Biology subjects

Yue, Y.

Publications and source records attributed to Yue, Y..

At least 37 records · Page 2Linked to original sources

bioRxiv2025

Features fusion or not: harnessing multiple pathological foundation models using Meta-Encoder for downstream tasks fine-tuning

The emergence of diverse pathological foundation models has empowered computational pathology tasks, including tumor classification, biomarker prediction, and RNA expression prediction. However, variations in model architecture and data sources lead to inconsistent downstream performance and complicate centralized training. Specifically, the lack of data sharing makes retraining foundation models with pooled data infeasible. Alternatively, the release of model parameters enables combining multiple models during fine-tuning. Inspired by the meta-analysis method, we propose the Meta-Encoder framework, which integrates features from multiple foundation models to generate a comprehensive representation, improving downstream fine-tuning task performance. Comparative experiments demonstrate that Meta-Encoder is more effective than individual foundation models, with its strengths more pronounced in handling complex tasks. While single models may perform sufficiently well for simple tasks, Meta-Encoder can match or even surpass the best-performing single model, alleviating concerns over model selection. Moderately challenging tasks benefit from Meta-Encoders concatenation or self-attention strategies, with the latter demonstrating superior performance in more challenging scenarios. For highly complex tasks, such as high-dimensional gene expression prediction, self-attention proves to be the most effective Meta-Encoder strategy, balancing feature integration and computational efficiency. For three patch-level spatial gene expression prediction tasks (HEST-Benchmark, CRC-inhouse, and Her2ST), the self-attention strategy improved the Pearson correlation by 38.58%, 26.06%, and 20.39%, respectively, compared to the average performance of three patch-level single models. Similarly, for the TCGA-BRCA, TCGA-NSCLC, and TCGA-CRC WSI-level bulk gene expression prediction tasks, the Pearson correlation increased by 14.36%, 9.27%, and 42.55%, respectively, compared to the average performance of two WSI-level single models. By leveraging multiple pathological foundation models using Meta-Encoder, it can further improve molecular characterization in pathology images to advance precision oncology.

bioinformatics↗

bioRxiv2025

Tumor Treating Fields inhibit cell proliferation and promote immunogenic cell death in biliary tract cancer cells

Tumor Treating Fields (TTFields) are low-intensity, intermediate-frequency alternating electric fields that exert antimitotic effects on cancer cells. This study evaluated the in vitro efficacy of TTFields on biliary tract cancer (BTC) cell lines HCCC-9810 and RBE, investigating their sensitivity to TTFields across varying frequencies and electric field intensities. The results demonstrated that a frequency of 150 kHz induced the most pronounced cytotoxic effects, significantly impairing clonogenicity and migratory capacity while inducing abnormal mitotic processes. Furthermore, TTFields treatment triggered a marked upregulation of immunogenic cell death (ICD) biomarkers, including enhanced surface exposure of calreticulin (CRT) and increased extracellular release of high mobility group box 1 (HMGB1) and ATP. These findings indicate that TTFields represent a promising therapeutic approach for BTC, not only by suppressing tumor proliferation but also by enhancing ICD.

cancer biology↗

bioRxiv2024

Mapping Global Causal Responses to Noninvasive Modulation of Genetically and Spatially Targeted Neural Populations with Sonogenetic-fPET

Despite significant progress in brain circuit mapping over recent decades, a major challenge remains: no method currently allows for the noninvasive modulation of genetically and spatially defined neural populations while simultaneously monitoring their global effects throughout the brain and body. Here, we present sonogenetic-fPET, a technique that integrates sonogenetics with [18F]-2-fluoro-2-deoxy-D-glucose functional positron emission tomography (FDG-fPET) to overcome this challenge. Sonogenetics enables noninvasive, spatially targeted modulation of neurons genetically engineered to express the ultrasound-sensitive ion channel TRPV1, while FDG-fPET captures glucose metabolic changes triggered by this stimulation across the brain and body. We demonstrate the effectiveness of this technique by targeting neurons in the dorsal striatum, showcasing its capability to map global network responses to specific neuronal activation. Incorporating an acoustic hologram for sonogenetics further enables flexible modulation of different brain regions within a single mouse while concurrently mapping the resulting network activity. In summary, sonogenetic-fPET offers a tool for dissecting the global responses of the brain and body to the noninvasive modulation of genetically and spatially defined neuronal populations.

