Biology subjects

Luo, Y.

Publications and source records attributed to Luo, Y..

At least 253 records · Page 14Linked to original sources

bioRxiv2023

Distance-dependent reconfiguration of hubs in Alzheimer's disease:a cross-tissue functional network study

The hubs of the intra-grey matter (GM) network were sensitive to anatomical distance and susceptible to neuropathological damage. However, few studies examined the hubs of cross-tissue distance-dependent networks and their changes in Alzheimers disease (AD). Using resting-state fMRI data of 30 AD patients and 37 normal older adults (NC), we constructed the cross-tissue networks based on functional connectivity (FC) between GM and white matter (WM) voxels. In the full-ranged and distance-dependent networks (characterized by gradually increased Euclidean distances between GM and WM voxels), their hubs were identified with weight degree metrics (frWD and ddWD). We compared these WD metrics between AD and NC; using the resultant abnormal WDs as the seeds, we performed seed-based FC analysis. With increasing distance, the GM hubs of distance-dependent networks moved from the medial to lateral cortices, and the WM hubs spread from the projection fibers to longitudinal fascicles. Abnormal ddWD metrics in AD were primarily located in the hubs of distance-dependent networks around 20-100mm. Decreased ddWDs were located in the left corona radiation (CR), which had decreased FCs with the executive networks GM regions in AD. Increased ddWDs were located in the posterior thalamic radiation (PTR) and the temporal-parietal-occipital junction (TPO), and their FCs were larger in AD. Increased ddWDs were shown in the sagittal striatum, which had larger FCs with the salience networks GM regions in AD. The reconfiguration of cross-tissue distance-dependent networks possibly reflected the disruption in the neural circuit of executive function and the compensatory changes in the neural circuits of visuospatial and social-emotional functions in AD.

neuroscience↗

bioRxiv2023

STdGCN: accurate cell-type deconvolution using graph convolutional networks in spatial transcriptomic data

Spatially resolved transcriptomics performs high-throughput measurement of transcriptomes while preserving spatial information about the cellular organizations. However, many spatially resolved transcriptomic technologies can only distinguish spots consisting of a mixture of cells instead of working at single-cell resolution. Here, we present STdGCN, a graph neural network model designed for cell type deconvolution of spatial transcriptomic (ST) data that can leverage abundant single-cell RNA sequencing (scRNA-seq) data as reference. STdGCN is the first model incorporating the expression profiles from single cell data as well as the spatial localization information from the ST data for cell type deconvolution. Extensive benchmarking experiments on multiple ST datasets showed that STdGCN outperformed 14 published state-of-the-art models. Applied to a human breast cancer Visium dataset, STdGCN discerned spatial distributions between stroma, lymphocytes and cancer cells for tumor microenvironment dissection. In a human heart ST dataset, STdGCN detected the changes of potential endothelial-cardiomyocyte communications during tissue development.

bioinformatics↗

bioRxiv2023

Patient and cell-type specific hiPSC-modeling of a truncating titin variant associated with atrial fibrillation

BackgroundProtein truncating mutations in the titin gene are associated with increased risk of atrial fibrillation (AF). However, little is known regarding the underlying pathophysiology. MethodsWe identified a heterozygous titin truncating variant in a patient with unexplained early-onset AF using whole exome sequencing. We used atrial and ventricular patient induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs), CRISPR/Cas9 genetic correction, and engineered heart tissue (EHT) constructs to evaluate the impact of the titin truncating variant on electrophysiology, sarcomere structure, contractility, and gene expression. ResultsWe generated atrial and ventricular iPSC-CMs from the AF patient with the titin truncating variant and a CRISPR/Cas9 genome corrected isogenic control. We demonstrate that the titin truncating variant increases susceptibility to pacing-induced arrhythmia (prevalence of arrhythmogenic phenotypes, 85.7% versus 14.2%; P = 0.03), promotes sarcomere disorganization (mean {+/-} SEM, 66.3 {+/-} 6.8% versus 88.0 {+/-} 2.9%; P = 0.04) in atrial iPSC-CMs, and reduces contractile force (0.013 {+/-} 0.003 mN versus 0.027 {+/-} 0.004 mN; P < 0.01) in atrial EHTs compared to isogenic controls. In ventricular iPSC-CMs, this variant led to altered electrophysiology (90.0% versus 33.3%; P = 0.02) and sarcomere organization (62.0 {+/-} 3.9% versus 82.9 {+/-} 2.9%; P < 0.01) with no change in EHT contractility compared to isogenic controls. RNA-sequencing revealed an upregulation of cell adhesion and extracellular matrix genes in the presence of the titin truncating variant for both atrial and ventricular EHTs. ConclusionsIn a patient with early-onset unexplained AF and normal ventricular function, iPSC-CMs with a titin truncating variant showed structural and electrophysiological abnormalities in both atrial and ventricular preparations, while only atrial EHTs demonstrated reduced contractility. Whole transcriptome sequencing showed upregulation of genes involved in cell-cell and cell-matrix interactions in both atrial and ventricular EHTs. Together, these findings suggest titin truncating variants promote the development of AF through remodeling of atrial cardiac tissue and provide insight into the chamber-specific effects of titin truncating variants.

