Biology subjects

Chu, T.

Publications and source records attributed to Chu, T..

At least 37 records · Page 2Linked to original sources

bioRxiv2025

Unfolding the Black Box of Recurrent Neural Networks for Path Integration

Path integration is essential for spatial navigation. Experimental studies have identified neural correlates for path integration, but exactly how the neural system accomplishes this computation remains unresolved. Here, we adopt recurrent neural networks (RNNs) trained to perform a path integration task to explore this issue. After training, we borrow neuroscience prior knowledge and methods to unfold the black box of the trained model, including: clarifying neuron types based on their receptive fields, dissecting information flows between neuron groups by pruning their connections, and analyzing internal dynamics of neuron groups using the attractor framework. Intriguingly, we uncover a hierarchical information processing pathway embedded in the RNN model, along which velocity information of an agent is first forwarded to band cells, band and grid cells then coordinate to carry out path integration, and finally grid cells output the agent location. Inspired by the RNN-based study, we construct a neural circuit model, in which band cells form one-dimensional (1D) continuous attractor neural networks (CANNs) and serve as upstream neurons to support downstream grid cells to carry out path integration in the 2D space. Our study challenges the conventional view of considering grid cells as the principal velocity integrator, and supports a neural circuit model with the hierarchy of band and grid cells.

neuroscience↗

bioRxiv2025

Whole tissue spatial cellular analysis reveals increased macrophage infiltration in pancreata of autoantibody positive donors and patients with type 1 diabetes

While extensive efforts have characterized lymphoid populations that contribute to pancreatic insulitis in type 1 diabetes, significant gaps remain regarding key properties of macrophages in this inflammatory lesion. Hence, we quantified the spatial distribution of macrophages infiltrating human pancreatic tissue sections across disease states. Whole slide images of immunofluoresent stainings were analyzed with a semiautomated machine learning approach to study the distribution of macrophages throughout pancreatic sections from non-diabetic, auto-antibody positive (Aab+), and type 1 diabetes organ donors. Across the entire tissue section, macrophage infiltration was highest in donors with type 1 diabetes, and appeared contingent on the prescence of insulin-containing islets. Aab+ donors had higher macrophage densities in islet and peri-islet regions than non-diabetic individuals, but there was large case-by-case variation. Additionally, a greater proportion of macrophages highly expressed HLA class II (HLA-II) near and within islet regions of the type 1 diabetes group, which correlated with the proportion of stressed beta cells expressing HLA-II. The higher density of macrophages in the Aab+ and type 1 diabetes donors, taken together with higher expression of HLA-II in the type 1 diabetes pancreas, suggests that macrophages play a pivotal role in human type 1 diabetes initiation and progression. Article HighlightsO_LIMacrophage infiltration is increased in all pancreatic regions of patients with type 1 diabetes, and this infiltration appears contingent on the presence of insulin-containing islets. C_LIO_LIAab+ donors have higher macrophage densities in islet and peri-islet regions than non-diabetic individuals, but there is large case-by-case variation within the Aab+ group. C_LIO_LIThe proportion of macrophages highly expressing HLA-II is increased near and within islets only in patients with type 1 diabetes, and this correlates with the proportion of stressed beta cells expressing HLA-II within the islet. C_LIO_LIMacrophages are linked to human type 1 diabetes initiation and progression, hence forming a rationale to examine therapeutic strategies involving this cellular population. C_LI

pathology↗

bioRxiv2025

Localized Space Coding and Phase Coding Complement Each Other to Achieve Robust and Efficient Spatial Representation

Localized space coding and phase coding are two distinct strategies responsible, respectively, for representing abstract structure and sensory observations in neural cognitive maps. In spatial representation, localized space coding is implemented by place cells in the hippocampus (HPC), while phase coding is implemented by grid cells in the medial entorhinal cortex (MEC). Both strategies have their own advantages and disadvantages, and neither of them meets the requirement of representing space robustly and efficiently in the brain. Here, we show that through reciprocal connections between HPC and MEC, place and grid cells can complement each other to overcome their respective shortcomings. Specifically, we build a coupled network model, in which a continuous attractor neural network (CANN) with position coordinate models place cells, while multiple CANNs with phase coordinates model grid cell modules with varying spacings. The reciprocal connections between place and grid cells encode the correlation prior between the sensory cues processed by HPC and MEC, respectively. Using this model, we show that: 1) place and grid cells interact to integrate sensory cues in a Bayesian manner; 2) place cells complement grid cells in coding accuracy by eliminating non-local errors of the latter; 3) grid cells complement place cells in coding efficiency by enlarging the number of environmental maps stored stably by the latter. We demonstrate that the coupled network model explains the seemingly contradictory experimental findings about the remapping phenomena of place cells when grid cells are either inactivated or depolarized. This study gives us insight into understanding how the brain employs collaborative localized and phase coding to realize both robust and efficient information representation.