neuroscience↗

bioRxiv2024

Generalized cell phenotyping for spatial proteomics with language-informed vision models

We present DeepCell Types, a novel approach to cell phenotyping for spatial proteomics that addresses the challenge of generalization across diverse datasets with varying marker panels collected across different platforms. Our approach utilizes a transformer with channel-wise attention to create a language-informed vision model; this models semantic understanding of the underlying marker panel enables it to learn from and adapt to heterogeneous datasets. Leveraging a curated, diverse dataset named Expanded TissueNet with cell type labels spanning the literature and the NIH Human BioMolecular Atlas Program (HuBMAP) consortium, our model demonstrates robust performance across various cell types, tissues, and imaging modalities. Comprehensive benchmarking shows that our method outperforms existing approaches on cell-type prediction and, from the same model, predicts marker positivity competitively with a dedicated specialist; it further matches manual expert gating and adapts to new data with modest fine-tuning, well past what baselines reach when trained from scratch. This work equips the spatial proteomics community with a single, continuously improvable phenotyping model that generalizes to new marker panels and can be fine-tuned efficiently when needed. We release both DeepCell Types and Expanded TissueNet as open-source resources.

bioinformatics↗

bioRxiv2024

Evaluation of Machine Learning-Assisted Directed Evolution Across Diverse Combinatorial Landscapes

Various machine learning-assisted directed evolution (MLDE) strategies have been shown to identify high-fitness protein variants more efficiently than typical wet-lab directed evolution approaches. However, limited understanding of the factors influencing MLDE performance across diverse proteins has hindered optimal strategy selection for wet-lab campaigns. To address this, we systematically analyzed multiple MLDE strategies, including active learning and focused training using six distinct zero-shot predictors, across 16 diverse protein fitness landscapes. By quantifying landscape navigability with six attributes, we found that MLDE offers a greater advantage on landscapes which are more challenging for directed evolution, especially when focused training is combined with active learning. Despite varying levels of advantage across landscapes, focused training with zero-shot predictors leveraging distinct evolutionary, structural, and stability knowledge sources consistently outperforms random sampling for both binding interactions and enzyme activities. Our findings provide practical guidelines for selecting MLDE strategies for protein engineering.

bioengineering↗

bioRxiv2024

The mechanism of bacterial defense system DdmDE from Lactobacillus casei

Bacteria has developed a diverse array of defense mechanisms to protect against invasion by mobile genetic elements. Recent study identified a bacterial defense module DdmDE system which encodes a helicase-nuclease fusion protein DdmD and a putative prokaryotic Argonaute-like protein DdmE, imposing fitness advantages to the bacteria by eliminating invasive plasmids. However, the mechanistic basis of how DdmDE system detects and degrades plasmids is not fully understood. Here, by studying the DdmDE system from Lactobacillus casei (LcDdmDE), we found that LcDdmD is able to degrade ssDNA and nick plasmids in the presence of Mn2+, and it exhibits 5-3 DNA helicase activity in a ssDNA length-dependent mechanism. Meanwhile, LcDdmD serves as a sensor that utilizes DNA guide to recognize target DNA. We determined the cryo-EM structures of LcDdmD dimer bound with fork DNA, guide/target DNA-bound LcDdmE, and the complex of LcDdmDE-bubble DNA in intermediate state as well as active state. Together with functional analysis, we revealed the working mechanism of LcDdmDE system. In such a scenario, guided by ssDNA, LcDdmE recruits auto-inhibited LcDdmD dimer loading onto DNA target. Through substantial conformational changes, LcDdmD dimer dissociates into active monomer and unwind duplex for plasmid degradation. Our study provides structural insights into the mechanism of DdmDE, presenting pAgo-directed plasmid degradation by the allosterically regulated helicase-nuclease.

