Biology subjects

Levy, O.

Publications and source records attributed to Levy, O..

At least 19 recordsLinked to original sources

bioRxiv2026

The impacts of light and location on reproductive timing in the Orbicella species complex

Coral spawning events rely on tight synchronicity of gamete release to increase fertilization success by increasing gamete density in the water. Differences in spawning time can cause temporal reproductive isolation and minimize hybridization with closely related coral species. Sexual reproductive success is important as it generates the genetic diversity necessary for corals to survive large scale environmental stressors such as heat waves and bleaching events. The coordination of coral spawning events within a population maintains species boundaries and increases ecosystem-wide genetic diversity. Here we generate a database of spawning events in conjunction with depth and geographic location of the three species in the Orbicella species complex to assess variability in reproductive timing. The dataset, gathered from published papers, restoration practitioners, scientist throughout the Caribbean, contains almost three decades of data and includes 265 reproductive events. Our findings indicate that some aspects of each spawning season (month and day after the full moon) are influenced by latitude whereas others, such as spawning time in relation to local sunset, may exhibit variation based on geographic location. Combined, our results suggest that the reproductive timeframe of Orbicella spp. is triggered by factors associated with latitudinal gradients, such as the amount of energy from the sun, but the specific time of gamete release may be connected to other environmental factors that occur at a local scale or vary year-to-year.

ecology↗

bioRxiv2026

Bouncing back: Transient, but not long-lasting, effects of background disturbances on speech processing

Processing and understanding speech in face of occasional background disturbances can be difficult and often leads to distraction and reduced comprehension. However, despite the vast literature on the detrimental effects of processing speech-in-noise, little is known about the time-course of these effects. Here we studied whether reduced speech processing is limited only to the period when disruptions mask target speech, or if processing difficulties linger even after disruptions end, suggesting longer lasting distraction. Using an ecologically-relevant experimental design, we measured neural activity and skin-conductance from participants as they watched an educational video-lecture. Short background speech comments, mimicking plausible classroom disruptions, were occasional embedded in half of the lecture segments. Speech disruptions elicited robust neurophysiological orienting responses, reflected in a neural auditory ERP and a phasic increase in skin conductance. Importantly, when disruptions were presented, we observed a marked reduction in neural tracking of the lecture, increased inter-subject correlation (ISC), and decreased alpha power, suggesting a brief diversion of processing resources away from the target speech. However, these effects were transient in nature, with all neural measures returning to baseline immediately after disruption offset, showing no evidence for prolonged disruption-related costs or impaired comprehension. Interestingly, neurophysiological responses to disruption and return to baseline were not correlated with self-reported symptoms of attention-deficits (ADD). These findings highlight the resilience of auditory attention, demonstrating its ability to maintain a robust representations of target speech over time, despite of transient ecological disturbances.

neuroscience↗

bioRxiv2026

Testing the Cross-Paradigm Convergence of Behavioral and Neural Measures of Attention

Attention abilities, that are critical for most real-life cognitive task, can vary greatly across individuals. A plethora of behavioral and neural metric are commonly utilized to quantify attention; however, these often yield inconsistent and non-replicable results, raising questions as to which measures reliably account for and capture individual differences. To address this tension, here we employed a within-subject cross-paradigm design and quantified both behavioral and EEG-based neural measures associated with attention functioning, to test which measures converged across a gradient of artificial and ecological tasks. We found Inter-Subject Correlation (ISC), which quantifies the similarity in the pattern of neural response across individuals, captured consistent individual differences across both an artificial (Auditory Oddball) and ecological (Attention to Speech) task, with some correlation with performance. Moreover, we found that P300 neural responses to surprising events were qualitatively similar across artificial and ecological tasks, supporting their generalization across contexts. Interestingly, traditional cognitive-behavioral measures of attention - both from speeded response-time tasks and self-report of ADHD symptoms (ASRS) were not correlated with each other nor did they explain variance in neural metrics, nor were they explained by variation in general cognitive abilities (working memory or fluid intelligence). While these results demonstrate the shortcomings of many established measures of attention to generalize across contexts, they also point to the potential of neural measures, and specifically ISC, to serve as reliable indicators of individual differences and to bridge the gap between artificial and ecological studies of attention functioning.

