Biology subjects

Li, K. Y.

Publications and source records attributed to Li, K. Y..

10 recordsLinked to original sources

bioRxiv2026

A Manifold-Based Measure of Transcriptional Entropy for Quantifying Aging in Single Cells

Aging reshapes tissues through changes in cellular composition, coordinated transcriptional reprogramming, and loss of transcriptional coordination. Whereas the first two have been characterized across aging tissues, the third remains difficult to quantify. We introduce an unsupervised, first-principles framework for measuring transcriptional dyscoordination in single cells as deviation from a learned, predictable structure accounting for technical noise. Orthogonal validation links transcriptional dyscoordination to classical intrinsic noise and distinguishes it from coordinated change. In controlled perturbations, dyscoordination rises after genotoxic injury and senescence induction, then falls following senolytic depletion. Across mouse, rat, and human tissues, dyscoordination increases with chronological age, especially in regenerative compartments. In human T cells, dyscoordination increases with clonal expansion and effector function yet declines within persisting clones after checkpoint blockade. Cross-modal analyses further link dyscoordination to chromatin-based mitotic age and genome instability. These results identify loss of transcriptional coordination as a distinct and dynamic feature of cellular aging.

genomics↗

bioRxiv2025

A knowledge-guided approach to recovering important rare signals from high-dimensional single-cell data

Single-cell transcriptomic data are high-dimensional, with many genes profiled in each cell. Dimensionality reduction is routinely applied to improve interpretability, remove noise and redundancy, and enable visualization. Most existing methods aim at preserving the most prominent data properties, which can lead to omission of rare but important signals. Here we propose a novel framework that uses knowledge-derived genes of interest to guide dimensionality reduction, which can help cluster rare cells and separate highly similar cell sub-populations. We demonstrate the utility of our framework in identifying endocrine cell subtypes in the pancreatic islet, highly similar hematopoietic sub-populations, and rare senescent cells.

bioinformatics↗

bioRxiv2025

Targeting CyclinD1-CDK6 to Mitigate Senescence-Driven Inflammation and Age-Associated Functional Decline

Cellular senescence contributes to aging and age-related diseases by driving chronic inflammation through the Senescence Associated Secretory Phenotype (SASP) and interferon-stimulated genes (ISGs). Cyclin D1 (CCND1), a key cell cycle regulator, is paradoxically upregulated in these non-proliferating cells. We show that CCND1 and its kinase partner CDK6 drive SASP and ISG expression in senescent cells by promoting DNA damage accumulation. This leads to the formation of cytoplasmic chromatin fragments (CCFs) that activate pro-inflammatory CGAS-STING signaling. The tumor suppressor p53 (TP53) and its target p21 (CDKN2A) antagonize this CCND1-CDK6-dependent DNA damage accumulation pathway to suppress the SASP. In aged mouse livers, senescent hepatocytes show increased Ccnd1 expression. Hepatocyte-specific Ccnd1 knockout or treatment with the Cdk4/6 inhibitor Palbociclib reduces DNA damage and ISGs in aged mouse liver. Notably, Palbociclib also suppresses frailty and improves physical performance of aged mice. These findings reveal a novel role for CCND1/CDK6 in regulating DNA damage and inflammation in senescence and aging, highlighting it as a promising therapeutic target.

cell biology↗

bioRxiv2025

Few-shot in-context learning with large language models for antibody characterization