molecular biology↗

bioRxiv2024

Cholinergic feedback for context-specific modulation of sensory representations

The brains ability to prioritize behaviorally relevant sensory information is crucial for adaptive behavior, yet the underlying mechanisms remain unclear. Here, we investigated the role of basal forebrain cholinergic neurons in modulating olfactory bulb (OB) circuits in mice. Calcium imaging of cholinergic feedback axons in OB revealed that their activity is strongly correlated with orofacial movements, with little responses to passively experienced odor stimuli. However, when mice engaged in an odor discrimination task, OB cholinergic axons rapidly shifted their response patterns from movement-correlated activity to odor-aligned responses. Notably, these odor responses during olfactory task engagement were absent in cholinergic axons projecting to the dorsal cortex. The level of odor responses correlated with task performance. Inactivation of OB-projecting cholinergic neurons during task engagement impaired performance and reduced odor responses in OB granule cells. Thus, the cholinergic system dynamically modulates sensory processing in a modality-specific and context-dependent manner, providing a mechanism for a flexible and adaptive sensory prioritization.

neuroscience↗

bioRxiv2024

Haplotype-resolved genome assembly of upland switchgrass provides insights into cold tolerance

Switchgrass (Panicum virgatum L.) is a bioenergy and forage crop. Upland switchgrass exhibits superior cold tolerance than lowland ecotype, but the underlying molecular mechanisms remain unclear. Here, we presented a high-quality haplotype-resolved genome of the upland ecotype Jingji31 and conducted multi-omics analysis to understand its cold tolerance. The divergence between upland and lowland ecotypes of switchgrass occurred after the differentiation of the two subgenomes (K and N). Under cold stress, the K subgenome has more differentially expressed genes (DEGs). Transcriptome analysis revealed ecotype-specific differential expressions among members of the cold-responsive (COR) gene families. Specifically, certain members of the AFB1, ATL80, HOS10, and STRS2 gene families exhibited opposite expression changes between the two ecotypes, potentially contributing to their differential cold tolerance. By using haplotype-resolved genome, we identified more cold-induced allele-specific expressions (ASEs) in the upland ecotype, and these ASEs were significantly enriched in the COR gene families. Genome-wide association study detected an association signal on Chr3K related to overwintering rate, which overlapped with a selective sweep region and contained a cytochrome P450 (CYP450) gene highly expressed under cold stress. Heterologous overexpression of CYP450 in rice alleviated leaf wilting and improved cold tolerance. Our study provides a high-quality haplotype-resolved genome of upland switchgrass, to advance conceptual understanding of plant cold tolerance for breeding crops with enhanced cold adaptation.

genomics↗

bioRxiv2024

Active Learning-Assisted Directed Evolution

Directed evolution (DE) is a powerful tool to optimize protein fitness for a specific application. However, DE can be inefficient when mutations exhibit non-additive, or epistatic, behavior. Here, we present Active Learning-assisted Directed Evolution (ALDE), an iterative machine learning-assisted DE workflow that leverages uncertainty quantification to explore the search space of proteins more efficiently than current DE methods. We apply ALDE to an engineering landscape that is challenging for DE: optimization of five epistatic residues in the active site of an enzyme. In three rounds of wet-lab experimentation, we improve the yield of a desired product of a non-native cyclopropanation reaction from 12% to 93%. We also perform computational simulations on existing protein sequence-fitness datasets to support our argument that ALDE can be more effective than DE. Overall, ALDE is a practical and broadly applicable strategy to unlock improved protein engineering outcomes.

bioengineering↗

bioRxiv2024

Systemic response to nitrogen starvation in the cyanobacteria hosting hornwort and liverwort provides little evidence for extensive priming to the cyanobiont

