Biology subjects

Kang, M.

Publications and source records attributed to Kang, M..

At least 37 records · Page 2Linked to original sources

bioRxiv2026

Microfluidic Agarose Microdroplets for DNA-Encoded Chemical Library Screening

DNA-encoded library (DEL) technology enables high-throughput small-molecule discovery but is typically performed using purified proteins under in vitro conditions that do not reflect native intracellular environments. Here, we present a microfluidic agarose -droplet platform for cellular-context DEL screening. The porous hydrogel droplets provide mechanically stable yet permeable microenvironments that protect weak protein-ligand interactions while enabling efficient buffer exchange and ligand diffusion. Importantly, mild cell permeabilization within droplets selectively retained chromatin-associated proteins, allowing screening directly in a cellular context. Using BRD4 as a model target, we validated intracellular ligand engagement by fluorescence imaging and super-resolution microscopy. Small-scale DEL screening selectively enriched JQ1 in both bead-based and cell-based formats, and large-scale DEL screening across millions of encoded compounds successfully identified hit molecules by sequencing. This agarose -droplet-based strategy expands DEL technology toward biologically relevant and chromatin-associated targets under near-native conditions.

bioengineering↗

bioRxiv2026

Genetic Modifiers of Cognitive Function: Novel Genetic Associations with Memory, Executive Function, and Attention in the Long Life Family Study

Cognitive performance is central to health and quality of life. Studying the factors that sustain performance may offer insights into maintaining cognitive function into advanced ages and may inform strategies to promote healthy cognitive aging. To identify single nucleotide polymorphisms (SNPs) underlying cognitive function, we performed genome-wide association studies (GWAS) of nine neuropsychological test scores capturing performance in three cognitive domains in 2,455 participants of the Long Life Family Study (LLFS). We identified 12 variants in seven tests and three domains that reached genome-wide significance (p < 5x10-). Three rare (minor allele frequency, European population MAF<0.01) protective, intronic variants, rs190287985 (CCSER1), rs75730801 (FHOD3), and rs552842447 (LINC00508), were uniquely associated with semantic fluency, phonemic fluency and number span forward, respectively. Two rare deleterious variants, rs556333682 and rs188304645, were associated with performance on the Hopkins Verbal Learning Test-Revised (HVLT-R) learning trials and lie in ischemia-related genes SH3TC1 and RPH3A. Another variant, rs180691759, associated with HVLT-R delayed recall, was proximal to RSPO3 and linked to decreased ECHDC1 expression, implicating ischemic and unexplained Ethylmalonic acid (EMA) pathways. We compared the results with GWASs of a general cognitive factor (GCF) and reaction time (RT) in the UK Biobank and meta-analysis results of the UK Biobank, CHARGE and COGENT by Davies et al. We found that rs535509651 associates with HVLT-R delayed recall in the LLFS and nominally associates (p<0.05) with GCF, and that rs10424537 associates with Immediate Logical Memory in the LLFS and nominally associates with RT. Furthermore, we identified 5 genome-wide significant loci in Davies et al. that reached loci adjusted significance in the LLFS (GCF p<0.05/128 and RT p<0.05/39). Each locus was associated with a single cognitive domain in the LLFS. We annotated the genome-wide significant results with quantitative trait loci (QTL) analyses of whole blood transcriptomic, serum metabolomic data, and gene set enrichment analyses (GSEA) using all nominally significant transcripts (p < 0.05/12). QTL analyses discovered 1 SNP-transcript (ECHDC1, p < 3 x 10-6), no SNP-lipid (p < 2 x 10-4), and no SNP-polar metabolite (p < 2 x 10-4) associations. Gene set enrichment analyses identified three pathways at FDR < 0.05. Our findings provide insight into the domain-specific genetic architecture of cognitive function.

genetics↗

bioRxiv2026

Knowledge-Guided Learning with Curated Prior Genetic Biomarkers for Robust Model Interpretation

