Biology subjects

Lee, H.

Publications and source records attributed to Lee, H..

At least 163 records · Page 9Linked to original sources

bioRxiv2025

Implementing N-terminomics and machine learning to probe in vivo Nt-arginylation

N-terminal arginylation (Nt-arginylation) serves as a protein degradation signal in both the ubiquitin-proteasome system and the autophagy-lysosomal pathway. However, the scarcity of arginylated proteins in cells and limitations in current identification methods have hindered progress in this field. In this study, we developed a novel integrated approach that combines N-terminomics (N-terminomic enrichment, LC-MS/MS, and tandem database search) with advanced machine learning-based filtering strategies to successfully identify in vivo Nt-arginylation with unprecedented sensitivity and specificity. By utilizing Arg-starting peptides from missed cleavage products as physicochemical proxies for ATE1-mediated Nt-arginylation, we trained a transfer learning-based model to predict MS2 spectra and retention times of candidate peptides. Additionally, near-isobaric modifications were filtered by statistically analyzing mass error deviations in MS2 fragment ions. Using this approach, we identified 134 Nt-arginylation sites in thapsigargin-treated HeLa cells, revealing a significant increase of Nt-arginylome under unfolded protein response (UPR) stress. Arginylated proteins originate from various organelles, including ER, nucleus and mitochondria. Notably, arginylation frequently occurred at sites processed by caspases or where signal peptides had been cleaved. Several proteins identified in our arginylome study were validated for their interaction with the R-catcher, an Nt-Arginylation bait derived from the p62 ZZ domain. Temporal profiling of N-terminal arginylation sites post-stress induction via targeted proteomics revealed a sequential response pattern. The UPR signal transducer ATF4 exhibited the most rapid increase, followed by N-terminal arginylation of caspase-3 substrates, and subsequently by arginylation at signal peptide cleavage sites among endoplasmic reticulum proteins. Our novel machine learning-based filtering methodologies enable the discovery of rare post-translational modifications by implementing specialized filtering strategies tailored to the unique physicochemical properties of terminal modifications, with potential applications in biomarker discovery, drug target identification, and elucidation of disease-specific protein regulation mechanism.

systems biology↗

bioRxiv2025

Structured interactions explain the absence of keystone species in synthetic microcosms

In complex ecosystems, the loss of certain species can trigger a cascade of secondary extinctions and invasions. However, our understanding of the prevalence of these critical "keystone" species and the factors influencing their emergence remains limited. To address these questions, we experimentally assembled microcosms from 16 marine bacterial species and found that multiple extinctions and invasions were exceedingly rare upon removal of a species from the initial inoculation. This was true across eight different environments with either simple carbon sources (e.g., glucose) and more complex ones (e.g. glycogen). By employing a generalized Lotka-Volterra model, we could reproduce these results when interspecies interactions followed a hierarchical pattern, wherein species impacted strongly by one species were also more likely to experience strong impacts from others. Such a pattern naturally emerges due to observed variation in carrying capacities and growth rates. Furthermore, using both statistical inference and spent media experiments, we inferred interspecies interaction strengths and found them consistent with structured interactions. Our results suggest that the natural emergence of structured interactions may provide community resilience to extinctions.

systems biology↗

bioRxiv2025

Linoleic acid metabolite, 13-S-hydroxyoctadecadienoic acid, suppresses cancer cell growth by inhibiting mTOR

Mechanistic target of rapamycin (mTOR) is a key protein kinase that integrates various internal and external signals to control biological events including cell growth. Whereas substantial efforts were made to elucidate protein subunits interacting with mTOR, endogenous metabolite-mTOR interactions remain largely unknown. Using affinity protein purification and mass spectrometry, we identified direct binding of mTOR to 13-S-hydroxyoctadecadienoic acid (13-S-HODE) which is an oxygenated metabolite of linoleic acid, a polyunsaturated essential fatty acid. Interaction of 13-S-HODE with the catalytic ATP-binding domain of mTOR prevented its kinase activity in an ATP-competitive manner. Furthermore, either 13-S-HODE treatment or expression of arachidonate 15-lipoxygenase (ALOX15), an enzyme responsible for 13-S-HODE production, reduced mTOR signaling, thereby suppressing the growth of cancer cells as well as tumor xenografts. Our results highlight the importance of 13-S-HODE serving as a tumor suppressive, mTOR-inhibiting metabolite that links polyunsaturated fatty acid metabolism and the mTOR signaling in controlling cancer cell growth.