neuroscience↗

bioRxiv2026

Improved coral thermal tolerance through modulation of antioxidant defenses

Mass coral bleaching events, driven by rising ocean temperatures, are pushing reef ecosystems toward collapse on a global scale. Because oxidative stress is an early driver of coral bleaching, strategies that enhance coral antioxidant defenses may improve coral resilience under thermal stress. Here, we tested a targeted antioxidant supplementation designed to enhance oxidative stress regulation in three representative Red Sea scleractinian coral species subjected to a thermal challenge. While responses varied among species and physiological metrics, supplemented corals consistently maintained higher photophysiological performance under heat stress. In Stylophora pistillata, antioxidant supplementation was associated with enhanced catalase activity, maintenance of glutathione redox homeostasis, and lower intracellular reactive oxygen species levels. In contrast, non-fed corals exhibited oxidative imbalance, increased lipid peroxidation, and impaired photophysiological recovery, while corals receiving a non-enriched heterotrophic diet showed an intermediate response characterized by increased catalase activity but persistent glutathione oxidation and elevated ROS during recovery. Together, the dietary treatments revealed a gradient in oxidative regulation, ranging from insufficient antioxidant protection in autotrophic corals to enhanced oxidative homeostasis in antioxidant-supplemented corals. Our findings demonstrate the potential of targeted nutritional antioxidant supplementation to enhance coral oxidative regulation and physiological performance under elevated temperatures, highlighting a promising complementary approach for coral conservation and restoration efforts.

physiology↗

bioRxiv2026

The aging genome exhibits organized vulnerability to somatic mutations

Somatic mutations accumulate throughout life and have been hypothesized to drive organismal decline. Yet whether these mutations are distributed randomly or whether cells shield their most critical components has remained unresolved. Here we analyze over a million somatic mutations across thirteen human tis-sues, finding that the aging genome exhibits organized vulnerability, captured by the existence of hypo-mutated genes and longevity-associated pathways that have significantly lower mutation burden. Highly connected network hubs are systematically protected from mutation, while peripheral, condition-specific genes accumulate disproportionate burdens. We show that this organized vulnerability arises from the interplay of two independent mechanisms: transcription-coupled repair, and selective filtering. Finally, we validate our findings under experimental mutagenesis, demonstrating intrinsic mechanisms of protection rather than tissue-specific confounders. These findings reframe the somatic mutation hypothesis: organismal decline may not reflect total mutational burden, but where those mutations fall within the cellular network.

genomics↗

bioRxiv2026

Artificial light at night reshapes diel brain transcriptomics in Dascyllus aruanus damselfish

Artificial light at night (ALAN) disrupts natural light cycles and interferes with light-dependent biological processes. However, the effect of ALAN on cellular processes in wildlife is unclear. We examined diel brain transcriptomic alterations in the diurnal damselfish Dascyllus aruanus by comparing fish exposed to three consecutive nights of ALAN with control fish, sampled during both the day and night. ALAN partially disrupted circadian regulation transcription, altering diel expression of the core clock regulator bmal1 and glucocorticoid-regulated genes. At night, ALAN triggered activation of genes indicative of neuronal activity and acute neural stress, along with suppression of restorative nocturnal processes. The following day, the transcriptomic divergence between ALAN-exposed and control fish expanded, with widespread downregulation of genes governing vascular homeostasis, coagulation, and immune function. Together, these findings indicate that ALAN reshapes brain transcriptomic programs across the entire diel cycle, identifying molecular signatures of physiological disruption in light-polluted marine environments.