MotivationKnowledge-guided learning offers effective and robust model training strategies in data-scarce settings by incorporating established domain knowledge, thereby enhancing generalization, robustness, and interpretability. By contrast, conventional deep learning approaches rely purely on data-driven learning, which can limit robust model interpretability, particularly in high-dimensional settings with limited size samples. In computational biology, knowledge-guided learning has primarily leveraged network- and structural-based knowledge, leading to biologically interpretable representations and enhanced predictive performance compared to conventional approaches. However, curated biomarkers, one of the most accessible forms of biological knowledge, remain largely unexplored within knowledge-guided paradigms. ResultsIn this study, we propose a model-agnostic training paradigm, Biomarker-driven Explainable Prior-guided Learning (BioExPL), that can be applied to any neural networks that incorporates curated prior knowledge. BioExPL enforces neural networks to reflect curated biomarker priors in their latent representations through a novel knowledge-alignment loss. BioExPL consistently demonstrated significantly improved predictive performance and enhanced model interpretability with minimized computational overhead in simulation studies and intensive experiments on multiple cancer datasts. BioExPL not only integrates prior curated knowledge into the model but also accurately identifies unknown associated signals additionally. BioExPL is model-agnostic and domain-independent, enabling its integration into diverse neural network architectures. Availability and implementationThe open-source is publicly available at: https://github.com/datax-lab/BioExPL.

bioinformatics↗

bioRxiv2026

Fairness-aware Supervised Hierarchical Contrastive Semantic Learning for Sexual Dimorphism Analysis

MotivationSexual dimorphism is a fundamental biological determinant driving systematic differences in disease susceptibility, progression, and clinical outcomes. However, current AI-based genomic models often exhibit algorithmic bias and fail to capture these sex-specific mechanisms, creating a critical barrier to unbiased precision medicine. Ensuring fairness in the context of sexual dimorphism requires understanding and addressing the distinct biological mechanisms functioning in each sex, rather than focusing solely on equalizing predictive performance. ResultsWe propose a fairness-aware supervised hierarchical contrastive learning approach, called FairHICON, to discover unbiased sex-common and sex-specific genomic drivers. Evaluations on cancer and asthma transcriptomic datasets demonstrate that FairHICON significantly outperforms state-of-the-art benchmarks, improving predictive performance by up to 9% while effectively reducing the performance gap between male and female cohorts. Furthermore, prognostic validation confirms that the identified sex-specific pathways stratify patient survival significantly better within their corresponding sex groups. This validates FairHICON to elucidate the molecular heterogeneity of sexual dimorphism, advancing inclusive precision medicine. Availability and implementationThe source code and data is available at https://github.com/datax-lab/FairHICON.

bioinformatics↗

bioRxiv2026

PIMO: Pathway-based Interpretable Multi-Omics interactions for multi-omics integration

MotivationModeling inter-omics interactions across multiple molecular levels is critical for deciphering the mechanisms underlying complex diseases. Epigenomic and structural alterations, such as DNA methylation and copy number alterations, modulate gene expression and collectively influence disease progression and patient survival outcomes. Despite advancements in deep learning-based multi-omics analysis, gene-level interactions of inter-omics have been seldom considered, due to combinational complexity and power, which limits interpretability and mechanistic insight. ResultsWe propose a Pathway-based Interpretable deep learning Multi-Omics interaction model, PIMO, that explicitly captures regulatory effects across omics layers. Experiments on multiple TCGA cancer datasets showed that PIMO consistently outperformed state-of-the-art baselines in survival analysis, up to 13% increase in the C-index. PIMO provides biologically interpretable analyses that identify important pathways, genes, and inter-omics interactions with DNA methylation and copy number alterations. Availability and implementationThe source code and data is available at https://github.com/datax-lab/PIMO.

bioinformatics↗

bioRxiv2025

A machine learning framework for predicting and modulating condition-dependent protein phase separation