Large language models (LLMs) can learn new tasks by in-context learning (ICL), but it is unknown whether this ability reliably transfers to biological sequence classification. Here, we systematically evaluate how demonstration selection, shot count, and prompting strategies affect performance across 20 general-purpose LLMs. Using antibody characterization as a representative test case, we compare zero-shot, few-shot, and chain-of-thought (CoT) ICL on three classification tasks: humanness, antigen specificity, and isotype. Our results reveal a clear performance hierarchy: while zero-shot prompting performs near chance, few-shot prompting with randomly selected demonstrations improves performance, showing that LLMs can perform ICL using biological sequences from minimal supervision. However, matching protein-language model (pLM)-based classifier accuracy is only achieved when using label-diverse demonstrations drawn from antibodies similar to the query sequence. To leverage this insight, we introduce Sim-ICL, a framework that automatically retrieves such demonstrations. Using only 32-shot prompting, Sim-ICL achieves performance competitive with pLM-based classifiers, matching or outperforming them in two of the three tasks. Furthermore, reasoning-oriented prompts yield marginal gains and often produce fluent but biologically incorrect rationales, suggesting that current CoT explanations function as after-the-fact rationalizations rather than capturing mechanistic determinants of antibody properties. From these experiments, we derive practical design principles for ICL on biological sequences: use similarity-based, label-diverse demonstrations and modest shot counts, and treat reasoning prompts primarily as post hoc narratives rather than drivers of performance. Sim-ICL implements these principles in a streamlined, prompt-based framework for antibody sequence classification and, in principle, could be adapted to other biological sequence tasks.

bioinformatics↗

bioRxiv2024

A Comprehensive Assessment of Methylation-Based Age Prediction Methods

DNA methylation (DNAm) clock is widely used to measure biological age, helping to identify key biomarkers associated with aging, infer the progression of aging, and have promise for elucidating, delaying, or even reversing aging. During the past decade, a large number of epigenetic clocks have been developed. However, they are decentralized, with applicable scopes overlapping. We benchmark 15 of these methods on 142 Illumina DNAm array datasets in five criteria and analyze the biological significance of CPGs about aging and overlapping. There are many exciting commons in models performance. We found the optimal model closely related to the numbers and characteristics of the training data. We provided a comprehensive assessment process to guide DNAm clock research at (https://dnamclock.com), the corresponding data and evaluation pipeline are freely available (https://github.comyNENUBioCompute/MethylationEvaluation), this study will aid in the development of improved tools designed to analyze increasingly large DNAm datasets.

bioinformatics↗

bioRxiv2024

SUMO Inhibition Plus CD40 Agonism Increases Anti-Tumor Immunogenicity Through Interferon Mediated Macrophage Activation

Resistance to immunotherapy is a cardinal feature of pancreatic ductal adenocarcinoma (PDAC). Inhibition of Small Ubiquitin-like MOdifier (SUMO), a post-translational modification with important immune regulatory functions, augments responsiveness to immunotherapy in non-PDAC models via pro-immunogenic effects on myeloid cells, cancer cells, and T-cells. Recently, it has been reported that SUMO inhibition has direct immunogenic effects on PDAC. Here, we report that the novel combination of SUMO inhibition with a small molecule, TAK-981, plus antibody-mediated CD40 agonism improves survival in an aggressive orthotopic mouse model of PDAC by enhancing anti-tumoral immunogenicity. This combination amplifies CD8+ T-cell tumor infiltration and induces significant changes among macrophages. TAK-981 also leads to enhanced cancer specific MHC-I expression both in vitro and in vivo by augmenting interferon signaling. We show that the improvement in survival is mediated by macrophages. Our findings show that SUMO inhibition complements CD40 agonism to enhance immune activity in PDAC via interferon signaling, improving survival in an aggressive pre-clinical model of PDAC and translating previous findings to a characteristically immunosuppressive and highly aggressive solid malignancy.

cancer biology↗

bioRxiv2024

Overcoming artificial structures in resolution-enhanced Hi-C data by signal decomposition and multi-scale attention