cell biology↗

bioRxiv2026

Molecular insights into protein disulfide isomerase antagonism by punicalagin

Punicalagin, an ellagic acid polyphenol from pomegranate, has been proposed as an antagonist of protein disulfide isomerase (PDI) and endoplasmic reticulum resident protein 57 (ERp57), thiol oxidoreductases that regulate protein folding and extracellular thrombotic signaling. Here, biochemical oxidase and reductase assays on PDI show that punicalagin inhibits both activities with micromolar potency, thereby extending earlier work that described only disulfide reductase inhibition. In parallel, thiol labeling of catalytic cysteines revealed no change in the redox state, supporting a noncovalent, allosteric of inhibition. Molecular docking and molecular dynamics simulations showed that punicalagin binds stably and preferentially to defined sites on the Nterminal domains of PDI through extensive hydrogen bonding and van der Waals contacts, which is an alternative binding mode to previously reported C-terminal binding. Finally, artificial intelligence-driven network analysis identified PDI as a high-confidence target of punicalagin and related galloylated polyphenols, alongside additional signaling proteins. Together, these findings provide further mechanistic framework for punicalagin-mediated antagonism of PDI and highlight galloylated polyphenols as promising scaffolds for protein disulfide isomerase-targeted therapeutics. HighlightsO_LIPunicalagin, a galloylated polyphenol, antagonizes not only the reductase activity but also the oxidase activity of protein disulfide isomerase C_LIO_LIProtein disulfide isomerase inhibition by punicalagin is through N-terminal binding C_LIO_LIPunicalagin inhibits conformationally rather than catalytic cysteine modification C_LIO_LIArtificial intelligence network analysis reveals pathway inhibition by punicalagin C_LI

biochemistry↗

bioRxiv2025

From Big Data to Small Scales: Machine Learning Enhances Microclimate Model Predictions

1. Microclimates are critical for understanding how organisms interact with their environments, influencing behaviour, physiology, and species distributions. However, traditional physical heat-balance models for predicting ground temperatures in microhabitats often exhibit biases due to unaccounted environmental complexities and poorly constrained parameters. These limitations can hinder ecological research and conservation planning, particularly in the context of climate change. 2. In this study, we demonstrate how high-resolution drone-based mapping and machine learning can improve the accuracy of microclimate models. Using drone imagery, we generated detailed environmental maps, including solar radiation, vegetation indices, and skyview factors, to parameterize a physical heat-balance model. Validation with thermal maps derived from drone-mounted infrared cameras revealed systematic errors in the physical models predictions, including over- and underestimations under specific environmental conditions. To address these errors, we applied a random forest machine learning model to predict and correct biases in new prediction maps. 3. Our results show that machine learning reduced mean absolute errors by over 30% and mean square errors by 50%, while consistently narrowing the range of prediction inaccuracies. Key factors driving biases, such as vegetation cover, solar radiation, and height above ground, were identified, offering valuable insights for improving physical models. The machine learning corrections not only improved accuracy but also highlighted parameters and processes that were previously underrepresented or oversimplified in traditional models. 4. These findings illustrate the potential of combining machine learning with physical modelling to enhance microclimate predictions. This approach provides ecologists and conservation practitioners with a powerful tool to generate accurate, fine-scale microclimate maps, enabling better understanding of species responses to climate change and informing climate-resilient habitat management and conservation strategies.

ecology↗

bioRxiv2025

Human iPSC derived alveolar macrophages reveal macrophage subtype specific functions of itaconate in M. tuberculosis host defense

Mycobacterium tuberculosis (Mtb) must survive within multiple macrophage populations during infection, including alveolar macrophages (AM) and recruited inflammatory macrophages. In mice, itaconate, produced in macrophages by ACOD1 mediated decarboxylation of aconitate, has direct antimicrobial activity, modulates inflammatory cytokines, and is required for resistance to M. tuberculosis (Mtb) infection. The role of itaconate in human macrophages is less clear and whether itaconate mediates distinct effects in macrophage subtypes is unknown. Here, we investigated the role of itaconate in human iPSC-derived macrophages, either induced by GM-CSF to resemble alveolar macrophages (AM-Like cells), or treated with M-CSF to generate control macrophages (MCDM cells). Both types of human macrophages produce substantially less itaconate than mouse macrophages and AM-Ls produced 4-fold less itaconate than MCDMs. Surprisingly, ACOD1 deficient AM-L macrophages, but not MCDM macrophages, were permissive for Mtb growth. Moreover, itaconate functioned to dampen the Mtb induced inflammatory response in MCDMs, but not AM-L macrophages, affecting both the Type I IFN and TNF pathways. These results indicate that itaconate is involved in human macrophage responses to TB, with distinct roles in different macrophage subsets. These results also show that genetically tractable hiPSC-derived macrophages are a robust and versatile model to dissect cellular host pathogen interactions.