Protein phase separation is a fundamental process in organizing membraneless organelles and is implicated in a wide range of pathological conditions. Importantly, rather than being a static feature of specific proteins, phase separation is a condition-dependent phenomenon governed by environmental parameters, including protein concentration, temperature, and solvent composition. However, most existing machine learning models infer phase-separation propensity solely from amino-acid sequences, failing to capture these context-dependent behaviors. Here, we present LLPSense, a machine learning framework that integrates pre-trained protein language model embeddings with environmental parameters to achieve accurate, condition-aware predictions of protein phase separation. We demonstrate LLPSenses predictive power and utility through three key experimental demonstrations. First, the model revealed that SGTA, previously unrecognized as a phase-separating protein, exhibits complex, temperature-dependent reentrant phase behavior. Second, LLPSense accurately predicted mutations in -synuclein that either enhance or suppress phase separation, enabling systematic mapping of residues potentially relevant to Parkinsons disease. Third, using model-guided mutagenesis, we inverted the phase behavior of UBQLN4, shifting it from high-temperature to low-temperature separation. Collectively, LLPSense provides a robust computational tool for interrogating the condition-dependent landscape of protein phase separation, enabling mechanistic studies of disease-associated phase separation and the rational design of programmable condensates.

bioinformatics↗

bioRxiv2025

ETV4 restrains the formation of CD74+ cytotoxic exhausted CD8+ T cells

Exhausted CD8+ T (Tex) cells comprise heterogeneous subsets with varying cytotoxic and dysfunctional properties, making inhibitory receptor profiles unreliable indicators of function. Therefore, it is essential to define transcriptional regulators or markers that accurately reflect these states. Here, we identified ETS variant 4 (ETV4) as a transcription factor that restricts cytotoxic Tex differentiation within tumors. ETV4 deficiency promotes the emergence of a CD74+ cytotoxic Tex subset while preserving the progenitor Tex (Tpex) pool. CD74 serves as both a marker and regulator, reinforcing Tpex features and enhancing Tex cytotoxicity. Mechanistically, ETV4 loss derepresses T-bet expression, thereby increasing IFN{gamma} production that induces CD74 expression in a paracrine manner. Single-cell analyses across human cancers revealed that CD74+ Tex cells are enriched for proliferative and metabolic programs and correlate with improved survival. Collectively, we establish the ETV4-CD74 axis as a regulator of Tex fate and promising therapeutic target to augment anti-tumor immunity.

immunology↗

bioRxiv2025

Chromatin is a long-range force generator that regulates plasma membrane tension and cell integrity independently of gene expression

Primarily studied for its role in gene expression, chromatin organization is emerging as an important regulator of nuclear mechanics. Although the nucleus is in mechanical equilibrium with the cell, we do not know whether and how chromatin reorganization actively regulates the mechanical properties and downstream behaviors of cells. Here, we tested the hypothesis that as a dynamic crosslinked polymer, chromatin directly impacts cell mechanics independently of transcription by studying NETosis: a transcription-independent process where chromatin decompacts and the plasma membrane (PM) ruptures. Using high resolution microscopy and ATAC-see, we found that chromatin accessibility progressively increases during NETosis suggesting that chromatin binding proteins (CBPs) dissociate from chromatin during NETosis. To determine the identity and dynamics of these dissociated CBPs, we used fluorescent recovery after photobleaching to measure the mobility and localization of the linker histone H1, the nucleosomal histone H3 and the heterochromatin binding protein HP1. We found that the mobile fraction of nuclear H1 increases during NETosis while fractions of HP1 and H3 diffuse outside of the nucleus suggesting that they become cytosolic osmolytes and potentially alter the mechanical state of cells. Consistently, we found that plasma membrane tension and cell volume increase as chromatin decompacts during NETosis. In non-NETing U2OS cells, we found that inducing chromatin decompaction increases plasma membrane tension, independently of the cytoskeleton, indicating a causal relationship between chromatin organization, cell volume and plasma membrane tension. Our findings reveal a novel non- genetic role of chromatin in cellular biophysics: regulating cell volume, PM tension, and thus, overall cell mechanics. Considering the critical role of cell mechanics in biological processes such as cell migration, proliferation and pathogen killing, our work broadens our understanding of how chromatin regulates cell physiology and pathology. Significance StatementChromatin organizes our DNA inside the nucleus and is important for gene expression. However, chromatin is also a polymer which can passively regulate the rigidity of the nucleus, but whether and how chromatin can actively regulate the mechanical properties of the whole cell remains unknown. Here, we leverage the immune process of NETosis to show that the organization of chromatin inside the nucleus actively regulates the volume and tension of cells. Our work establishes chromatin as a long-range force generator in cells, broadening our understanding of the roles of this crucial polymer network in cells and opening the door to new strategies for controlling the mechanical properties of cells as needed by their physiology.