Mutualistic plant-microbe symbiotic interactions are thought to have evolved from a loose association between host plants and microbes when nutrients are limited. Therefore, the molecular network enabling intimate mutualistic plant-microbe symbioses may have evolved from a nutrient starvation response shared by all land plants. While the molecular link between nutrient status and symbiotic interaction is well-established, it remains poorly understood in some systems. This is especially true for the symbiotic associations between plants and cyanobacteria. To test the conservation of the starvation network across land plants as well as to investigate the link between nutrient starvation and symbiosis initiation in the plant-cyanobacteria symbiosis, here we explore the transcriptional responses to nutrient starvation in two non-vascular plant species, a hornwort Anthoceros agrestis and a liverwort Blasia pusilla, forming plant-cyanobacteria endophytic symbioses. We observe a deep conservation of the systemic starvation response across land plants. However, very few if any components of the starvation network appear to be specific to cyanobacteria hosting plants, providing little evidence for extensive and specific priming to the cyanobiont. Moreover, we found that some bioactive molecules known to be important in initiating the plant-mycorrhiza and nodule-forming bacteria symbioses, may also have a similar role in plant-cyanobacteria symbioses. HighlightOur results suggest that the most critical step in establishing plant-cyanobacteria interactions using non-host plants is the attraction of the cyanobiont. This finding has significant impact on crop engineering.

evolutionary biology↗

bioRxiv2024

Rapid Response of Neutrophils Determined by Shortest Differentiation Trajectories and Absence of Self-Renewal in committed Progenitors

Despite substantial advances in our understanding of hematopoietic stem cell self-renewal, differentiation, and proliferation, the underlying mechanisms and cell cycle kinetics remain elusive. Neutrophils, as primary rapid responders during infections, quickly generate an enormous number of mature cells for immune defense. This immune phenomena is effected by cell cycle kinetics. Our previous study redefined progenitors and reformed the hierarchy of hematopoiesis. How exactly lineage commitment and cell cycle are coordinated remain to be elucidated. Here, we aim to elucidate the cell cycle kinetics and chronological order of progenitor states, providing new insights into immune phenomena. We revealed the changes in cell cycle kinetics from differentiation (lineage commitment) to proliferation (cell division) in peripheral hematopoiesis. Differentiating hematopoietic progenitor cells (HPCs) are maintained in the G1 phase of cell cycle, regulated by DNA replication. CDCA7 was identified as an essential factor for DNA replication, facilitating the transition of progenitor cells from G1 to S phase. Once the progenitors complete their commitment, cell cycle states of progenitors convert from G1 to S phase, switching from lineage commitment to cell division with DNA replication begins. Lineage commitment and cell division of MPCs are independent processes, and fate determination takes precedence over proliferation. Since the differentiating HPCs do not undergo cell division, a committing progenitor will destined to mature into only one committed progenitor cell along its designated differentiation trajectory. The chronological order of quiescence, self-renewal, differentiation, and proliferation, was clearly delineated. Our study proposed the concept of "temporal sequence control," emphasizing that fate determination necessarily precedes division during lineage commitment, thereby providing a new dimension for modeling hematopoietic dynamics. Our study reveals that initial progenitors in adult peripheral blood exhibit self-renewal capabilities, which are absent after lineage commitment is completed during hematopoiesis. Neutrophil present the shortest differentiation trajectory in the hematopoietic hierarchy, which decide their role as rapid responders. Our results provide a clear characterization of the cell cycle kinetics of HPCs lineage commitment and cell division, facilitating advancements in inflammation and disease treatment. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=132 SRC="FIGDIR/small/590178v3_ufig1.gif" ALT="Figure 1"> View larger version (21K): org.highwire.dtl.DTLVardef@12f9c62org.highwire.dtl.DTLVardef@c3fe21org.highwire.dtl.DTLVardef@1615922org.highwire.dtl.DTLVardef@22d225_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOGraphical Abstract:C_FLOATNO Lineage commitment progresses independently of cell division, where each progenitor differentiates into a single terminally committed cell (Panel B), challenging the conventional paradigm that cell division occurs at each step of lineage commitment (A).s C_FIG

cell biology↗

bioRxiv2024

Integration of molecular coarse-grained model into geometric representation learning framework for protein-protein complex property prediction