neuroscience↗

bioRxiv2025

spRefine Denoises and Imputes Spatial Transcriptomics with a Reference-Free Framework Powered by Genomic Language Model

The analysis of spatial transcriptomics is hindered by high noise levels and missing gene measurements, challenges that are further compounded by the higher cost of spatial data compared to traditional single-cell data. To overcome this challenge, we introduce spRefine, a deep learning framework that leverages genomic language models to jointly denoise and impute spatial transcriptomic data. Our results demonstrate that spRefine yields more robust celland spot-level representations after denoising and imputation, substantially improving data integration. In addition, spRefine serves as a strong framework for model pre-training and the discovery of novel biological signals, as highlighted by multiple downstream applications across datasets of varying scales. Notably, spRefine enhances the accuracy of spatial ageing clock estimations and uncovers new aging-related relationships associated with key biological processes, such as neuronal function loss, which offers new insights for analyzing ageing effect with spatial transcriptomics.

genomics↗

bioRxiv2025

The Chromosome 19 miRNA Cluster Guards Trophoblasts Against Overacting Innate Immunity

To maintain pregnancy health, the human placenta delicately balances protection of the developing fetus from invading pathogens with suppression of excessive inflammation that could lead to fetal and neonatal autoimmune disorders. Previous research, including our own, has shown that small RNA products of the Chromosome 19 MicroRNA Cluster (C19MC) promote viral resistance in non-trophoblastic cells. However, the role of C19MC products in placental trophoblasts remained unclear. Here, we analyzed chromatin accessibility in the C19MC enhancer and identified a previously unknown regulatory domain. Deletion of this domain silenced the expression of C19MC microRNA and Alu elements in trophoblasts. This silencing unexpectedly led to marked activation of cellular innate immune response and strikingly increased Toll-like receptor 3 (TLR3)-mediated sensitivity to poly(I:C), a viral RNA mimic. Our data suggest that C19MC non-coding RNAs interfere with endosomal TLR3 activation in trophoblasts, highlighting a previously unrecognized mechanism for hindrance of excessive innate immune activation.

developmental biology↗

bioRxiv2025

Accelerated Discovery of Cell Migration Regulators Using Label-Free Deep Learning-Based Automated Tracking

Cell migration plays a key role in normal developmental programs and in disease, including immune responses, tissue repair, and metastasis. Unlike other cell functions, such as proliferation which can be studied using high-throughput assays, cell migration requires more sophisticated instruments and analysis, which decreases throughput and has led to more limited mechanistic advances in our understanding of cell migration. Current assays either preclude single-cell level analysis, require tedious manual tracking, or use fluorescently labeled cells, which greatly limit the number of extracellular conditions and molecular manipulations that can be studied in a reasonable amount of time. Using the migration of cancer cells as a testbed, we established a workflow that images large numbers of cells in real time, using a 96-well plate format. We developed and validated a machine-vision and deep-learning analysis method, DeepBIT, to automatically detect and track the migration of individual cells from time-lapsed videos without cell labeling and user bias. We demonstrate that our assay can examine cancer cell motility behavior in many conditions, using different small-molecule inhibitors of known and potential regulators of migration, different extracellular conditions such as different contents in extracellular matrix and growth factors, and different CRISPR-mediated knockouts. About 1500 cells per well were tracked in 840 different conditions, for a total of ~1.3M tracked cells, in 70h (5 min per condition). Manual tracking of these cells by a trained user would take ~5.5 years. This demonstration reveals previously unidentified molecular regulators of cancer cell migration and suggests that collagen content can change the sign of how cytoskeletal molecules can regulate cell migration.