cell biology↗

bioRxiv2025

Proteolytic cleavage of G3BP1 by calpain 1 couples NMDAR activation to mTOR-dependent local translation

Ribonucleoprotein (RNP) granules are dynamic, membraneless organelles that sequester translationally repressed mRNAs and RNA-binding proteins, playing a pivotal role in the regulation of localized protein synthesis. While disassembly of RNP granules is essential for reactivating translation, the mechanisms by which neuronal activity regulates this process remain poorly understood. In this study, we show that stimulation of N-methyl-D-aspartate (NMDA) receptor (NMDAR) triggers calcium influx, leading to activation of calpain 1 and subsequent proteolytic cleavage of Ras-GTPase-activating protein binding protein 1 (G3BP1), a core component of stress granules. This cleavage results in the disassembly of G3BP1 granules in the neurites and promotes mTOR-dependent local translation, thereby linking synaptic activity to spatially restricted protein synthesis. Finally, we demonstrate that the NMDAR-calpain 1-G3BP1-mTOR signaling axis contributes to axonal regeneration, establishing proteolytic remodeling of RNP granules as a key mechanism of activity-dependent neural repair, with potential implications for therapeutic intervention in brain injury.

neuroscience↗

bioRxiv2025

VEGFR2 blockade overcomes acquired KRAS G12D inhibitor resistance driven by PI3Kγ activation

KRAS G12D mutation is a key oncogenic driver in many solid tumors, including pancreatic, gastric, and colorectal cancers. While recent studies have characterized features associated with primary and acquired resistance to KRAS inhibitors, strategies to overcome such resistance, particularly in the context of gastrointestinal cancers, remain underexplored. Here, we have generated nine human gastrointestinal cancer models, including three patient-derived organoids (PDOs), with acquired resistance to the KRAS G12D-selective inhibitor MRTX1133. Using single-cell RNA sequencing analysis, we identified the enrichment of angiogenesis, hypoxia, and epithelial-to-mesenchymal transition (EMT) signatures in the resistant model compared to the parental PDO. Across all resistant models, VEGFA expression and VEGFR2 phosphorylation were uniformly elevated, which were driven by AKT activation and SP1 nuclear translocation. Mechanistic investigations uncovered increased PI3K{gamma} activity in MRTX1133-resistant models via complex formation of KRAS with p110{gamma} and p101. This leads to an autocrine VEGFA-VEGFR2 signaling loop formation and EMT induction. Therapeutically, the disruption of VEGFA-VEGFR2 signaling restored MRTX1133 sensitivity and inhibited EMT. Furthermore, cancer-endothelial paracrine signaling amplified angiogenesis, hypoxia, and EMT signatures in cancer cells and simultaneously promoted endothelial cell proliferation, reinforcing an adaptive feedback mechanism. In a mouse model of MRTX1133-resistant tumor xenograft, a combination of anti-VEGFR2 therapy and MRTX1133 more effectively reduced tumor growth, angiogenesis, and proliferation markers than monotherapy without significant body weight change. These findings establish VEGFA-VEGFR2 signaling by PI3K{gamma} activation as a key driver of acquired resistance to KRAS G12D inhibition and provide a rationale for combining VEGFA-VEGFR2 inhibition with KRAS blockade in KRAS-mutant cancers. HighlightO_LIVEGFA-VEGFR2 signaling activation is a common feature of MRTX1133 resistance in KRASG12D cancer cells C_LIO_LINuclear translocation of SP1 by AKT activation promotes VEGFA transcription in MRTX1133-resistant models C_LIO_LIInteraction of p110{gamma}-p101 with KRAS activates PI3K{gamma} in the resistant models C_LIO_LIVEGFA-VEGFR2 inhibition reverses MRTX1133 resistance in vitro and in vivo C_LI

