Biology subjects

Li, C.

Publications and source records attributed to Li, C..

At least 217 records · Page 12Linked to original sources

bioRxiv2025

Bio-synthesis of 15N-Labeled G-Quadruplexes to Investigate the Structure and Interactions in the Cell Lysate using Nuclear Magnetic Resonance

G-quadruplexes (G4) play key roles in biology, making it critical to understand their structure and ligand-binding behavior in cellular environments for advancing G4-targeted therapeutics. While in-cell nuclear magnetic resonance (NMR) is a powerful technique for studying G4 in situ, its application is limited by the challenge of producing isotope-labeled single-stranded DNA (ssDNA). Here, we introduce Restriction Endonuclease Digestion (RED), a simple and cost-effective method to generate 15N-labeled ssDNA. This approach combines molecular cloning and enzymatic design processing by propagating plasmids in E. coli cultured with 15NH4Cl, followed by double restriction digestion and isolation target 15N-ssDNA. Using RED, we produced milligram-scale quantities of 96%-enriched 15N-labeled human telomeric G4 ssDNA (wtTel23c, CTAGGG(TTAGGG)3), ideal for NMR analysis. The NMR spectra revealed that wtTel23c adopts G4 topology and undergoes multiple conformations of wtTel23c in potassium-containing solutions and in Xenopus laevis cell lysate. Interaction studies with the ligand TMPyP4 showed distinct binding profiles in cellular and dilute environments. In dilute solution, TMPyP4 binds to the top tetrad of wtTel23c, while it binds to the loop in cellular environments. The RED method offers an efficient strategy for producing stable isotope-labeled ssDNA, opening new avenues for studying G4 structures and their ligand interactions in complex biological contexts.

biochemistry↗

bioRxiv2025

A novel 18F-labeled brain penetrant PET ligand for imaging poly(ADP-ribose) polymerase-1

Poly(ADP-ribose) polymerase-1 (PARP-1) is a key mediator of DNA repair, and its inhibition has become a validated therapeutic strategy in homologous recombination-deficient cancers. However, tools for non-invasive assessment of PARP-1-specific expression remain limited. Here, we evaluated [18F]AZD9574, a next-generation PARP-1-selective PET radiotracer. [18F]AZD9574 was tested in a panel of breast, glioblastoma, prostate, and pancreatic cancer cell lines. Uptake correlated with PARP-1 expression and was dose-dependently blocked by a variety of clinically relevant PARP inhibitors, confirming its binding specificity. In 22Rv1 xenograft mouse models, the tracer demonstrated significant tumor accumulation that was specific to PARP-1, and ex vivo biodistribution confirmed organ uptake consistent with specific tumor binding and PARP-1 expression. Together, these findings establish [18F]AZD9574 as a promising PARP-1-targeted imaging agent with strong potential for advancing cancer research and therapeutic monitoring.

cancer biology↗

bioRxiv2025

An ELISA for discovering protein-protein interaction inhibitors: blocking lysinoalanine crosslinking between subunits of the spirochete flagellar hook as a test case

The inhibition of a specific protein-protein interaction is often difficult to achieve in targeted drug design. We report the development and optimization of a general-purpose, readily implemented enzyme-linked immunosorbent assay (ELISA) for high-throughput screening to identify small- molecule inhibitors of protein interactions. This ELISA does not involve the use of any capture antibodies, probes, or compounds coated on the plate and represents a general strategy to identify inhibitors of a given protein-protein interaction. We demonstrate its utility in blocking lysinoalanine crosslinking between subunits of the spirochete flagellar hook by targeting the native form of the FlgE protein, which differs from the strategies used in previous assays. The flagellar hook protein FlgE self-catalyzes the formation of a lysinoalanine (Lal) inter-subunit crosslink that is essential for the motility, and thus, infectivity of spirochetes. Prevention of Lal crosslinking through inhibition with small molecules thus represents an avenue for therapeutic development against spirochete-related diseases, such as Lyme and syphilis. Screening a library of [~]700 compounds with the ELISA confirmed that hexachlorophene, currently the only known inhibitor of Lal crosslinking in FlgE, effectively inhibits the crosslinking reaction. In addition, the assay identified two new potential inhibitors, honokiol and zafirlukast, and several activators which belong to well-known classes of antibiotics.