cell biology↗

bioRxiv2025

Tuning antibody stability and function by rational designs of framework mutations

Artificial intelligence and machine learning models have been developed to engineer antibodies for specific recognition of antigens, however these approaches often focus on the antibody complementarity determining region (CDR) whilst ignoring the immunoglobulin framework (FW) which provides structural rigidity and support for the flexible CDR loops. Here we present an integrated computational-experimental workflow, combining static structure analyses, molecular dynamics simulations and in vitro physicochemical and functional assays to generate rational designs of FW mutations for modulating antibody stability and activity. We first showed that recent antibody-specific language models lacked insights in FW mutagenesis, in comparison to approaches which utilised antibody structure information. Using the widely-used breast cancer therapeutic trastuzumab as a use case, we designed stabilising mutants which were distal to the CDR and preserved the antibodys functionality to engage its cognate antigen (HER2) and induce antibody-dependent cellular cytotoxicity (ADCC). Interestingly, guided by local backbone motions predicted using molecular dynamics simulations, we designed a FW mutation on the trastuzumab light chain which retained antigen-binding effects but lost Fab-mediated and Fc-mediated effector functions. This highlighted effects of FW on immunological functions engendered in distal areas of the antibody, and the importance to consider attributes other than binding affinity when assessing antibody function. Our approach incorporates interdomain dynamics and distal effects between FW and the Fc domains, expands the scope of antibody engineering beyond the CDR, and underscores the importance of a holistic perspective that considers the entire antibody structure as a whole in optimising antibody stability, developability and function.

immunology↗

bioRxiv2025

Transcriptomic plasticity is a hallmark of metastatic pancreatic cancer

Metastasis is the leading cause of cancer deaths; nonetheless, how tumor cells adapt to vastly different organ contexts is largely unknown. To investigate this question, we generated a transcriptomic atlas of primary tumor and diverse metastatic samples from a patient with pancreatic ductal adenocarcinoma who underwent rapid autopsy. Unsupervised archetype analysis identified both shared and site-specific gene programs, including lipid metabolism and gastrointestinal programs prevalent in peritoneum and stomach wall lesions, respectively. We developed a probabilistic approach for inferring clonal phylogeny from single-cell and matched whole-exome data. Distantly related genetic clones in the peritoneum express the lipid metabolism program, likely due to signaling by the adipocyte-rich peritoneum environment, and cells in most clones express multiple programs, suggesting that transcriptomic plasticity is a prevalent feature of metastatic cells. These deeply annotated analyses using a patient-centric platform provide a model for investigating metastatic mechanisms and plasticity in advanced cancer.

bioinformatics↗

bioRxiv2025

Lancet2: Improved and accelerated somatic variant calling with joint multi-sample local assembly graph

Here, we present Lancet2, an open-source somatic variant caller designed to improve detection of small variants in short-read sequencing data. Lancet2 introduces significant enhancements, including: 1) Improved variant discovery and genotyping through partial order multiple sequence alignment of assembled haplotype contigs and re-alignment of sample reads to the best supporting allele. 2) Optimized somatic variant scoring with Explainable Machine Learning models leading to better somatic filtering throughout the sensitivity scale. 3) Integration with Sequence Tube Map for enhanced visualization of variants with aligned sample reads in graph space. When benchmarked against enhanced two-tech truth sets generated using high-coverage short-read (Illumina) and long-read (Oxford Nanopore) data from four well characterized matched tumor/normal cell lines, Lancet2 outperformed other industry-leading tools in variant calling performance, especially for InDels. In addition, significant runtime performance improvements compared to Lancet1 ([~]10x speedup and 50% less peak memory usage) and most other state of the art somatic variant callers (at least 2x speedup with 8 cores or more) make Lancet2 an ideal tool for accurate and efficient somatic variant calling.

bioinformatics↗

bioRxiv2025

Identification of a multi-omics factor predictive of long COVID in the IMPACC study