cancer biology↗

bioRxiv2025

Prediction of bacterial protein-compound interactions with only positive samples

Prediction of Compound-Protein Interactions (CPI) in bacteria is crucial to advance various pharmaceutical and chemical engineering fields, including bio-catalysis, drug discovery, and industrial processing. However, current CPI models cannot be applied for bacterial CPI prediction due to the lack of curated negative interaction samples. This paper introduces a novel Positive-Unlabeled (PU) learning framework, named BIN-PU, to address this limitation. BIN-PU generates pseudo positive and negative labels from known positive interaction data, enabling effective training of deep learning models for CPI prediction. We also propose a weighted positive loss function that weights to truly positive samples. We have validated BIN-PU with multiple CPI backbone models, comparing the performance with the existing PU model using bacterial cytochrome P450 (CYP) data. Extensive experiments demonstrate the superiority of BIN-PU over the benchmark model in predicting CPIs with only truly positive samples. Furthermore, we have validated BIN-PU on additional bacterial proteins obtained from literature review, human CYP datasets, and uncurated data for its reproducibility. We have also validated the CPI prediction for the uncurated CYP data with biological and biophysical experiments. BIN-PU represents a significant advancement in CPI prediction for bacterial proteins, opening new possibilities for improving predictive models in related biological interaction tasks.

bioinformatics↗

bioRxiv2025

Plastic-hydrolytic enzyme classification using explainable deep learning

The rapid accumulation of plastic waste has emerged as a critical environmental threat, driving the need for scalable and effective biodegradation solutions. Hydrolytic plastic-degrading enzymes (PDEs) offer a promising solution, yet their functional classification remains limited by insufficient annotations and enzymatic diversity. In this study, we present an explainable deep learning framework, PEPIC, to classify nine types of PDEs directly from protein sequences. Using a curated dataset of experimentally validated enzymes and an expanded homologous dataset, we built an explainable deep learning model based on convolutional neural networks (PEPIC) for plastic-degrading enzyme prediction. We benchmarked PEPICs performance against state-of-the-art approaches. First, PEPIC demonstrated statistically significant improvements in predictive performance compared to state-of-the-art methods. Second, PEPIC calculates contribution scores for each amino acid in the protein sequence, indicating their influence on the predictions. The model interpretation revealed that regions highlighted by high contribution scores matched conserved catalytic triads and substrate-binding clefts across PET-, PCL-, and PLA-degrading enzymes. Furthermore, structural modeling confirmed the trustworthiness of PEPICs predictions. Finally, PEPIC predicted an uncurated enzyme as a PET-degrading enzyme, which was biologically validated to hydrolyze bis(2-hydroxyethyl) terephthalate (BHET). These findings demonstrate that PEPIC provides accurate and trustworthy predictions of PDEs, facilitating the discovery of novel enzymes and supporting the development of sustainable plastic biodegradation technologies.

bioinformatics↗

bioRxiv2025

Single-cell metabolome and RNA-seq multiplexing on single plant cells

Plants produce valuable natural products used for a wide variety of industrial applications. Since these molecules have important applications in a variety of industrial sectors, there is enormous interest in elucidating the biosynthetic pathways that are responsible for the production of these compounds. Identification of the genes that comprise these biosynthetic pathways has been enabled by gene-to-metabolite networks that are generated from transcriptomic and metabolomic datasets. Recent advances in both single-cell RNA-seq (scRNA-seq) and single-cell mass spectrometry metabolomics (scMS) have enabled the measurement of either gene expression or metabolite levels in individual cells. However, these individual datasets can only be used to indirectly correlate gene expression levels with metabolite concentrations at the single cell level. Here we demonstrate that both scRNA-seq and scMS can be applied to the same plant cell, thereby enabling direct comparisons between gene expression and metabolite levels. This multiplexing approach reveals both qualitative and quantitative correlations between metabolite levels and biosynthetic gene expression in individual cells. This integrated approach sheds light on the underlying processes driving complex plant biosynthesis.

plant biology↗

bioRxiv2025

Longitudinal deep phenotyping in a mouse model of West syndrome reveals temporal dynamics of synapse remodeling, gliosis, and proteomic and lipidomic changes during seizure evolution