cell biology↗

bioRxiv2023

Excessive apoptosis of naïve T cells resulting hyperactivation as a cause of mammalian aging

In mammals, the most remarkable T cell variations with aging are the shrinking of naive T cell pool and enlargement of memory T cell pool, which are partially caused by thymic involution. However, it remains an enigma whether these T-cell-related changes are consequences or causes of mammalian aging. In this study, we find that the T-cell specific Rip1 KO mice present similar age-related T cell changes and exhibit signs of accelerated aging, including inflammation, multiple age-related diseases and a shorter lifespan. Mechanistically, T cells lacking RIP1 displayed excessive apoptosis, leading to T cell compensatory proliferation, hyperactivation, increased inflammation, and ultimately premature death. Consistent with this, blocking apoptosis by co-deletion of Fadd in Rip1 deficient T cells significantly recovered the lymphopenia and imbalance between naive and memory T cell, substantially restored ageing-related phenotypes, and prolonged life span in T-cell specific Rip1 KO mice. These results suggest that changes in T cells play a causal role in mammalian aging. Therefore, replenishing or blocking apoptosis of naive T cells could offer new therapeutic approaches for aging and age-related diseases.

immunology↗

bioRxiv2023

G-quadruplexes are promoter elements controlling nucleosome exclusion and RNA polymerase II pausing

Despite their central role in transcription, it has been difficult to define universal sequences associated to eukaryotic promoters. Within chromatin context, recruitment of the transcriptional machinery requires opening of the promoter but how DNA elements could contribute to this process has remained elusive. Here, we show that G-quadruplex (G4) secondary structures are highly enriched mammalian core promoter elements. G4s are located at the deepest point of nucleosome exclusion at promoters and correlate with maximum promoter activity. We found that experimental G4s exclude nucleosomes both in vivo and in vitro and display a strong positioning potential. At model promoters, impairing G4s affected both transcriptional activity and chromatin opening. G4 destabilization also resulted in an inactive promoter state and affected transition to effective RNA production in live imaging experiments. Finally, G4 stabilization resulted in global reduction of proximal promoter pausing. Altogether, our data introduce G4s as bona fide promoter elements allowing nucleosome exclusion and facilitating pause release by the RNA Polymerase II.

genomics↗

bioRxiv2023

Lower shear stress exacerbates atherosclerosis by inducing the generation of neutrophil extracellular traps via Piezo1-mediated mechanosensation