Following SARS-CoV-2 infection, [~]10-35% of COVID-19 patients experience long COVID (LC), in which often debilitating symptoms persist for at least three months. Elucidating the biologic underpinnings of LC could identify therapeutic opportunities. We utilized machine learning methods on biologic analytes and patient reported outcome surveys provided over 12 months after hospital discharge from >500 hospitalized COVID-19 patients in the IMPACC cohort to identify a multi-omics "recovery factor". IMPACC participants who experienced LC had lower recovery factor scores compared to participants without LC. Biologic characterization revealed increased levels of plasma proteins associated with inflammation, elevated transcriptional signatures of heme metabolism, and decreased androgenic steroids in LC patients. The recovery factor was also associated with altered circulating immune cell frequencies. Notably, recovery factor scores were predictive of LC occurrence in patients as early as hospital admission, irrespective of acute disease severity. Thus, the recovery factor identifies patients at risk of LC early after SARS-CoV-2 infection and reveals LC biomarkers and potential treatment targets.

systems biology↗

bioRxiv2025

Balancing tumour proliferation and sustained cell cycle arrest through proteostasis remodelling drives immune niche compartmentalisation in breast cancer

Uncontrolled proliferation is a hallmark of cancer, yet tumour cells can enter G0 arrest by halting the cell cycle reversibly (quiescence) or irreversibly (senescence) to survive under stress and in hostile tumour microenvironments (TME). G0 arrested cells contribute to drug tolerance, metastasis and recurrence, but their identification remains challenging due to their rarity and elusive regulatory pathways. Here, we quantify G0 arrest and proliferation decisions in single-cell and spatially profiled breast primary tumours to unveil adaptive responses driving immune compartmentalisation. We identify a genomically-constrained and prolonged G0 arrest state resembling that of dormant precursors of cancer progression. This state featured adaptive transcriptional reprogramming, including unfolded protein response stress, reduced mitochondrial translation and epithelial-mesenchymal plasticity enabled by semaphorin signalling. Spatial transcriptomics revealed G0 arrest pockets encapsulated by APOE+ lipid-associated macrophages, myofibroblastic CAFs and immature perivascular cells, suggestive of an immunosuppressive niche contrasting with cytotoxic environments in proliferative hotspots and displaying distinct drug sensitivities. To facilitate future research, we provide a foundation model capturing G0 arrest and proliferation with 94% accuracy in single cell data, available at https://github.com/secrierlab/G0-LM. Our findings provide new insights into the spatial organisation of cell cycle arrest in breast cancer, highlighting the role of G0 states in tumour heterogeneity and adaptation.

cancer biology↗

bioRxiv2024

A systems model of alternating theta sweeps via firing rate adaptation

Place and grid cells provide a neural system for self-location and tend to fire in sequences within each cycle of the hippocampal theta rhythm when rodents run on a linear track. These sequences correspond to the decoded location of the animal sweeping forward from its current location ("theta sweeps"). However recent findings in open-field environments show alternating left-right theta sweeps, and propose a circuit for their generation. Here, we present a computational model of this circuit, comprising head direction cells, conjunctive grid x direction cells, and pure grid cells, based on continuous attractor dynamics, firing rate adaptation, and modulated by the medial-septal theta rhythm. Due to firing rate adaptation, the head-direction ring attractor exhibits left-right sweeps coding for internal direction, providing an input to the grid cell attractor network shifted along the internal direction, via an intermediate layer of conjunctive grid x direction cells, producing left-right sweeps of position by grid cells. Our model explains the empirical findings, including the alignment of internal position and direction sweeps and the dependence of sweep length on grid spacing. It makes predictions for thetamodulated head-direction cells, including specific relationships between theta phase precession during turning, theta skipping, anticipatory firing and directional tuning width. These predictions are verified in experimental data from anteroventral thalamus. The model also makes several predictions for the relationships between position and direction sweeps, running speed and dorsal-ventral location within the entorhinal cortex. Overall, a simple intrinsic mechanism explains the complex theta dynamics of the spatial circuit, with testable predictions.

neuroscience↗

bioRxiv2024

Dynamical Modulation of Hippocampal Replay Sequences through Firing Rate Adaptation