immunology↗

bioRxiv2024

A Framework for Autonomous AI-Driven Drug Discovery

The exponential increase in biomedical data offers unprecedented opportunities for drug discovery, yet overwhelms traditional data analysis methods, limiting the pace of new drug development. Here we introduce a framework for autonomous artificial intelligence (AI)-driven drug discovery that integrates knowledge graphs with large language models (LLMs). It is capable of planning and carrying out automated drug discovery programs at a massive scale while providing details of its research strategy, progress, and all supporting data. At the heart of this framework lies the focal graph - a novel construct that harnesses centrality algorithms to distill vast, noisy datasets into concise, transparent, data-driven hypotheses. We demonstrate that even small-scale applications of this highly scalable approach can yield novel, transparent insights relevant to multiple stages of the drug discovery process, including chemical structure-based target prediction, and present the implementation of a system which autonomously plans and executes a multi-step target discovery workflow. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=65 SRC="FIGDIR/small/629024v3_ufig1.gif" ALT="Figure 1"> View larger version (18K): org.highwire.dtl.DTLVardef@9d7d7dorg.highwire.dtl.DTLVardef@199d768org.highwire.dtl.DTLVardef@10d0335org.highwire.dtl.DTLVardef@14d8d1c_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2024

Network Analysis Reveals Protein Modules Associated with Childhood Respiratory Diseases

BackgroundThe first year of life is a period of rapid immune development that can impact health trajectories and the risk of developing respiratory-related diseases, such as asthma, recurrent infections, and eczema. However, the biology underlying subsequent disease development remains unknown. MethodsUsing weighted gene correlation network analysis (WGCNA), we derived modules of highly correlated immune-related proteins in plasma samples from children at age 1 year (N=294) from the Vitamin D Antenatal Asthma Reduction Trial (VDAART). We applied regression analyses to assess relationships between protein modules and development of childhood respiratory diseases up to age 6 years. We then characterized genomic, environmental, and metabolomic factors associated with modules. ResultsWGCNA identified four protein modules at age 1 year associated with incidence of childhood asthma and/or recurrent wheeze (Padj range: 0.02-0.03), respiratory infections (Padj range: 6.3x10-9-2.9x10-6), and eczema (Padj=0.01) by age 6 years; three modules were associated with at least one environmental exposure (Padj range: 2.8x10-10-0.03) and disrupted metabolomic pathway(s) (Padj range: 2.8x10-6-0.04). No genome-wide SNPs were identified as significant genetic risk factors for any protein module. Relationships between protein modules with clinical, environmental, and omic factors were temporally sensitive and could not be recapitulated in protein profiles at age 6 years. ConclusionThese findings suggested protein profiles as early as age 1 year predicted development of respiratory-related diseases through age 6 and were associated with changes in pathways related to amino acid and energy metabolism. These may inform new strategies to identify vulnerable individuals based on immune protein profiling.

bioinformatics↗

bioRxiv2024

Neurophysiological Patterns of Attention and Distraction during Realistic Virtual-Reality Classroom Learning in Adults with and without ADHD

Many people, and particularly individuals with Attention Deficit (Hyperactivity) Disorder (AD(H)D), find it difficult to maintain attention during classroom learning. However, traditional paradigms used to evaluate attention do not capture the complexity and dynamic nature of real-life classrooms. Using a novel Virtual Reality platform, coupled with measurement of neural activity, eye-gaze and skin conductance, here we studied the neurophysiological manifestations of attention and distractibility, under realistic learning conditions. Individuals with AD(H)D exhibited higher neural responses to irrelevant sounds and reduced speech tracking of the teacher, relative to controls. Additional neurophysiological measures, such the power of alpha-oscillations and frequency of gaze-shifts away from the teacher, contributed to explaining variance in self-reported AD(H)D symptoms across the sample. These ecologically-valid findings provide critical insight into the neurophysiological mechanisms underlying individual differences in the capacity for sustained attention and the proneness to distraction and mind-wandering, experienced in real-life situations.

neuroscience↗

bioRxiv2024

An ecological investigation of the effect of background noise on speech processing in a Virtual Classroom