BACKGROUNDAtherosclerosis is a chronic lipid-driven inflammatory disease, largely influenced by hemodynamics. Neutrophil extracellular traps (NETs)-mediated inflammation plays an important role in atherosclerosis. However, little is known about the mechanism of the generation of NETs under different shear stress and subsequent damage to endothelial cells. We sought to identify a novel mechanical signal provokes NETs generation and to investigate its potential role in atherosclerosis. METHODSApoE-/- mice were fed with high-fat diet (HFD) to induce atherosclerosis. The model of lower shear stress (LSS) with a partial ligation of the left carotid artery was established to assess the role of LSS in NETs generation and atherosclerotic lesions development. Furthermore, the underlying mechanism of LSS promoting NETs generation and injuring endothelial cells was deciphered in neutrophil-like human promyelocytic leukemia (HL-60) cells in parallel-plate flow chamber. RESULTSWe found that LSS correlated spatially with both NETs and atherosclerosis, while inhibition of NETosis could significantly reduce plaque formation in ApoE-/- mice. In vitro, LSS could promote NETs generation directly through down-regulation of Piezo1, a mechanosensitive ion channel. downexpression of Piezol could activate neutrophils and promote NETosis in static. Conversely, Yoda1-evoked activation of Piezo1 attenuated LSS-induced NETosis. Mechanistically, the downexpression of Piezo1 resulted in decreased Ca2+ influx and increased histone deacetylase 2 (HDAC2), which increase reactive oxygen species levels, then led to NETosis. LSS-induced NETs generation promoted the apoptosis and adherence of endothelial cells. CONCLUSIONSLSS directly promotes NETosis through piezo1-HDAC2 axis in atherosclerosis progression. This study uncovers the essential role of Piezo1-mediated mechanical signaling in NETs generation and plaque formation, which provides a promising therapeutic strategy for atherosclerosis. Graphic Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=131 SRC="FIGDIR/small/529165v1_ufig1.gif" ALT="Figure 1"> View larger version (52K): org.highwire.dtl.DTLVardef@583911org.highwire.dtl.DTLVardef@aa3e2org.highwire.dtl.DTLVardef@8d83e5org.highwire.dtl.DTLVardef@ce55a0_HPS_FORMAT_FIGEXP M_FIG C_FIG Proposed mechanism for lower shear stress LSS exacerbating atherosclerosis. LSS stimuli decrease Piezo1 expression in the neutrophils, resulting in decreased intracellular Ca2+ concentration, as well as the higher expression level of HDAC2, which could activate oxidative stress and promote intracellular reactive oxygen species formation, and ultimately lead to NETs generation. NETs could aggravate endothelial cells injury and exasperate atherosclerosis. Highlights{blacksquare} Lower shear stress (LSS) promotes Neutrophil extracellular traps (NETs) formation, which is critical for lipid deposits and plaque formation in Atherosclerosis. {blacksquare}Atherosclerotic plaque formation was significantly reduced in the aorta of high fat diet fed ApoE-/- mice intraperitoneal injected with NETs inhibitor, GSK484, especially in the lower shear stress regions. {blacksquare}Piezo1 is a key molecule in the process of neutrophils sense lower shear stress. {blacksquare}lower shear stress inhibits the activation of Piezo1 and promotes NETosis through piezo1-HDAC2 axis. {blacksquare}LSS-induced NETs promote the apoptosis and adhesion of endothelial cells.

cell biology↗

bioRxiv2023

mRNA interactions with disordered regions control protein activity

The cytoplasm is compartmentalized into different translation environments. mRNAs use their 3'UTRs to localize to distinct cytoplasmic compartments, including TIS granules (TGs). Many transcription factors, including MYC, are translated in TGs. It was shown that translation of proteins in TGs enables the formation of protein complexes that cannot be established when these proteins are translated in the cytosol, but the mechanism is poorly understood. Here we show that MYC protein complexes that involve binding to the intrinsically disordered region (IDR) of MYC are only formed when MYC is translated in TGs. TG-dependent protein complexes require TG-enriched mRNAs for assembly. These mRNAs bind to a new and widespread RNA-binding domain in neutral or negatively charged IDRs in several transcription factors, including MYC. RNA-IDR interaction changes the conformational ensemble of the IDR, enabling the formation of MYC protein complexes that act in the nucleus and control functions that cannot be accomplished by cytosolically-translated MYC. We propose that certain mRNAs have IDR chaperone activity as they control IDR conformations. In addition to post-translational modifications, we found a novel mode of protein activity regulation. Since RNA-IDR interactions are prevalent, we suggest that mRNA-dependent control of protein functional states is widespread.

molecular biology↗

bioRxiv2023

Inhibiting Bet1-mediated transport of MMP14 to plasma membrane impaired GBM cell invasion