During periods of immobility and sleep, the hippocampus generates diverse self-sustaining sequences of "replay" activity, exhibiting stationary, diffusive, and super-diffusive dynamical patterns. However, the neural mechanisms underlying this diversity in hippocampal sequential dynamics remain largely unknown. Here, we propose such a mechanism demonstrating that modulation of firing rate adaptation in a continuous attractor model of place cells causes the emergence of different types of replay. Our model makes several key predictions. First, more diffusive replay sequences positively correlate with longer theta sequences across animals (both reflecting stronger adaptation). Second, replay diffusivity varies within an animal across behavioural states that affect adaptation (such as wake and sleep). Third, increases in neural excitability, incorporated with firing rate adaptation, reduce the step size of decoded movements within individual replay sequences. We provide new experimental evidence for all three predictions. These insights suggested that the diverse replay dynamics observed in the hippocampus can be reconciled through a simple yet effective neural mechanism, shedding light on its role in hippocampal-dependent cognitive functions and its relationship to other aspects of hippocampal electrophysiology.

neuroscience↗

bioRxiv2024

Placental Network Differences Among Obstetric Syndromes Identified With An Integrated Multiomics Approach

The placenta is essential for a healthy pregnancy, and placental pathology can endanger both maternal and fetal health. Placental function is affected by dynamic, complex, and interconnected molecular, cellular, and environmental events; therefore, we need a systems biology approach to study disease in normal physiological placenta function. We use placental multiomics (short and bulk transcriptomics, untargeted metabolomics, and targeted proteomics) paired with clinical data and placental histopathology reports from 321 placentas across multiple obstetric conditions: fetal growth restriction (FGR), FGR with pregnancy-related hypertension (FGR+HDP), preeclampsia (PE), and spontaneous preterm delivery (PTD). We first performed cellular deconvolution to estimate cell type numbers from bulk transcriptomes: FGR+HDP placentas were the most different from control placentas driven by a higher estimated number of extravillous trophoblast (p<0.0001). Next, we evaluated the impact of fetal sex and gestational age on analyte levels, adjusting for these confounders. We then generated obstetric condition-specific correlation networks and identified communities of related analytes associated with physiology and disease. We demonstrated how network connectivity and its disruption in disease can be used to identify signatures unique to a clinical outcome. We examined a community defined in control placentas for which the most connected node was miR-365a-3p in contrast to the corresponding community in FGR+HDP placentas for which the most connected node was hypoxia-induced miR-210-3p. From this community, we identified a signature containing mRNA transcripts implicated in placental dysfunction (e.g. FLT1, FSTL3, HTRA4, LEP, and PHYHIP). This signature distinguishes between FGR+HDP placentas and placentas of differing clinical outcomes in high-dimensional space. These findings illustrate the power of systems biology-driven interomics network analysis in a single tissue type, laying the groundwork for future multi-tissue studies.

systems biology↗

bioRxiv2024

Precision and Accuracy of Single-Cell/Nuclei RNA Sequencing Data

Single-cell and single-nucleus RNA sequencing (sc/snRNA-Seq) have become essential tools for profiling gene expression across different cell types in biomedical research. While factors like RNA integrity, cell count, and sequencing depth are known to influence data quality, quantitative benchmarks and actionable guidelines are lacking. This gap contributes to variability in study designs and inconsistencies in downstream analyses. In this study, we systematically evaluated quantitative precision and accuracy in expression measures across 23 sc/snRNA-Seq datasets comprising 3,682,576 cells from 339 samples. Precision was assessed using technical replicates based on pseudo-bulks created from subsampling. Accuracy was evaluated using sample-matched scRNA-Seq and pooled-cell RNA-Seq data of mononuclear phagocytes from four species. Our results show that precision and accuracy are generally low at the single-cell level, with reproducibility being strongly influenced by cell count and RNA quality. We establish data-driven thresholds for optimizing study design, recommending at least 500 cells per cell type per individual to achieve reliable quantification. Furthermore, we show that signal-to-noise ratio is a key metric for identifying reproducible differentially expressed genes. To support future research, we developed VICE (Variability In single-Cell gene Expressions), a tool that evaluates sc/snRNA-seq data quality and estimates the true positive rate of differential expression results based on sample size, observed noise levels, and expected effect size. These findings provide practical, evidence-based guidelines to enhance the reliability and reproducibility of sc/snRNA-seq studies.

genomics↗

bioRxiv2024

BAITSAO: Building A Foundation Model for Drug Synergy Analysis Powered by Language Models