biochemistry↗

bioRxiv2025

A statistical framework for inferring genetic requirements from embryo-scale single-cell sequencing experiments

Improvements in single-cell sequencing have enabled phenotyping at organism-scale and molecular resolution, but interpreting such experiments poses computational challenges. Identifying the genes and cell types directly impacted by genetic, chemical, or environmental perturbations requires explicit modeling of lineage relationships amongst many cell types, over time, from datasets with millions of cells collected from thousands of specimens. We describe two software tools, "Hooke" and "Platt", which exploit the rich statistical patterns within single-cell datasets to characterize the direct molecular and cellular consequences of experimental perturbations. We apply Hooke and Platt to a single-cell atlas of thousands of perturbed zebrafish embryos to synthesize a coherent map of lineage dependencies and leverage it to reveal previously unappreciated roles for fate-determining transcription factors. We show that cell type covariation in single-cell datasets is a powerful source of information for inferring how cells depend on genes and one another in the program of vertebrate development.

bioinformatics↗

bioRxiv2025

Embryo-scale single-cell chemical transcriptomics reveals dependencies between cell types and signaling pathways

Organogenesis is a highly organized process that is conserved across vertebrates and is heavily dependent on intercellular signaling to achieve cell type identity. We lack a comprehensive understanding of how developing cell types in each organ and tissue depend on developmental signaling pathways. To address this gap in knowledge, we captured the molecular consequences of inhibiting each of the seven major developmental signaling pathways in zebrafish, using large-scale whole embryo single cell RNA-seq from over two million cells. This approach allowed us to detect signaling pathway regulation even in very rare cell types. By focusing on the development of the pectoral fin, we uncovered two new cell types (distal mesenchyme and tenocytes) and multiple novel signaling dependencies during pectoral fin development. This resource serves as a valuable tool for investigators seeking to rapidly assess the role of the major signaling pathways during the formation of their tissue of interest.

developmental biology↗

bioRxiv2025

Designed Peptides as Affinity Ligands for Extracellular-Vesicle-based Cancer Biomarker Detection

Assays for cancer diagnosis via the analysis of tumor biomarkers on circulating extracellular vesicles (EVs) have shown great potential. Single EV imaging that can measure the abundance of protein biomarkers in EVs can help in detecting the presence, stage, and progression of disease. Antibodies are typically used to detect EV proteins. Controlling the quality of antibody-based immunoassays can, however, be challenging, as they may exhibit unintended cross-reactivity with non-target proteins or variability in binding affinity across different batches, even for monoclonal antibodies. Here, we report short peptides that are a promising alternative to antibodies for detecting protein biomarkers in EVs. We describe an effort that combines a Peptide Binding Design (PepBD) algorithm and molecular-level simulations to identify peptides that can recognize (1) the extracellular domain of EpCAM (a known cancer biomarker) and (2) the extracellular domain of tetraspanin CD81(a protein commonly expressed on the surface of EVs). The peptides designed for EpCAM and CD81 were labeled with a fluorescent dye and their binding to the target protein was evaluated using fluorescence ELISA. The results demonstrated that one of the computationally designed peptides EP-2.1 exhibited an affinity for the target EpCAM protein comparable to that of an antibody. Further testing involved single EV imaging to gauge the peptides affinity towards EVs. Designed peptides for both the target proteins showed similar affinity as antibodies to EVs.

cancer biology↗

bioRxiv2025

Temporal and Spatial Scales of Resting-state Human Cortical Activity Throughout Lifespan