biochemistry↗

bioRxiv2025

NF-κB restrains nutrient-dependent transcription programs through chromatin modulation in Drosophila

The co-evolution of immune and metabolic systems has endowed immune signaling pathways with distinct control of cellular metabolism. Innate immune transcription factors, such as nuclear factor {kappa}B (NF-{kappa}B), have thus emerged as key regulators of adaptive metabolic responses to changes in diet and nutrition. Utilizing chromatin accessibility genomics, we found that Drosophila NF-{kappa}B (Relish) can restrain nutrient-dependent metabolic transcriptional programs that control cellular catabolism of energy substrates, divergent from the proteins canonical role as a transcriptional activator. NF-{kappa}B/Relish restricts chromatin accessibility through modulating histone acetylation at metabolic target gene loci, which restrains metabolic gene transcription and blocks excessive activation of nutrient-dependent metabolic programs. Targeted genetic screening revealed that histone deacetylase 6 (HDAC6) interacts with NF-{kappa}B/Relish at NF-{kappa}B DNA regulatory motifs to limit chromatin accessibility and repress metabolic transcriptional programs. These results highlight that innate immune transcription factors can epigenetically restrain cellular catabolism to fine-tune nutrient-dependent metabolic adaptation.

physiology↗

bioRxiv2025

Global shifts in vegetation compositional resilience over the past 8,000 years

Vegetation resilience is critical for sustaining biodiversity and ecosystem services, yet its long-term dynamics and underlying drivers remain poorly resolved. Here, we assess changes in continental-scale vegetation compositional resilience over the past 8,000 years by applying established leading indicators (including autocorrelation, standard deviation, skewness, and kurtosis) to 482 fossil pollen records from all major landmasses except Antarctica. Using a machine learning approach, we evaluate the relative contributions of anthropogenic land cover change (ALCC) and climatic parameters (annual mean temperature, temperature seasonality, annual precipitation, and precipitation seasonality) in driving resilience shifts across continents. We further employ structural equation modeling (SEM) to disentangle how abiotic factors (ALCC and climatic variables) and biotic factors (species richness, evenness, synchrony, and temporal {beta} diversity) jointly shape long-term trends in vegetation resilience at the continental scale. Our results reveal a persistent, millennia-scale decline in resilience during the past 1.6 to 4.4 thousand years before present, with intensified anthropogenic land-use change identified as the dominant global driver. This finding suggests that the global decline in vegetation resilience observed over recent decades likely represents a continuation of the long-term trend established during the preceding Holocene. Notably, North America exhibited a resurgence in resilience over the past 1,200 years, primarily attributed to enhanced resilience in tundra and savanna ecosystems. SEM analyses uncover complex pathways through which abiotic and biotic predictors interact to shape millennial-scale resilience changes across continents. Despite the complexity of these underlying mechanisms, the analyses demonstrate that biotic factors collectively exert substantial, and often predominant, influences on long-term resilience dynamics. This underscores the indispensable role of biotic attributes in modulating long-term vegetation resilience, highlighting their potential as critical targets for ecosystem conservation and restoration.

ecology↗

bioRxiv2025

Dedifferentiation-Driven Oncogenic Stemness Promotes Tumor-Sustaining Adaptability in the Intestinal Epithelium

