to model tumour microenvironments as local subgraphs captures distinctive cellular interactions associated with differential clinical outcomes. We applied this spatial cellular-graph strategy to specimens of human head-and-neck and colorectal cancers assayed with 40-plex immunofluorescence imaging to identify spatial motifs associated with cancer recurrence and with patient survival after treatment. The graph deep learning model was substantially more accurate in predicting patient outcomes than deep learning approaches that model spatial data on the basis of the local composition of cell types, and it generated insights into the effect of the spatial compartmentalization of tumour cells and granulocytes on patient prognosis. Local graphs may also aid in the analysis of disease-relevant motifs in histology samples characterized via spatial transcriptomics and other -omics techniques.

Cerebral Venous Sinus Thrombosis Associated With Coronavirus Disease 2019: Case Report and Review of the Literature.

INTRODUCTION: Coronavirus disease 2019 (COVID-19) is associated with significant risk of acute thrombosis. We present a case report of a patient with cerebral venous sinus thrombosis (CVST) associated with COVID-19 and performed a literature review of CVST associated with COVID-19 cases. CASE REPORT: A 38-year-old woman was admitted with severe headache and acute altered mental status a week after confirmed diagnosis of COVID-19. Magnetic resonance imaging brain showed diffuse venous sinus thrombosis involving the superficial and deep veins, and diffuse edema of bilateral thalami, basal ganglia and hippocampi because of venous infarction. Her neurological exam improved with anticoagulation (AC) and was subsequently discharged home. We identified 43 patients presenting with CVST associated with COVID-19 infection. 56% were male with mean age of 51.8±18.2 years old. The mean time of CVST diagnosis was 15.6±23.7 days after onset of COVID-19 symptoms. Most patients (87%) had thrombosis of multiple dural sinuses and parenchymal changes (79%). Almost 40% had deep cerebral venous system thrombosis. Laboratory findings revealed elevated mean D-dimer level (7.14/mL±12.23 mg/L) and mean fibrinogen level (4.71±1.93 g/L). Less than half of patients had prior thrombotic risk factors. Seventeen patients (52%) had good outcomes (mRS <=2). The mortality rate was 39% (13 patients). CONCLUSION: CVST should be in the differential diagnosis when patients present with acute neurological symptoms in this COVID pandemic. The mortality rate of CVST associated with COVID-19 can be very high, therefore, early diagnosis and prompt treatment are crucial to the outcomes of these patients.

Invasion of the body snatchers: the role of parasite introduction in host distribution and response to salinity in invaded estuaries.

In dynamic systems, organisms are faced with variable selective forces that may impose trade-offs. In estuaries, salinity is a strong driver of organismal diversity, while parasites shape species distributions and demography. We tested for trade-offs between low-salinity stress and parasitism in an invasive castrating parasite and its mud crab host along salinity gradients of two North Carolina rivers. We performed field surveys every six to eight weeks over 3 years to determine factors influencing parasite prevalence, host abundance, and associated taxa diversity. We also looked for signatures of low-salinity stress in the host by examining its response (time-to-right and gene expression) to salinity. We found salinity and temperature significantly affected parasite prevalence, with low-salinity sites (less than 10 practical salinity units (PSU)) lacking infection, and populations in moderate salinities at warmer temperatures reaching prevalence as high as 60%. Host abundance was negatively associated with parasite prevalence. Host gene expression was plastic to acclimation salinity, but several osmoregulatory and immune-related genes demonstrated source-dependent salinity response. We identified a genetic marker that was strongly associated with salinity against a backdrop of no neutral genetic structure, suggesting possible selection on standing variation. Our study illuminates how selective trade-offs in naturally dynamic systems may shape host evolutionary ecology.

Uncertainty Quantification Using Neural Networks for Molecular Property Prediction.

Uncertainty quantification (UQ) is an important component of molecular property prediction, particularly for drug discovery applications where model predictions direct experimental design and where unanticipated imprecision wastes valuable time and resources. The need for UQ is especially acute for neural models, which are becoming increasingly standard yet are challenging to interpret. While several approaches to UQ have been proposed in the literature, there is no clear consensus on the comparative performance of these models. In this paper, we study this question in the context of regression tasks. We systematically evaluate several methods on five regression data sets using multiple complementary performance metrics. Our experiments show that none of the methods we tested is unequivocally superior to all others, and none produces a particularly reliable ranking of errors across multiple data sets. While we believe that these results show that existing UQ methods are not sufficient for all common use cases and further research is needed, we conclude with a practical recommendation as to which existing techniques seem to perform well relative to others.

Seizure and neuropsychological outcomes in a large series of selective amygdalohippocampectomies with a minimally invasive subtemporal approach.