PurposeGlioblastoma (GBM) is the most aggressive and common form of brain cancer in adults. GBM is characterised by poor survival as the lack of effective therapies. This research aims to detect the roles of SNAREs in GBM and improve our knowledge of targeting therapy for GBM. Materials and methodsthe expression of SNAREs and their correlation with overall survival (OS) in GBM are investigated using the GEPIA. The level of BET1 in GBM cell lines was tested by RT-qPCR, and its biological functions in GBM cells were tested by Transwell assay and CCK8 kit. The effect of BET1 on the location of MMP14 is identified by Immunofluorescence. ResultsThe expression profile of SNARE family members in GBM tissue is changed dramatically. Among them, the mRNA levels of BET1 and VAMP3 are up-regulated, and their expression negatively correlates with OS. BET1 is also increased in GBM Cell Lines, and it is required for efficient GBM cell migration and invasion partly because it mediates the transport of MMP14 to the plasma membrane. ConclusionGBM has highly diffusive and infiltrative ability in nature, making complete surgical resection almost impossible. Our data shows that BET1 is highly expressed in GBM tissue, negatively correlated with OS, and essential for GBM cell migration and invasion. These results indicate that SNARE BET1 may present a potential target for GBM treatment.

cell biology↗

bioRxiv2023

Learning virus genotype-fitness landscape in embedding space

The relationship between genotype and fitness is fundamental to evolution, but quantitatively mapping genotypes to fitness has remained challenging. We propose the Phenotypic-Embedding theorem (P-E theorem) that bridges genotype-phenotype through an encoder-decoder deep learning framework. Inspired by this, we proposed a more general first principle for correlating genotype-phenotype, and the Phenotypic-Embedding theorem provides a computable basis for the application of first principle. As an application example of the P-E theorem, we developed the Co-attention based Transformer model to bridge Genotype and Fitness (CoT2G-F) model, a Transformer-based pre-train foundation model with downstream supervised fine-tuning (SFT) that can accurately simulate the neutral evolution of viruses and predict immune escape mutations. Accordingly, following the calculation path of the P-E theorem, we accurately obtained the basic reproduction number (R0) of SARS-CoV-2 from first principles, quantitatively linked immune escape to viral fitness, and plotted the genotype-fitness landscape. The theoretical system we established provides a general and interpretable method to construct genotype-phenotype landscapes, providing a new paradigm for studying theoretical and computational biology.

evolutionary biology↗

bioRxiv2023

xTrimoDock: Rigid Protein Docking via Cross-Modal Representation Learning and Spectral Algorithm

Protein-protein interactions are the basis for the formation of protein complexes which are essential for almost all cellular processes. Knowledge of the structures of protein complexes is of major importance for understanding the biological function of these protein-protein interactions and designing protein drugs. Here we address the problem of rigid protein docking which assumes no deformation of the involved proteins during interactions. We develop a method called, xTrimoDock, which leverages a cross-modal representation learning to predict the protein distance map, and then uses a spectral initialization and gradient descent to obtain the roto-translation transformation for docking. We show that, on antibody heavy-chain and light-chain docking, and antibody-antigen docking, xTrimoDock consistently outperforms the state-of-the-art such as AlphaFold-Multimer and HDock, and can lead to as much as a 10% improvement in DockQ metric. xTrimoDock has been applied as a useful tool in protein drug design at BioMap.

bioinformatics↗

bioRxiv2023

Resistance to vincristine in cancerous B-cells by disruption of p53-dependent mitotic surveillance

The frontline therapy R-CHOP for patients with diffuse large B-cell lymphoma (DLBCL) has remained unchanged for two decades despite numerous phase III clinical trials investigating new alternatives. Multiple large studies have uncovered genetic subtypes of DLBCL enabling a targeted approach. To further pave the way for precision oncology, we perform genome-wide CRISPR screening to uncover the cellular response to one of the components of R-CHOP, vincristine, in the DLBCL cell line SU-DHL-5. We discover important pathways and subnetworks using gene-set enrichment analysis and protein-protein interaction networks and identify genes related to mitotic spindle organization that are essential during vincristine treatment. Inhibition of KIF18A, a mediator of chromosome alignment, using the small molecule inhibitor BTB-1 causes complete cell death in a synergistic manner when administered together with vincristine. We also identify the genes KIF18B and USP28 for which CRISPR/Cas9-directed knockout induces vincristine resistance across two DLBCL cell lines. Mechanistic studies show that lack of KIF18B or USP28 counteracts a vincristine-induced p53 response involving the mitotic surveillance pathway (USP28-53BP1-p53). Collectively, our CRISPR screening data uncover potential drug targets and mechanisms behind vincristine resistance, which may support the development of future drug regimens. Key pointsO_LIInhibition of the mitotic surveillance pathway (USP28-53BP1-p53) and KIF18B induces resistance to vincristine C_LIO_LISubstantial synergistic effects observed when using the KIF18A-inhibitor BTB-1 with vincristine in eradicating GCB-subtype DLBCL cells C_LI

cancer biology↗

bioRxiv2023

Innate immunity can distinguish beneficial from pathogenic rhizosphere microbiota