Drug synergy prediction is a challenging and important task in the treatment of complex diseases including cancer. In this manuscript, we present a novel unified Model, known as BAITSAO, for tasks related to drug synergy prediction with a unified pipeline to handle different datasets. We construct the training datasets for BAITSAO based on the context-enriched embeddings from Large Language Models for the initial representation of drugs and cell lines. After demonstrating the relevance of these embeddings, we pre-train BAITSAO with a large-scale drug synergy database under a multi-task learning framework with rigorous selections of tasks. We demonstrate the superiority of the model architecture and the pre-trained strategies of BAITSAO over other methods through comprehensive benchmark analysis. Moreover, we investigate the sensitivity of BAITSAO and illustrate its unique functions including new drug discoveries, drug combinations-gene interaction, and multi-drug synergy predictions.

bioinformatics↗

bioRxiv2024

The neuroendocrine transition in prostate cancer is dynamic and dependent on ASCL1

Lineage plasticity is a recognized hallmark of cancer progression that can shape therapy outcomes. The underlying cellular and molecular mechanisms mediating lineage plasticity remain poorly understood. Here, we describe a versatile in vivo platform to identify and interrogate the molecular determinants of neuroendocrine lineage transformation at different stages of prostate cancer progression. Adenocarcinomas reliably develop following orthotopic transplantation of primary mouse prostate organoids acutely engineered with human-relevant driver alterations (e.g., Rb1-/-; Trp53-/-; cMyc+ or Pten-/-; Trp53-/-; cMyc+), but only those with Rb1 deletion progress to ASCL1+ neuroendocrine prostate cancer (NEPC), a highly aggressive, androgen receptor signaling inhibitor (ARSI)-resistant tumor. Importantly, we show this lineage transition requires a native in vivo microenvironment not replicated by conventional organoid culture. By integrating multiplexed immunofluorescence, spatial transcriptomics and PrismSpot to identify cell type-specific spatial gene modules, we reveal that ASCL1+ cells arise from KRT8+ luminal epithelial cells that progressively acquire transcriptional heterogeneity, producing large ASCL1+;KRT8- NEPC clusters. Ascl1 loss in established NEPC results in transient tumor regression followed by recurrence; however, Ascl1 deletion prior to transplantation completely abrogates lineage plasticity, yielding adenocarcinomas with elevated AR expression and marked sensitivity to castration. The dynamic feature of this model reveals the importance of timing of therapies focused on lineage plasticity and offers a platform for identification of additional lineage plasticity drivers.

cell biology↗

bioRxiv2024

Senolytic therapy preserves blood-brain barrier integrity and promotes microglia homeostasis in a tauopathy model

Cellular senescence, characterized by expressing the cell cycle inhibitory proteins, is evident in driving age-related diseases. Senescent cells play a crucial role in the initiation and progression of tau-mediated pathology, suggesting that targeting cell senescence offers a therapeutic potential for treating tauopathy associated diseases. This study focuses on identifying non-invasive biomarkers and validating their responses to a well-characterized senolytic therapy combining dasatinib and quercetin (D+Q), in a widely used tauopathy mouse model, PS19. We employed human-translatable MRI measures, including water extraction with phase-contrast arterial spin tagging (WEPCAST) MRI, T2 relaxation under spin tagging (TRUST), longitudinally assessed brain physiology and high-resolution structural MRI evaluated the brain regional volumes in PS19 mice. Our data reveal increased BBB permeability, decreased oxygen extraction fraction, and brain atrophy in 9-month-old PS19 mice compared to their littermate controls. (D+Q) treatment effectively preserves BBB integrity, rescues cerebral oxygen hypometabolism, attenuates brain atrophy, and alleviates tau hyperphosphorylation in PS19 mice. Mechanistically, D+Q treatment induces a shift of microglia from a disease-associated to a homeostatic state, reducing a senescence-like microglial phenotype marked by increased p16/INK4a. D+Q-treated PS19 mice exhibit enhanced cue-associated cognitive performance in the tracing fear conditioning test compared to the vehicle-treated littermates, implying improved cognitive function by D+Q treatment. Our results pave the way for application of senolytic treatment as well as these noninvasive MRI biomarkers in clinical trials in tauopathy associated neurological disorders.

neuroscience↗