Sensorimotor and cognitive abilities undergo substantial changes throughout the human lifespan, but the corresponding changes in the functional properties of cortical networks remain poorly understood. This can be studied using temporal and spatial scales of functional magnetic resonance imaging (fMRI) signals, which provide a robust description of the topological structure and temporal dynamics of neural activity. For example, timescales of resting-state fMRI signals parsimoniously predict a significant amount of the individual variability in functional connectivity networks identified in adult human brains. In the present study, we quantified and compared temporal and spatial scales in resting-state fMRI data collected from 2,352 subjects between the ages of 5 and 100 in Developmental, Young Adult, and Aging datasets from the Human Connectome Project. For most cortical regions, we found that both temporal and spatial scales decreased with age throughout the lifespan, with the visual cortex and the limbic network consistently showing the largest and smallest scales, respectively. For some prefrontal regions, however, these two scales displayed non-monotonic trajectories and peaked around the same time during adolescence and decreased throughout the rest of the lifespan. We also found that cortical myelination increased monotonically throughout the lifespan, and its rate of change was significantly correlated with the changes in both temporal and spatial scales across different cortical regions in adulthood. These findings suggest that temporal and spatial scales in fMRI signals, as well as cortical myelination, are closely coordinated during both development and aging. Significance StatementTemporal and spatial scales of resting-state cortical activity in humans measured by fMRI largely decreased throughout the lifespan, except that for some regions in the prefrontal cortex they peaked similarly during adolescence. In addition, whereas cortical myelination consistently increased throughout the lifespan, its variation across different cortical networks and the rate of age-related changes were correlated with the dynamics of temporal and spatial scales of rs-fMRI activity, suggesting that the spatio-temporal scales of cortical activity and cortical myelination might be co-regulated during development and aging.

neuroscience↗

bioRxiv2025

Generative frame interpolation enhances tracking of biological objects in time-lapse microscopy

Object tracking in microscopy videos is crucial for understanding biological processes. While existing methods often require fine-tuning tracking algorithms to fit the image dataset, here we explored an alternative paradigm: augmenting the image time-lapse dataset to fit the tracking algorithm. To test this approach, we evaluated whether generative video frame interpolation can augment the temporal resolution of time-lapse microscopy and facilitate object tracking in multiple biological contexts. We systematically compared the capacity of Latent Diffusion Model for Video Frame Interpolation (LDMVFI), Real-time Intermediate Flow Estimation (RIFE), Compression-Driven Frame Interpolation (CDFI), and Frame Interpolation for Large Motion (FILM) to generate synthetic microscopy images derived from interpolating real images. Our testing image time series ranged from fluorescently labeled nuclei to bacteria, yeast, cancer cells, and organoids. We showed that the off-the-shelf frame interpolation algorithms produced bio-realistic image interpolation even without dataset-specific retraining, as judged by high structural image similarity and the capacity to produce segmentations that closely resemble results from real images. Using a simple tracking algorithm based on mask overlap, we confirmed that frame interpolation significantly improved tracking across several datasets without requiring extensive parameter tuning and capturing complex trajectories that were difficult to resolve in the original image time series. Taken together, our findings highlight the potential of generative frame interpolation to improve tracking in time-lapse microscopy across diverse scenarios, suggesting that a generalist tracking algorithm for microscopy could be developed by combining deep learning segmentation models with generative frame interpolation.

cell biology↗

bioRxiv2025

randPedPCA: Rapid approximation of principal components from large pedigrees

BackgroundPedigrees continue to be extremely important in agriculture and conservation genetics with the pedigrees of modern breeding programmes easily comprising millions of records. The structure of such pedigrees is challenging to visualise. Being directed acyclic graphs, pedigrees can be represented as matrices. Common choices are the numerator relationship matrix, A, and the adjacency matrix T. With these matrices, the structure of pedigrees can then, in principle, be visualised via principal component analysis (PCA). However, the naive PCA of matrices for large pedigrees is challenging due to computational and memory constraints. ResultsWe present the open-access R package randPedPCA for rapid pedigree PCA using sparse matrices and randomised linear algebra. Our rapid pedigree PCA builds on the fact that matrix-vector multiplications with the numerator relationship matrix can be carried out implicitly using the extremely sparse inverse Cholesky factor of the numerator relationship matrix. We demonstrate the utility of randPedPCA using several simulated datasets of pedigrees and SNPs. We then demonstrate the performance of randPedPCA by analysing the pedigree of the UK Labrador Retriever breeding population of almost 1.5 million individuals. ConclusionsThe structure of pedigrees can be efficiently and rapidly visualised using scatter plots of principal component scores. For large pedigrees, this is considerably faster than rendering plots of a pedigree graph.