OBJECTIVE: Debate continues over proper surgical treatment for mesial temporal lobe epilepsy (MTLE). Few large comprehensive studies exist that have examined outcomes for the subtemporal selective amygdalohippocampectomy (sSAH) approach. This study describes a minimally invasive technique for sSAH and examines seizure and neuropsychological outcomes in a large series of patients who underwent sSAH for MTLE. METHODS: Data for 152 patients (94 women, 61.8%; 58 men, 38.2%) who underwent sSAH performed by a single surgeon were retrospectively reviewed. The sSAH technique involves a small, minimally invasive opening and preserves the anterolateral temporal lobe and the temporal stem. RESULTS: All patients in the study had at least 1 year of follow-up (mean [SD] 4.52 [2.57] years), of whom 57.9% (88/152) had Engel class I seizure outcomes. Of the patients with at least 2 years of follow-up (mean [SD] 5.2 [2.36] years), 56.5% (70/124) had Engel class I seizure outcomes. Preoperative and postoperative neuropsychological test results indicated no significant change in intelligence, verbal comprehension, perceptual reasoning, attention and processing, cognitive flexibility, visuospatial memory, or mood. There was a significant change in word retrieval regardless of the side of surgery and a significant change in verbal memory in patients who underwent dominant-side resection (p < 0.05). Complication rates were low, with a 1.3% (2/152) permanent morbidity rate and 0.0% mortality rate. CONCLUSIONS: This study reports a large series of patients who have undergone sSAH, with a comprehensive presentation of a minimally invasive technique. The sSAH approach described in this study appears to be a safe, effective, minimally invasive technique for the treatment of MTLE.

A Deep Learning Approach to Antibiotic Discovery.

Due to the rapid emergence of antibiotic-resistant bacteria, there is a growing need to discover new antibiotics. To address this challenge, we trained a deep neural network capable of predicting molecules with antibacterial activity. We performed predictions on multiple chemical libraries and discovered a molecule from the Drug Repurposing Hub-halicin-that is structurally divergent from conventional antibiotics and displays bactericidal activity against a wide phylogenetic spectrum of pathogens including Mycobacterium tuberculosis and carbapenem-resistant Enterobacteriaceae. Halicin also effectively treated Clostridioides difficile and pan-resistant Acinetobacter baumannii infections in murine models. Additionally, from a discrete set of 23 empirically tested predictions from >107 million molecules curated from the ZINC15 database, our model identified eight antibacterial compounds that are structurally distant from known antibiotics. This work highlights the utility of deep learning approaches to expand our antibiotic arsenal through the discovery of structurally distinct antibacterial molecules.

Deep learning for automated classification and characterization of amorphous materials.

It is difficult to quantify structure-property relationships and to identify structural features of complex materials. The characterization of amorphous materials is especially challenging because their lack of long-range order makes it difficult to define structural metrics. In this work, we apply deep learning algorithms to accurately classify amorphous materials and characterize their structural features. Specifically, we show that convolutional neural networks and message passing neural networks can classify two-dimensional liquids and liquid-cooled glasses from molecular dynamics simulations with greater than 0.98 AUC, with no a priori assumptions about local particle relationships, even when the liquids and glasses are prepared at the same inherent structure energy. Furthermore, we demonstrate that message passing neural networks surpass convolutional neural networks in this context in both accuracy and interpretability. We extract a clear interpretation of how message passing neural networks evaluate liquid and glass structures by using a self-attention mechanism. Using this interpretation, we derive three novel structural metrics that accurately characterize glass formation. The methods presented here provide a procedure to identify important structural features in materials that could be missed by standard techniques and give unique insight into how these neural networks process data.

Temporal Horn Choroid Plexus Papilloma Presenting with Seizures in Adulthood: Clinical Case Report and Review of the Literature.

BACKGROUND: Choroid plexus papillomas (CPPs) are benign World Health Organization grade I tumors that comprise 2%-4% of all brain tumors among children and less than 1% of brain tumors in adults. Most adult cases occur in the fourth ventricle, with only 1 previous report describing an adult patient with a temporal horn CPP. CASE DESCRIPTION: We report a rare case of a temporal horn CPP presenting in an adult with seizures. We performed a minimally invasive subtemporal approach for gross total resection of the lesion. CONCLUSIONS: CPP presenting in the temporal horn is rare among adults. We discuss the surgical nuances of the subtemporal approach for resection and review the literature regarding adult presentation of CPP and the treatment strategies for adult CPP.

Analyzing Learned Molecular Representations for Property Prediction.

Advancements in neural machinery have led to a wide range of algorithmic solutions for molecular property prediction. Two classes of models in particular have yielded promising results: neural networks applied to computed molecular fingerprints or expert-crafted descriptors and graph convolutional neural networks that construct a learned molecular representation by operating on the graph structure of the molecule. However, recent literature has yet to clearly determine which of these two methods is superior when generalizing to new chemical space. Furthermore, prior research has rarely examined these new models in industry research settings in comparison to existing employed models. In this paper, we benchmark models extensively on 19 public and 16 proprietary industrial data sets spanning a wide variety of chemical end points. In addition, we introduce a graph convolutional model that consistently matches or outperforms models using fixed molecular descriptors as well as previous graph neural architectures on both public and proprietary data sets. Our empirical findings indicate that while approaches based on these representations have yet to reach the level of experimental reproducibility, our proposed model nevertheless offers significant improvements over models currently used in industrial workflows.