Many real-life situations can be extremely noisy. Psychoacoustic studies have shown that background noise can have a detrimental effect on the ability to process and understand speech. However, most studies use stimuli and task designs that are highly artificial, limiting their generalization to more realistic contexts. Moreover, to date, we do not fully understand the neurophysiological consequences of trying to pay attention to speech in a noisy place. To address this lab to real-life gap and increase the ecological validity of speech in noise research, here we introduce a novel audiovisual Virtual Reality (VR) experimental platform. Combined with neurophysiological measurements of neural activity (EEG), eye-gaze and skin conductance (GSR) we studied the effects of background noise in a realistic context where the ability to process and understand continuous speech is especially important: A VR Classroom. Participants (n=32) sat in a VR Classroom and were told to pay attention to mini-lecture segments by a virtual teacher. Trials were either Quiet or contained background construction noise, emitted from outside the classroom window, which was either Continuous (drilling) or Intermittent (air hammers). Result show that background noise had a detrimental effect on learning outcomes, which was also accompanied by reduced neural tracking of the teachers speech. Comparison of the two noise types showed that the intermittent construction noise was more disruptive than continuous noise, as index by both behavioral and neural measures, and it also elicited higher skin-conductance levels, reflecting heightened arousal. Interesting, eye-gaze dynamics were not affected by the presence of noise. This study advances our understanding of the neurophysiological effects of background noise and extends it to more ecologically relevant contexts. It also emphasizes the role that temporal dynamics play for processing speech in noise, highlighting the need to consider the features of realistic noises, as we expand speech in noise research to increasingly realistic circumstances.

neuroscience↗

bioRxiv2023

Integrated longitudinal multi-omics study identifies immune programs associated with COVID-19 severity and mortality in 1152 hospitalized participants

Hospitalized COVID-19 patients exhibit diverse clinical outcomes, with some individuals diverging over time even though their initial disease severity appears similar. A systematic evaluation of molecular and cellular profiles over the full disease course can link immune programs and their coordination with progression heterogeneity. In this study, we carried out deep immunophenotyping and conducted longitudinal multi-omics modeling integrating ten distinct assays on a total of 1,152 IMPACC participants and identified several immune cascades that were significant drivers of differential clinical outcomes. Increasing disease severity was driven by a temporal pattern that began with the early upregulation of immunosuppressive metabolites and then elevated levels of inflammatory cytokines, signatures of coagulation, NETosis, and T-cell functional dysregulation. A second immune cascade, predictive of 28-day mortality among critically ill patients, was characterized by reduced total plasma immunoglobulins and B cells, as well as dysregulated IFN responsiveness. We demonstrated that the balance disruption between IFN-stimulated genes and IFN inhibitors is a crucial biomarker of COVID-19 mortality, potentially contributing to the failure of viral clearance in patients with fatal illness. Our longitudinal multi-omics profiling study revealed novel temporal coordination across diverse omics that potentially explain disease progression, providing insights that inform the targeted development of therapies for hospitalized COVID-19 patients, especially those critically ill.

immunology↗

bioRxiv2023

Early cellular and molecular signatures correlate with severity of West Nile Virus infection

Infection with West Nile Virus (WNV) can drive a wide range of responses, from asymptomatic to flu-like symptoms/fever or severe cases of encephalitis and death. To identify cellular and molecular signatures distinguishing WNV severity, we employed systems profiling of peripheral blood from asymptomatic and severely ill individuals infected with WNV. We interrogated immune responses longitudinally from acute infection through convalescence at 3 months and 1 year employing multiplexed single cell protein and transcriptional profiling (CyTOF and Seq-Well) complemented with matched serum proteomics and metabolomics. At the acute time point, we detected both an elevated proportion of pro-inflammatory markers in innate immune cell types and reduced frequency of regulatory T cell activity in participants with severe infection compared to those with asymptomatic infection. Single-cell transcriptomics of paired samples revealed that asymptomatic donors had higher expression of genes associated with innate immune pathways, in particular anti-inflammatory CD16+ monocytes at the acute time point. A multi-omics analysis identified factors--beyond those from individual analyses--that distinguished immune state trajectory between severity groups. Here we highlighted the potential of systems immunology using multiple cell-type and cell-state-specific analyses to identify correlates of infection severity and host cellular activity contributing to an effective anti-viral response.

immunology↗

bioRxiv2023

Uncovering the function of the evolutionary conserved Clock gene in the sea anemone Nematostella vectensis