genetics↗

bioRxiv2025

INTACT: A method for integration of longitudinal physical activity data from multiple sources

Wearable devices and digital phenotyping are increasingly used in observational and interventional studies to measure real-time biosignals such as physical activity. However, integrating and comparing data across studies and cohorts remains challenging due to variability in device types, acquisition protocols, and preprocessing methods. A key challenge is removing unwanted study- or device-specific effects while preserving meaningful biological signals. These difficulties are exacerbated by the longitudinal and within-day correlations inherent in high-resolution time-varying data collected from wearable sensors. To address this, we propose INTACT (INtegration of Time-varying data from weArable sensors for physiCal acTivity), a novel method for harmonizing time-varying physical activity intensity data from accelerometers. INTACT models shared information through common eigenvalues and eigenfunctions while allowing for source-specific scale and rotation adjustments. We apply the proposed method to two real-world applications: (i) integration of accelerometer data from two waves of the National Health and Nutrition Examination Survey (NHANES), measured using different devices and reported in different units; and (ii) integration of NHANES accelerometry data with accelerometer and gyroscope measures from commercial devices. Across both applications, INTACT outperforms existing approaches in mitigating source effects while preserving biological variation, enabling more reliable cross-study comparisons of physical activity patterns.

bioinformatics↗

bioRxiv2025

Seed structure and phosphorylation in the fuzzy coat impact tau seeding competency

Tau is a pathogenic protein in Alzheimers (AD) and other neurodegenerative diseases. The misfolding of tau into {beta}-sheet rich elongated filaments is thought to be a key event in disease pathogenesis, followed by subsequent templated recruitment of monomeric tau into this pathogenic form. Cryo-electron microscopy has revealed that specific tau conformations characterize different diseases. In this study, we explored how tau filament core structure and post-translational modifications in its disordered fuzzy coat influence its seeding capacity in primary neurons and mice. We show that the structure of the seeds affects seeding capacity, but that the AD tau core structure alone is insufficient to capture the full seeding capacity of AD tau. Proteolytic cleavage of AD tau which removes the fuzzy coat causes a loss in seeding capacity, as does removal of phosphorylation from the fuzzy coat by phosphatase treatment. Re-phosphorylation of phosphatase-treated AD tau by kinase treatment partially restores seeding activity. Finally, we find that filaments of recombinant tau with twelve phospho-mimetic residues (PAD12 tau) with the AD fold are able to recapitulate the seeding capacity of AD tau. Combined, these results suggest that the structure of the ordered core, together with phosphorylation in the fuzzy coat, confers the seeding capacity of tau filaments.

neuroscience↗

bioRxiv2025

On ARGs, pedigrees, and genetic relatedness matrices

Genetic relatedness is a central concept in genetics, underpinning studies of population and quantitative genetics in human, animal, and plant settings. It is typically stored as a genetic relatedness matrix (GRM), whose elements are pairwise relatedness values between individuals. This relatedness has been defined in various contexts based on pedigree, genotype, phylogeny, coalescent times, and, recently, ancestral recombination graph (ARG). ARG-based GRMs have been found to better capture the structure of a population and improve association studies relative to the genotype GRM. However, calculating GRMs and further operations with them is fundamentally challenging due to inherent quadratic time and space complexity. Here, we first discuss the different definitions of relatedness in a unifying context, making use of the additive model of a quantitative trait to provide a definition of "branch relatedness" and the corresponding "branch GRM". We explore the relationship between branch relatedness and pedigree relatedness through a case study of French-Canadian individuals that have a known pedigree. Through the tree sequence encoding of an ARG, we then derive an efficient algorithm for computing products between the branch GRM and a general vector, without explicitly forming the branch GRM. This algorithm leverages the sparse encoding of genomes with the tree sequence and hence enables large-scale computations with the branch GRM. We demonstrate the power of this algorithm by developing a randomized principal components algorithm for tree sequences that easily scales to millions of genomes. All algorithms are implemented in the open source tskit Python package. Taken together, this work consolidates the different notions of relatedness as branch relatedness and by leveraging the tree sequence encoding of an ARG it provides efficient algorithms that enable computations with the branch GRM that scale to mega-scale genomic datasets.