Intestinal tumorigenesis can occur via two distinct routes: bottom-up tumorigenesis occurs from mutations sustained in the Lgr5 stem cells, whereas top-down tumorigenesis is driven by dedifferentiation of epithelial cells near the intestinal lumen. While sporadic human colon adenomas exhibit features of top-down tumorigenesis, their biological determinants remain elusive. Here, using a Smad4 loss-of-function and {beta}-catenin gain-of-function (Smad4LOF:{beta}-cateninGOF) mouse model, we demonstrate that dedifferentiation-derived oncogenic stem cells sustain tumorigenesis more effectively than endogenous mutant stem cells harboring the mutation. The dedifferentiating villi epithelial cells showed early expression of CD44 and Lgr5, supporting oncogenic stemness. Aberrant Notch signaling in the villi epithelium was also detected at the onset of dedifferentiation, suggesting its contribution to dedifferentiation. Single-cell RNA sequencing revealed a distinct population of dedifferentiation-derived stem cells enriched for proliferative, metabolic, and mouse embryonic stem cell-like gene signatures, consistent with enhanced plasticity and tumorigenic potential. These mutant cells exhibited growth factor independence, indicating a capacity for niche-independent proliferation and metabolic adaptation to sustain tumor growth. These findings identify dedifferentiation-driven stemness, aberrant Notch activation, and metabolic plasticity as cooperative mechanisms that promote top-down intestinal tumorigenesis. This study provides insight into how oncogenic dedifferentiation contributes to tumor heterogeneity and persistence and has implications for therapeutic resistance in colorectal cancer.

cancer biology↗

bioRxiv2025

Aβ-42 sidechain deamidation at Q15 and N27 modulate protein aggregation and microglial responses via altered cytokine production and CD68 expression

The progressive aggregation of amyloid beta (A{beta}) monomers into oligomers is a critical factor in Alzheimers disease (AD) pathogenesis. Although mutated forms of A{beta} have been shown to display altered aggregation dynamics, the specific effects of deamidated A{beta} on microglial function remain understudied. Our research group previously found that the deamidated variant A{beta}-42-N27D modified A{beta} aggregation, reduced neurotoxicity, and reduced microglial reactivity, but the impact of A{beta}-42 side chain deamidation in general on such parameters remained unclear. Here, we expanded on our prior work by investigating how two site-specific A{beta}-42 mutations (Q15E & N27D), where neutral amide side chains are replaced with negatively charged carboxylic acids, affect aggregation and microglial immune response using a mouse microglial cell line. Size exclusion chromatography revealed that A{beta}-42-Q15E and A{beta}-42-N27D exhibit distinct aggregation profiles compared to A{beta}-42 wild type (WT). Multiplexed analysis of 8 cytokines secreted into the culture medium revealed that A{beta}-42-Q15E and A{beta}-42-N27D decrease the expression of inflammatory cytokines such as IL-6, IP-10, and MIP-1 relative to A{beta}-42-WT. Immunocytochemistry revealed that A{beta}-42-Q15E and A{beta}-42-N27D decrease CD68 expression relative to A{beta}-42-WT. These findings demonstrate that deamidation significantly alters A{beta}-42 aggregation and microglial activation, suggesting structural modifications to A{beta}-42 modulate inflammatory signaling in AD. This work provides a foundation for future studies on A{beta}-42 post-translational modifications as potential therapeutic targets in AD.S

neuroscience↗

bioRxiv2025

Coordination of local and global responses in individual mouse thalamic local interneurons

Inhibitory interneurons, featuring presynaptic release sites and spiking activity in dendrites, present a puzzle in their signal organization and coordination. We investigated this in local interneurons (LINs) of the primary visual thalamus using two-photon calcium imaging in awake mice with both local and full-screen visual stimuli. Dendrites in single LINs showed consistent global responses to full-screen stimuli but varied responses to local stimuli. Dendrites from the same LIN exhibited different receptive fields that followed retinotopic progression. The similarity between dendritic and somatic responses decreased with increasing distance. Weak local stimuli could excite dendrites locally, while salient local stimuli triggered global responses. Salient global stimuli further enhanced motion selectivity, preference consistency, and peak responses. Thus, LINs leverage both local and global dendritic activation to perform spatially-organized, feature-selective, and size-dependent motion processing, providing new insights into the complexity and precision of sensory processing in the early visual system. HighlightsO_LILocal interneurons (LINs) exhibit diverse yet sharp tuning to moving stimuli. C_LIO_LIdisplay an orderly shift of receptive field centers along their dendrites. C_LIO_LIThe propagation of calcium signals under local stimuli depends on stimulus strength. C_LIO_LIFull-screen stimuli enhance motion selectivity and peak responses. C_LI

neuroscience↗

bioRxiv2025

Ethylene receptors are functionally conserved in calcium permeability across the green lineage