The circadian clock enables anticipation of the day/night cycle in animals ranging from cnidarians to mammals. Circadian rhythms are generated through a transcription-translation feedback loop (TTFL or pacemaker) with CLOCK as a conserved positive factor in animals. However, CLOCKs functional evolutionary origin and mechanism of action in basal animals are unknown. In the cnidarian Nematostella vectensis, pacemaker gene transcript levels, including NvClk (the Clock ortholog), appear arrhythmic under constant darkness, questioning the role of NvCLK. Utilizing CRISPR/Cas9, we generated a NvClk allele mutant (NvClk{Delta}), revealing circadian behavior loss in constant light conditions (LL and DD) while a 24-hour rhythm was maintained under light-dark condition (LD). Transcriptomics showed distinct rhythmic genes in wild-type (WT) genes in LD compared to DD. The LD NvClk{Delta}/{Delta} showed comparable numbers of rhythmic genes, but were greatly reduced in DD. Furthermore, the LD NvClk{Delta}/{Delta} showed alterations in temporal pacemaker gene expression, affecting their potential interactions. Additionally, differential expression of non-rhythmic genes associated with cell division and neuronal differentiation was observed. These findings revealed that a light-responsive pathway can partially compensate for circadian clock disruption and that the Clock gene has evolved in cnidarians to synchronize rhythmic physiology and behavior to the diel rhythm of the earths biosphere.

evolutionary biology↗

bioRxiv2022

Reduced SARS-CoV-2 mRNA vaccine immunogenicity and protection in mice with diet-induced obesity and insulin resistance.

BackgroundObesity and Type 2 Diabetes Mellitus (T2DM) are associated with an increased risk of severe outcomes from infectious diseases, including COVID-19. These conditions are also associated with distinct responses to immunization, including an impaired response to widely used SARS-CoV-2 mRNA vaccines. ObjectiveTo establish a connection between reduced immunization efficacy via modeling the effects of metabolic diseases on vaccine immunogenicity that is essential for the development of more effective vaccines for this distinct vulnerable population. MethodsWe utilized a murine model of diet-induced obesity and insulin resistance to model the effects of comorbid T2DM and obesity on vaccine immunogenicity and protection. ResultsMice fed a high-fat diet (HFD) developed obesity, hyperinsulinemia, and glucose intolerance. Relative to mice fed a normal diet (ND), HFD mice vaccinated with a SARS-CoV-2 mRNA vaccine exhibited significantly lower anti-spike IgG titers, predominantly in the IgG2c subclass, associated with a lower type 1 response, along with a 3.83-fold decrease in neutralizing titers. Furthermore, enhanced vaccine-induced spike-specific CD8+ T cell activation and protection from lung infection against SARS-CoV-2 challenge were seen only in ND mice but not in HFD mice. ConclusionWe demonstrate impaired immunity following SARS-CoV-2 mRNA immunization in a murine model of comorbid T2DM and obesity, supporting the need for further research into the basis for impaired anti-SARS-CoV-2 immunity in T2DM and investigation of novel approaches to enhance vaccine immunogenicity among those with metabolic diseases. Capsule summaryObesity and type 2 diabetes impair SARS-CoV-2 mRNA vaccine efficacy in a murine model.

immunology↗

bioRxiv2022

Correspondence between the layered structure of deep language models and temporal structure of natural language processing in the human brain

Deep language models (DLMs) provide a novel computational paradigm for how the brain processes natural language. Unlike symbolic, rule-based models described in psycholinguistics, DLMs encode words and their context as continuous numerical vectors. These "embeddings" are constructed by a sequence of computations organized in "layers" to ultimately capture surprisingly sophisticated representations of linguistic structures. How does this layered hierarchy map onto the human brain during natural language comprehension? In this study, we used electrocorticography (ECoG) to record neural activity in language areas along the superior temporal gyrus and inferior frontal gyrus while human participants listened to a 30-minute spoken narrative. We supplied this same narrative to a high-performing DLM (GPT2-XL) and extracted the contextual embeddings for each word in the story across all 48 layers of the model. We next trained a set of linear encoding models to predict the temporally-evolving neural activity from the embeddings at each layer. We found a striking correspondence between the layer-by-layer sequence of embeddings from GPT2-XL and the temporal sequence of neural activity in language areas. In addition, we found evidence for the gradual accumulation of recurrent information along the linguistic processing hierarchy. However, we also noticed additional neural processes in the brain, but not in DLMs, during the processing of surprising (unpredictable) words. These findings point to a connection between human language processing and DLMs where the layer-by-layer accumulation of contextual information in DLM embeddings matches the temporal dynamics of neural activity in high-order language areas.

neuroscience↗