genomics↗

bioRxiv2025

Loss of JAK1 Function Causes G2/M Cell Cycle Defects Vulnerable to Kif18a Inhibition

To gain insight into biological mechanisms that cause resistance to DNA damage, we performed parallel pooled genetic CRISPR-Cas9 screening for survival in high risk HNSCC subtypes. Surprisingly, and in addition to ATM, DNAPK, and NFKB signaling, JAK1 was identified as a driver of tumor cell radiosensitivity. Knockout of JAK1 in HNSCC increases cell survival by enhancing the DNA damage-induced G2 arrest, and both knockout and JAK1 inhibition with abrocitinib prevent subsequent formation of radiation-induced micronuclei. Loss of JAK1 function does not affect canonical CDK1 signaling but does reduce activation of PLK1 and AURKA, kinases that regulate both G2 and M phase progression. Correspondingly, JAK1 KO was found to cause mitotic defects using both EdU labeling and live cell imaging techniques. Given this insight, we evaluated Kif18a inhibition as an approach to exacerbate mitotic stress and enhance the efficacy of radiation. These studies establish Kif18a inhibition as a novel strategy to counteract therapeutic resistance to DNA damage mediated by G2 cell cycle arrest.

cancer biology↗

bioRxiv2025

Inhibiting fibronectin assembly in the breast tumor microenvironment increases cell death and improves response to doxorubicin

PurposeEffective therapies for solid tumors, including breast cancers, are hindered by several roadblocks that can be largely attributed to the fibrotic extracellular matrix (ECM). Fibronectin (FN) is a highly upregulated ECM component in the fibrotic tumor stroma and is associated with poor patient prognosis. This study aimed to investigate the therapeutic potential of an anti-fibrotic peptide that specifically targets FN and blocks the fibrillar assembly of FN. MethodsTo target FN, we used PEGylated Functional Upstream Domain (PEG-FUD), which binds to the 70 kDa N-terminal region of FN with high affinity, localizes to mammary tumors, and potently inhibits FN assembly in vitro and in vivo. Here, we used the 4T1 tumor model to investigate the efficacy and mechanisms of PEG-FUD to inhibit tumor growth. ResultsOur data demonstrates that PEG-FUD monotherapy reduces tumor growth without systemic toxicity. Analysis of the tumor microenvironment revealed that PEG-FUD effectively inhibited FN matrix assembly within tumors and reduced adhesion-mediated signaling through 5 integrin and FAK leading to enhanced tumor cell death. Notably, signaling through FAK has been associated with resistance mechanisms to doxorubicin (DOX). Therefore, we tested the combination of PEG-FUD and Dox, which significantly reduced tumor growth by 60% compared to vehicle control and 30% compared to Dox monotherapy. ConclusionsOur findings demonstrate that PEG-FUD significantly modifies the peritumoral ECM of breast cancer, leading to increased tumor cell death, and potentiates the efficacy of conventional breast cancer therapy.

cancer biology↗

bioRxiv2025

Protein folding stability estimation with an explicit consideration of unfolded states

Folding stability is a critical requirement for the vast majority of proteins. Computational methods suggested to date for the absolute folding stability ({Delta}G) prediction - including those driven from protein structure prediction AIs - show clear limitations on reproducing quantitative experimental values. Here we present IFUM, a deep neural network that jointly estimates {Delta}G and the equilibrium ensemble of folded and unfolded states represented by their residue-pair distance probability distributions. This joint learning considerably enhances the {Delta}G prediction accuracy against the scenario where {Delta}G prediction was learned alone. To improve the model, we extend the dataset beyond previous related works to include the Mega-scale small proteins and disordered proteins for training as well as wild-type natural proteins with sizes up to 869 residues for validation. We show that IFUM is robust to various protein types and sizes, and is capable of accurately predicting more general types of mutational effects such as sequence insertions or deletions. The applicability of IFUM is demonstrated through two real-world design challenges. First, for blind-tested protein engineering scenarios containing many sequence substitutions and insertions, good correlation with experimental melting temperatures (Tm) is observed. Second, for the de novo design selection, IFUM shows considerably improved performance over broadly used AlphaFold-based metrics. IFUM is a free software available at github.com/HParklab/IFUM and also through Google Colab.