For optimal growth and development, hosts depend on their ability to promote healthy symbiotic interactions while restricting pathogen growth. To ask whether hosts can distinguish phylogenetically similar pathogens and beneficial bacteria, we used two closely related plant-associated strains of Pseudomonas fluorescens where one is beneficial and the other exhibits toxin-dependent virulence. We show that while the two strains co-exist in vitro, the beneficial outcompetes that pathogen in planta. Using several readouts for plant innate immunity, we found that the beneficial and pathogenic strains elicit mechanistically distinct immune responses that occur in distinct root compartments. We show that while both the pathogenic and beneficial bacterial have plant recognizable MAMPs, the pathogen uniquely induces MAMP-independent immune responses. We found that the pathogen induces both a toxin-independent and a unique toxin-dependent defense response that remains intact in immune signaling mutants including bak1/bkk1 and npr1/4D. We conclude that hosts can distinguish between phylogenetically similar microbes.

plant biology↗

bioRxiv2023

Crystal structure of a mammalian CMTM6 and its interaction model with PD-L1

CKLF-like MARVEL transmembrane domain-containing protein 6 (CMTM6) is a master regulator of PD-L1. By binding PD-L1 at the plasma membrane and recycling endosomes, CMTM6 prevents the lysosomal degradation of PD-L1 and maintains its cell surface expression, thus stabilizing the inhibitory PD-1/PD-L1 axis. However, the mechanism of CMTM6/PD-L1 interaction is unknown. Here we report the first experimentally determined structure of CMTM6 from bovine. Combined with a low-resolution cryo-EM map, computational docking analysis and a protein binding assay an interaction model between CMTM6 and PD-L1 was proposed, providing a structural framework for the CMTM6 regulation on PD-L1.

biophysics↗

bioRxiv2022

Comparing scientific abstracts generated by ChatGPT to original abstracts using an artificial intelligence output detector, plagiarism detector, and blinded human reviewers

BackgroundLarge language models such as ChatGPT can produce increasingly realistic text, with unknown information on the accuracy and integrity of using these models in scientific writing. MethodsWe gathered ten research abstracts from five high impact factor medical journals (n=50) and asked ChatGPT to generate research abstracts based on their titles and journals. We evaluated the abstracts using an artificial intelligence (AI) output detector, plagiarism detector, and had blinded human reviewers try to distinguish whether abstracts were original or generated. ResultsAll ChatGPT-generated abstracts were written clearly but only 8% correctly followed the specific journals formatting requirements. Most generated abstracts were detected using the AI output detector, with scores (higher meaning more likely to be generated) of median [interquartile range] of 99.98% [12.73, 99.98] compared with very low probability of AI-generated output in the original abstracts of 0.02% [0.02, 0.09]. The AUROC of the AI output detector was 0.94. Generated abstracts scored very high on originality using the plagiarism detector (100% [100, 100] originality). Generated abstracts had a similar patient cohort size as original abstracts, though the exact numbers were fabricated. When given a mixture of original and general abstracts, blinded human reviewers correctly identified 68% of generated abstracts as being generated by ChatGPT, but incorrectly identified 14% of original abstracts as being generated. Reviewers indicated that it was surprisingly difficult to differentiate between the two, but that the generated abstracts were vaguer and had a formulaic feel to the writing. ConclusionChatGPT writes believable scientific abstracts, though with completely generated data. These are original without any plagiarism detected but are often identifiable using an AI output detector and skeptical human reviewers. Abstract evaluation for journals and medical conferences must adapt policy and practice to maintain rigorous scientific standards; we suggest inclusion of AI output detectors in the editorial process and clear disclosure if these technologies are used. The boundaries of ethical and acceptable use of large language models to help scientific writing remain to be determined.