Neurodevelopmental disorders can have long-lasting effects, causing not only early pediatric symptoms but also a range of neurological issues throughout adulthood. West syndrome is a severe neurodevelopmental disorder marked by infantile spasms, an early symptom that typically subsides with age. However, many patients progress to other seizure forms, known as seizure evolution, which is closely linked to poor long-term outcomes. Despite its clinical significance, the neurobiological mechanisms behind seizure evolution in West syndrome remain poorly understood. Recent genetic studies have consistently identified the CYFIP2 p.Arg87Cys variant in West syndrome patients, and the Cyfip2+/R87C mouse model carrying this mutation has been shown to recapitulate key symptoms of the disorder, including infantile spasms. In this study, we aimed to gain deeper insight into seizure evolution by conducting longitudinal deep phenotyping of the Cyfip2+/R87C mouse model from the neonatal stage to seven months of age. We tracked seizure activity through behavioral and EEG recordings and employed multi-omic analyses, including tissue and single-cell level transcriptomics, ultrastructural analysis, proteomics, and lipidomics, to capture a comprehensive view of molecular and cellular changes. Our results showed that after an initial period of neonatal spasms, Cyfip2+/R87C mice entered a seizure-free phase, followed by spontaneous recurrent seizures in adulthood, ultimately leading to premature death. This progression was associated with synaptic remodeling, sequential activation of different glial cell types, lipid droplet accumulation in astrocytes, and significant proteomic and lipidomic changes in the brain. These findings suggest that seizure evolution in West syndrome involves complex, time-dependent interactions between neurons and glial cells, along with alterations in lipid metabolism. Our study highlights the potential of longitudinal multi-omic approaches to uncover underlying mechanisms of seizure evolution and suggests that targeting these changes could offer novel therapeutic strategies. Additionally, the dataset generated here may provide valuable insights for other epilepsy and neurodevelopmental disorder models.

neuroscience↗

bioRxiv2025

Stochastic LASSO for extremely high-dimensional genomic data

Accurate identification of significant features in high-dimensional data is indispensable in high-throughput genomic analysis and association studies. Least Absolute Shrinkage and Selection Operator (LASSO) and its derivatives have been widely adapted to discover potential biomarkers as a feature selection scheme in various biological systems. Recently, bootstrap-based LASSO models, such as Random LASSO and Hi-LASSO, have been effective solutions for extremely high-dimensional but low sample size (EHDLSS) genomic data. However, the bootstrap-based LASSO models still have several drawbacks, such as multicollinearity within bootstrap samples, missing predictors in draw, and randomness in predictor sampling. To tackle the limitations, we propose a new bootstrap-based LASSO, named Stochastic LASSO, that effectively reduces multicollinearity in bootstrap samples and mitigates randomness in predictor sampling, resulting in remarkably outperforming benchmarks in feature selection and coefficient estimation. Furthermore, Stochastic LASSO provides a two-stage t-test strategy for selecting statistically significant features. The performance of Stochastic LASSO was assessed by comparing the existing benchmark models in extensive simulation experiments. In the simulation experiments, Stochastic LASSO consistently showed significant improvements in performance compared to the state-of-the-art LASSO models for feature selection, coefficient estimation, and robustness. We also applied Stochastic LASSO for the gene expression data of publicly available TCGA cancer datasets and identified statistically significant genes associated with survival month prediction. The source code is publicly available at: https://github.com/datax-lab/StochasticLASSO.

bioinformatics↗

bioRxiv2025

Personalized Brain Morphometric Feature as a Transdiagnostic Predictor of Psychopathology: Insights from Dual Systems Models