Structure-based machine learning algorithms have been utilized to predict the properties of protein-protein interaction (PPI) complexes, such as binding affinity, which is critical for understanding biological mechanisms and disease treatments. While most existing algorithms represent PPI complex graph structures at the atom-scale or residue-scale, these representations can be computationally expensive or may not sufficiently integrate finer chemical-plausible interaction details for improving predictions. Here, we introduce MCGLPPI, a novel geometric representation learning framework that combines graph neural networks (GNNs) with the MARTINI molecular coarse-grained (CG) model to predict overall PPI properties accurately and efficiently. This framework maps proteins onto a concise CG-scale complex graph, where nodes represent CG beads and edges encode chemically plausible interactions. The GNN-based encoder is tailored to extract high-quality representations from this graph, efficiently capturing the overall properties of the protein complex structure. Extensive experiments on three different downstream PPI property prediction tasks demonstrate that MCGLPPI achieves competitive performance compared with the counterparts at the atom- and residue-scale, but with only a third of the computational resource consumption. Furthermore, the CG-scale pre-training on protein domain-domain interaction structures enhances its predictive capabilities for PPI tasks. MCGLPPI offers an effective and efficient solution for PPI overall property predictions, serving as a promising tool for the large-scale analysis of biomolecular interactions.

bioinformatics↗

bioRxiv2024

Feature Reuse and Scaling: Understanding Transfer Learning with Protein Language Models

Large pretrained protein language models (PLMs) have improved protein property and structure prediction from sequences via transfer learning, in which weights and representations from PLMs are repurposed for downstream tasks. Although PLMs have shown great promise, currently there is little understanding of how the features learned by pretraining relate to and are useful for downstream tasks. We perform a systematic analysis of transfer learning using PLMs, conducting 370 experiments across a comprehensive suite of factors including different downstream tasks, architectures, model sizes, model depths, and pretraining time. We observe that while almost all down-stream tasks do benefit from pretrained models compared to naive sequence representations, for the majority of tasks performance does not scale with pretraining, and instead relies on low-level features learned early in pretraining. Our results point to a mismatch between current PLM pretraining paradigms and most applications of these models, indicating a need for better pretraining methods.

bioengineering↗

bioRxiv2023

A Foundation Model for Cell Segmentation

Cells are a fundamental unit of biological organization, and identifying them in imaging data - cell segmentation - is a critical task for various cellular imaging experiments. While deep learning methods have led to substantial progress on this problem, most models are specialist models that work well for specific domains but cannot be applied across domains or scale well with large amounts of data. In this work, we present CellSAM, a universal model for cell segmentation that generalizes across diverse cellular imaging data. CellSAM builds on top of the Segment Anything Model (SAM) by developing a prompt engineering approach for mask generation. We train an object detector, CellFinder, to automatically detect cells and prompt SAM to generate segmentations. We show that this approach allows a single model to achieve human-level performance for segmenting images of mammalian cells, yeast, and bacteria collected across various imaging modalities. We show that CellSAM has strong zero-shot performance and can be improved with a few examples via few-shot learning. Additionally, we demonstrate how CellSAM can be applied across diverse bioimage analysis workflows. A deployed version of CellSAM is available at https://cellsam.deepcell.org/.

bioinformatics↗

bioRxiv2023

A wave of minor de novo DNA methylation initiates in mouse 8-cell embryos and co-regulates imprinted X- chromosome inactivation

DNA methylation is extensively reprogrammed during early stage of mammalian development and is essential for normal embryogenesis. It is well established that mouse embryos acquire genome-wide DNA methylation during implantation, referred to as de novo DNA methylation, from globally hypomethylated blastocysts. However, the fact that the main de novo DNA methyltransferase 3B (DNMT3B) is initially expressed as early as the 8-cell stage, contradicts the current knowledge about timing of initiation of de novo DNA methylation. Here, we reported that a previously overlooked minor wave of de novo DNA methylation initially occurs during the transition from the 8-cell to blastocyst stage, before the well-known large-scale de novo DNA methylation during implantation. Functional analyses indicated that minor de novo DNA methylation regulates proliferation, lineage differentiation and metabolic homeostasis of preimplantation embryos, and is critical for embryonic developmental potential and pregnancy outcomes. Furthermore, bioinformatic and functional analyses indicated that minor de novo DNA methylation preferentially occurs on the X chromosome and co-regulates imprinted X-chromosome inactivation via the interaction between DNMT3B and polycomb repressive complexes 2 core components during blastocyst formation. Thus, our study updates the current knowledge of embryonic de novo DNA methylation, thereby providing a novel insight of early embryonic epigenetic reprogramming. Summary statementA minor wave of de novo DNA methylation has been initiated prior to blastocyst formation, but not during the implantation period, and co-regulates imprinted X-chromosome inactivation.