scientific communication and education↗

bioRxiv2022

Multi-center integrated analysis of non-coding CRISPR screens

The ENCODE Consortiums efforts to annotate non-coding, cis-regulatory elements (CREs) have advanced our understanding of gene regulatory landscapes which play a major role in health and disease. Pooled, non-coding CRISPR screens are a promising approach for systematically investigating gene regulatory mechanisms. Here, the ENCODE Functional Characterization Centers report 109 screens comprising 346,970 individual perturbations across 13.3Mb of the genome, using a variety of methods, readouts, and statistical analyses. Across 332 functionally confirmed CRE-gene links, we identify principles for screening endogenous, non-coding elements for causal regulatory mechanisms. Nearly all CREs show strong evidence of open chromatin, and targeting accessibility peak summits is a critical component of our proposed sgRNA design rules. We provide experimental guidelines to accurately detect CREs with variable, often low, transcriptional effects. We discover a previously undescribed DNA strand-bias for CRISPRi in transcribed regions with implications for screen design and analysis. Benchmarking five screen analysis tools, we find CASA produces the most conservative CRE calls and is robust to artifacts of low-specificity sgRNAs. Together, we provide an accessible data resource, predesigned sgRNAs targeting 3,275,697 ENCODE SCREEN candidate CREs, and screening guidelines to accelerate functional characterization of the non-coding genome.

genomics↗

bioRxiv2022

Regulation of Nuclear Transcription by Mitochondrial RNA

Chromatin-associated RNAs (caRNAs) form a relatively poorly recognized layer of the epigenome. The caRNAs reported to date are transcribed from the nuclear genome. Here, leveraging a recently developed assay for detection of caRNAs and their genomic association, we report that mitochondrial RNAs (mtRNAs) are attached to the nuclear genome and constitute a subset of caRNA, which we termed mt-caRNA. In four human cell types analyzed, mt-caRNAs preferentially attach to promoter regions. In human endothelial cells (ECs), the level of mt-caRNA-promoter attachment changes in response to environmental stress that mimics diabetes. Suppression of a non-coding mt-caRNA in ECs attenuates stress-induced nascent RNA transcription from the nuclear genome, including that of critical genes regulating cell adhesion, and abolishes stress-induced monocyte adhesion, a hallmark of dysfunctional ECs. Finally, we report increased nuclear localization of multiple mtRNAs in the ECs of human diabetic donors, suggesting many mtRNA translocate to the nucleus in a cell stress and disease-dependent manner. These data nominate mt-caRNAs as messenger molecules responsible for mitochondrial-nuclear communication and connect the immediate product of mitochondrial transcription with the transcriptional regulation of the nuclear genome.

genomics↗

bioRxiv2022

A Unitary Mechanism Underlies Adaptation to Both Local and Global Environmental Statistics in Time Perception

Our sense of the passage of time flexibly adapts to the statistical properties of the temporal context. Humans and non-human species exhibit a perceptual bias towards the mean of durations previously observed as well as serial dependence, a perceptual bias towards the duration of recently processed events. Here we asked whether those two phenomena arise from a unitary mechanism or reflect the operation of two distinct systems that adapt separately to the global and local statistics of the environment. We employed a set of duration reproduction tasks in which the target duration was sampled from distributions with different variances and means. The central tendency and serial dependence biases were jointly modulated by the range and the variance of the prior. These effects were well-captured by a unitary mechanism model in which temporal expectancies are updated after each trial based on perceptual observations. Alternative models that assume separate mechanisms for global and local contextual effects failed to capture the empirical results. TeaserTime perception of humans is shaped by a common mechanism that is sensitive to short-term and long-term environmental changes.

neuroscience↗

bioRxiv2022

Contrastive learning of protein representations with graph neural networks for structural and functional annotations

Although protein sequence data is growing at an ever-increasing rate, the protein universe is still sparsely annotated with functional and structural annotations. Computational approaches have become efficient solutions to infer annotations for unlabeled proteins by transferring knowledge from proteins with experimental annotations. Despite the increasing availability of protein structure data and the high coverage of high-quality predicted structures, e.g., by AlphaFold, many existing computational tools still only rely on sequence data to predict structural or functional annotations, including alignment algorithms such as BLAST and several sequence-based deep learning models. Here, we develop PenLight, a general deep learning framework for protein structural and functional annotations. PenLight uses a graph neural network (GNN) to integrate 3D protein structure data and protein language model representations. In addition, PenLight applies a contrastive learning strategy to train the GNN for learning protein representations that reflect similarities beyond sequence identity, such as semantic similarities in the function or structure space. We benchmarked PenLight on a structural classification task and a functional annotation task, where PenLight achieved higher prediction accuracy and coverage than state-of-the-art methods.

bioinformatics↗