BackgroundAdolescents are particularly vulnerable to developing psychopathological symptoms, yet neurobiological markers that can identify these vulnerabilities in a personalized and interpretable manner remain limited. Dual Systems Models suggest that this vulnerability may result from asynchronous development of neural systems subserving cognitive-control and socioemotional functions. Given that rigorous empirical evidence is sparse, this study aimed to quantify developmental imbalances and evaluate their predictive value for psychopathology. MethodsBased on Dual Systems Models, the dorsolateral prefrontal cortex (DLPFC) and ventral striatum (VS) were selected as key regions for the two brain systems, respectively. Using longitudinal data from the Adolescent Brain Cognitive Development (ABCD) study (baseline: n=11,238, ages 9.92{+/-}0.625 years; 2-year follow-up: n=7,870, ages 12.2{+/-}0.652 years; 4-year follow-up: n=2972, ages 14.1{+/-}0.693 years), we derived a personalized imbalance score based on brain morphometric features, including surface area, thickness and gray-matter volume. Nested linear mixed models were employed to assess associations between the imbalance scores and psychopathology. Additionally, generalized additive models were used to capture developmental trajectories of the imbalance scores and explore potential non-linear associations with psychopathology. To examine reproducibility, data from the Lifespan Human Connectome Project in Development study (HCP-D, N=652, ages 8-21 years) were interrogated. ResultsThe imbalance score, quantified as the difference between VS volume and DLPFC surface area, exhibited strong reliability and validity. During adolescence, the score exhibited a declining trend, with decreasing growth rates and variability between individuals. The score was positively associated with externalizing symptoms and showed a U-shaped relationship with internalizing symptoms. These findings were replicated in the HCP-D sample. ConclusionsThe novel neuroanatomical imbalance score empirically supports Dual Systems Models and may function as a transdiagnostic marker for internalizing and externalizing psychopathology. These findings enhance understanding of the neurodevelopmental mechanisms underlying adolescent psychopathology and offer potential implications for precision prevention and intervention strategies.

developmental biology↗

bioRxiv2025

Fast in vivo deep-tissue 3D imaging with selective-illumination NIR-II light-field microscopy and aberration-corrected implicit neural representation

Near-infrared II (NIR-II) microscopy, which enables in vivo deep-tissue visualization of vasculature and cell activities, has been a promising tool for understanding physiological mechanisms. However, the volumetric image speed of the NIR-II microscopy is hindered by scanning strategy, causing limitations for observing instantaneous biological dynamics in 3D space. Here, we developed a NIR-II light-field microscopy (LFM) based on selective-illumination and self-supervised implicit neural representation (INR)-reconstruction, which allows ultra-fast 3D in vivo imaging (100 volumes/s). Through integrating INR with view-wise aberration correction, we could conquer the artifacts induced by the angular subsampling and refractive index variation, achieving single-cell resolution at a volume of 550 m diameter and 200 m thickness. The volumetric selective-illumination overcomes the influence of out-of-focus background, together with the low scattering advantage of NIR-II wavelength, extending the imaging depth to 600 m. The developed aberration-corrected implicit neural representation reconstruction (AIR) NIR-II LFM showcases its capability by monitoring hemodynamics of mouse brain under norepinephrine and flow redistribution of ischemic stroke in 3D vasoganglion, as well as noninvasively tracking immune cell activities inside subcutaneous solid tumor through intact skin. This approach represents a significant advancement in 3D in vivo imaging, holding great potential in biomedical research and preclinical studies.

neuroscience↗

bioRxiv2025

HIT-EC: Trustworthy prediction of enzyme commission numbers using a hierarchical interpretable transformer

Accurate and trustworthy prediction of Enzyme Commission (EC) numbers is critical for understanding enzyme functions and their roles in biological processes. Despite the success of recently proposed deep learning-based models, there remain limitations, such as low performance in underrepresented EC numbers, lack of learning strategy with incomplete annotations, and limited interpretability. To address these challenges, we propose a novel hierarchical interpretable transformer model, HIT-EC, for trustworthy EC number prediction. HIT-EC employs a four-level transformer architecture that aligns with the hierarchical structure of EC numbers, and leverages both local and global dependencies within protein sequences for this multi-label classification task. We also propose a novel learning strategy to handle incomplete EC numbers. HIT-EC, as an evidential deep learning model, produces trustworthy predictions by providing domain-specific evidence through a biologically meaningful interpretation scheme. The predictive performance of HIT-EC was assessed by multiple experiments: a cross-validation with a large dataset, a validation with external data, and a species-based performance evaluation. HIT-EC showed statistically significant improvement in predictive performance when compared to the current state-of-the-art benchmark models. HIT-ECs robust interpretability was further validated by identifying well-known conserved motifs and functional regions in the CYP106A2 enzyme family. HIT-EC would be a robust, interpretable, and reliable solution for EC number prediction, with significant implications for enzymology, drug discovery, and metabolic engineering. The open-source code is publicly available at: https://github.com/datax-lab/HIT-EC.

bioinformatics↗