developmental biology↗

bioRxiv2023

Accurate single-molecule spot detection for image-based spatial transcriptomics with weakly supervised deep learning

Image-based spatial transcriptomics methods enable transcriptome-scale gene expression measurements with spatial information but require complex, manually-tuned analysis pipelines. We present Polaris, an analysis pipeline for image-based spatial transcriptomics that combines deep learning models for cell segmentation and spot detection with a probabilistic gene decoder to quantify single-cell gene expression accurately. Polaris offers a unifying, turnkey solution for analyzing spatial transcriptomics data from MERFSIH, seqFISH, or ISS experiments. Polaris is available through the DeepCell software library (https://github.com/vanvalenlab/deepcell-spots) and https://www.deepcell.org.

bioinformatics↗

bioRxiv2023

Integrative analysis of microbial 16S gene and shotgun metagenomic sequencing data improves statistical efficiency

The most widely used technologies for profiling microbial communities are 16S marker-gene sequencing and shotgun metagenomic sequencing. Interestingly, many microbiome studies have performed both sequencing experiments on the same cohort of samples. The two sequencing datasets often reveal consistent patterns of microbial signatures, highlighting the potential for an integrative analysis to improve power of testing these signatures. However, differential experimental biases, partially overlapping samples, and differential library sizes pose tremendous challenges when combining the two datasets. Currently, researchers either discard one dataset entirely or use different datasets for different objectives. In this article, we introduce the first method of this kind, named Com-2seq, that combines the two sequencing datasets for the objective of testing differential abundance at the genus and community levels while overcoming these difficulties. We demonstrate that Com-2seq substantially improves statistical efficiency over analysis of either dataset alone and works better than two ad hoc approaches.

bioinformatics↗

bioRxiv2023

DeCOIL: Optimization of Degenerate Codon Libraries for Machine Learning-Assisted Protein Engineering

With advances in machine learning (ML)-assisted protein engineering, models based on data, biophysics, and natural evolution are being used to propose informed libraries of protein variants to explore. Synthesizing these libraries for experimental screens is a major bottleneck, as the cost of obtaining large numbers of exact gene sequences is often prohibitive. Degenerate codon (DC) libraries are a cost-effective alternative for generating combinatorial mutagenesis libraries where mutations are targeted to a handful of amino acid sites. However, existing computational methods to optimize DC libraries to include desired protein variants are not well suited to design libraries for ML-assisted protein engineering. To address these drawbacks, we present DEgenerate Codon Optimization for Informed Libraries (DeCOIL), a generalized method which directly optimizes DC libraries to be useful for protein engineering: to sample protein variants that are likely to have both high fitness and high diversity in the sequence search space. Using computational simulations and wet-lab experiments, we demonstrate that DeCOIL is effective across two specific case studies, with potential to be applied to many other use cases. DeCOIL offers several advantages over existing methods, as it is direct, easy-to-use, generalizable, and scalable. With accompanying software (https://github.com/jsunn-y/DeCOIL), DeCOIL can be readily implemented to generate desired informed libraries. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=114 SRC="FIGDIR/small/540424v1_ufig1.gif" ALT="Figure 1"> View larger version (29K): org.highwire.dtl.DTLVardef@16ce726org.highwire.dtl.DTLVardef@bb8a2aorg.highwire.dtl.DTLVardef@6be56borg.highwire.dtl.DTLVardef@1a2efb6_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioengineering↗