Computational enhancement is an important strategy for inferring high-resolution features from genome-wide chromosome conformation capture (Hi-C) data, which typically have limited resolution. Deep learning has been highly successful in this task but we show that it creates prevalent artificial structures in the enhanced data due to the need to divide the large contact matrix into small patches. In addition, previous deep learning methods largely focus on local patterns, which cannot fully capture the complexity of Hi-C data. Here we propose Smooth, High-resolution, and Accurate Reconstruction of Patterns (SHARP) for enhancing Hi-C data. It uses the novel approach of decomposing the data into three types of signals, due to one-dimensional proximity, contiguous domains, and other fine structures, and applies deep learning only to the third type of signals, such that enhancement of the first two is unaffected by the patches. For the deep learning part, SHARP uses both local and global attention mechanisms to capture multi-scale contextual information. We compare SHARP with state-of-the-art methods extensively, including application to data from new samples and another species, and show that SHARP has superior performance in terms of resolution enhancement accuracy, avoiding creation of artificial structures, identifying significant interactions, and enrichment in chromatin states.

bioinformatics↗

bioRxiv2022

Regulatory roles of three-dimensional structures of topologically associating domains

1Transcriptional enhancers usually, but not always, regulate genes within the same topologically associating domain (TAD). We hypothesize that this incomplete insulation is due to three-dimensional structures of corresponding chromatin domains in individual cells: Whereas enhancers and genes buried inside the "core" of a domain interact mostly with other regions in the same domain, those on the "surface" can more easily interact with the outside. Here we show that a simple measure, the intra-TAD ratio, can quantify the "coreness" of a region with respect to single-cell domains it belongs. We show that domain surfaces are permissive for high gene expression, and cell type-specific active cis-regulatory elements (CREs), active histone marks, and transcription factor binding sites are enriched on domain surfaces, most strongly in chromatin subcompartments typically considered inactive. These findings suggest a "domain surface CRE" model of gene regulation. We also find that disease-associated non-coding variants are enriched on domain surfaces.

bioinformatics↗

bioRxiv2021

Single-molecule imaging of chromatin remodelers reveals role of ATPase in promoting fast kinetics of target search and dissociation from chromatin

Conserved ATP-dependent chromatin remodelers establish and maintain genome-wide chromatin architectures of regulatory DNA during cellular lifespan, but the temporal interactions between remodelers and chromatin targets have been obscure. We performed live-cell single-molecule tracking for RSC, SWI/SNF, CHD1, ISW1, ISW2, and INO80 remodeling complexes in budding yeast and detected hyperkinetic behaviors for chromatin-bound molecules that frequently transition to the free state for all complexes. Chromatin-bound remodelers display notably higher diffusion than nucleosomal histones, and strikingly fast dissociation kinetics with 4-7 s mean residence times. These enhanced dynamics require ATP binding or hydrolysis by the catalytic ATPase, uncovering an additional function to its established role in nucleosome remodeling. Kinetic simulations show that multiple remodelers can repeatedly occupy the same promoter region on a timescale of minutes, implicating an unending tug-of-war that controls a temporally shifting window of accessibility for the transcription initiation machinery.

molecular biology↗

bioRxiv2020

Spatio-Temporal Coordination Of Transcription Preinitiation Complex Assembly In Live Cells

Transcription initiation by RNA polymerase II (Pol II) requires preinitiation complex (PIC) assembly at gene promoters. In the dynamic nucleus where thousands of promoters are broadly distributed in chromatin, it is unclear how ten individual components converge on any target to establish the PIC. Here, we use live-cell, single-molecule tracking in S. cerevisiae to document subdiffusive, constrained exploration of the nucleoplasm by PIC components and Mediators key functions in guiding this process. On chromatin, TBP, Mediator, and Pol II instruct assembly of a short-lived PIC, which occurs infrequently but efficiently at an average promoter where initiation-coupled disassembly may occur within a few seconds. Moreover, PIC exclusion by nucleosome encroachment underscores regulated promoter accessibility by chromatin remodeling. Thus, coordinated nuclear exploration and recruitment to accessible targets underlies dynamic PIC establishment in yeast. Collectively, our study provides a global spatio-temporal model for transcription initiation in live cells.

molecular biology↗