bioinformatics↗

bioRxiv2025

A CD25-CCR7 complex initiates non-canonical IL-2 signaling

IL-2, a central regulator of immune function, binds to its receptor subunit CD25 (IL-2R), promoting IL-2 interaction with {beta} and {gamma} subunits to trigger the canonical IL-2 signaling pathway. An anti-mouse CD25 antibody, PC61, triggers alternative IL-2 signaling, leading to integrin activation. PC61 induces a complex formed by the IL-2-dependent association of CD25 with CCR7, suggesting that the formation of this complex initiates alternative IL-2 signaling. Here, we used structure-based design together with combinatorial screening to identify an IL-2 mutant (denoted IL-2(E52K)) that spares canonical IL-2 signaling but disrupts both PC61-induced complex formation and integrin activation while retaining the full CD25 affinity of the parent molecule. We also report that heparan sulfate (HS), a physiological ligand of IL-2 that triggers alternative signaling, induced IL-2-dependent CD25-CCR7 association, whereas IL-2(E52K) failed to support both HS-induced CD25-CCR7 complex formation and integrin activation. Thus, both anti-CD25 antibody and HS require common features of IL-2 needed for CD25-CCR7 complex assembly and resulting integrin activation. Collectively, these data show that IL-2 promotes CD25 interaction with CCR7, thereby forming the signal initiating complex. Furthermore, canonical and alternative IL-2 signaling can be decoupled by an IL-2 mutation, creating a tool to specify the biological role of alternative IL-2 signaling in immune responses.

cell biology↗

bioRxiv2025

Gut-Brain Hydraulics: Brain motion and CSF circulation is driven by mechanical coupling with the abdomen

The brain moves within the skull, but the drivers and function of this motion are not understood. We visualized brain motion relative to the skull in awake head-fixed mice using high-speed, multi-plane two-photon microscopy. Brain motion was primarily rostrally and laterally directed, and was tightly correlated with locomotion, but not with respiration or the cardiac cycle. Electromyography recordings in abdominal muscles and microCT reconstructions of the trunk and spinal vasculature showed that brain motion was driven by abdominal muscle contractions that activate a hydraulic-like vascular connection between the nervous system and the abdominal cavity. Externally-applied abdominal pressure generated brain motion similar to those seen during abdominal muscle contractions. Simulations showed that brain motion drives substantial volumes of interstitial fluid through and out of the brain (at volumetric rates several times higher than production) into the subarachnoid space, in the opposite direction of fluid flow seen during sleep. The brain is hydraulically linked to the abdominal compartment, and fluid flow in the brain is coupled to body movements, providing a mechanism by which the mechanics of the viscera directly impact brain health.

neuroscience↗

bioRxiv2025

Self-Contemplating In-Context Learning Enhances T Cell Receptor Generation for Novel Epitopes

Computational design of T cell receptors (TCRs) that bind to epitopes holds the potential to revolutionize targeted immunotherapy. However, computational design of TCRs for novel epitopes is challenging due to the scarcity of training data, and the absence of known cognate TCRs for novel epitopes. In this study, we aim to generate high-quality cognate TCRs particularly for novel epitopes with no known cognate TCRs, a problem that remains under-explored in the field. We propose to incorporate in-context learning, successfully used with large language models to perform new generative tasks, to the task of TCR generation for novel epitopes. By providing cognate TCRs as additional context, we enhance the models ability to generate high-quality TCRs for novel epitopes. We first unlock the power of in-context learning by training a model to generate new TCRs based on both a target epitope and a small set of its cognate TCRs, so-called in-context training (ICT). We then self-generate its own TCR contexts based on a target epitope, as novel epitopes lack known binding TCRs, and use it as an inference prompt, referred to as self-contemplation prompting (SCP). Our experiments first demonstrate that aligning training and inference distribution by ICT is critical for effectively leveraging context TCRs. Subsequently, we show that providing context TCRs significantly improves TCR generation for novel epitopes. Furthermore, we show TCR generation using SCP-synthesized context TCRs achieves performance comparable to, and sometimes surpassing, ground-truth context TCRs, especially when combined with refined prompt selection based on binding affinity and authenticity metrics.

bioinformatics↗