The gaseous hormone ethylene plays a key role in regulating plant growth and stress responses. Although Ca{superscript 2} has long been implicated in ethylene signaling, the identity of molecules controlling Ca{superscript 2} permeability involved has remained elusive. Here we show that Arabidopsis ethylene receptors ETR1 and ERS1 are Ca2+ permeable, and that ETR1 mediates ethylene-induced cytosolic Ca2+ increase and regulates hypocotyl elongation. Homologs of ETR1 from eight land plant and algal species also exhibit Ca{superscript 2} permeability, suggesting an evolutionarily conserved function. We further demonstrate ethylene enhances the Ca2+ permeability of ETR1 and its homologue from the charophyte Klebsormidium flaccidum. These findings uncover a previously unrecognized but conserved role of ethylene receptors in Ca{superscript 2} permeability regulation in the green lineage, with broad implications for Ca{superscript 2} signaling in plant development and environmental adaptation.

plant biology↗

bioRxiv2025

Rice Ethylene Receptors Function as Ca2+-Permeable Ion Channels to Orchestrate Calcium-Dependent Antagonism of Ethylene Responses in Roots

The gaseous phytohormone ethylene governs essential plant processes spanning development, productivity, and stress resistance. As an essential nutrient and second messenger, calcium (Ca2+) is widely implicated in diverse plant physiological activities; however, its role in the ethylene signal transduction pathway remains elusive. Here, we identified a calcium-dependent antagonism of ethylene response (CAER) specifically modulates root elongation in the model cereal rice (Oryza sativa). Interestingly, we demonstrate that the ethylene receptors OsERS1 and OsERS2 both function as Ca2+-permeable channels, with OsERS1 further acting as a non-selective cation channel capable of permeating both monovalent and divalent cations. Mutagenesis analysis reveals that OsERS1 channel activity relies on homomeric assembly sites (C4/C6) rather than its ethylene-binding site (C65), indicating a clear decoupling of the molecular modules governing receptor signaling and ion channel function. The loss-of-function mutant Osers1/2 does not exhibit the CAER phenotype of the wild type (WT), confirm that this calcium-dependent regulatory mechanism is dependent on both OsERS1 and OsERS2. These findings uncover an unexpected ion-channel function of ethylene receptors, redefining their molecular identity beyond canonical signaling proteins and establishing the novel "hormone receptor-type ion channel (HRIC)" paradigm--one that fundamentally expands our understanding of how plant hormones transduce signals at the membrane.

developmental biology↗

bioRxiv2025

O-GlcNAc clusters attenuate BRD4 phase separation for transcription regulation

Post-translational modifications (PTMs) regulate the liquid-liquid phase separation (LLPS) that organizes biomolecular condensates essential for transcription, yet for the monosaccharide O-GlcNAc, the general rules and their links to protein function remain undefined, with evidence largely limited to isolated cases rather than systematic analyses. Here we combine proteome-scale analyses with experiments to show that O-GlcNAc is a widespread LLPS modulator: glycosites are enriched in intrinsically disordered regions and form dense O-GlcNAc clusters. Using BRD4 as a model, we identified twelve clustered sites in its C-terminal IDR and showed that O-GlcNAc decreases condensate size while increasing fluidity in vitro and in cells. Removing O-GlcNAc strengthened BRD4 binding at active enhancers and promoted LLPS-mediated co-recruitment of transcriptional factors, including YTHDC1, leading to altered expression programs linked to the cell cycle and DNA repair. Our findings define O-GlcNAc clusters as regulators of condensate material properties and transcriptional outcomes, supporting a general paradigm in which PTMs fine-tune the molecular grammar of biomolecular condensates.

biochemistry↗

bioRxiv2025

Single-cell RNA sequencing and large-scale bulk combination with machine learning reveal gastric cancer-related macrophage heterogeneity

BackgroundThe tumor microenvironment (TME) significantly impacts cancer progression and overall patient survival. However, the complexity of tumor cell-TME interactions in gastric cancer (GC) and their underlying molecular basis remain to be systematically elucidated. MethodsWe performed single-cell RNA sequencing (scRNA-seq) on paired tumor and adjacent tissues from 3 treatment-naive GC patients. We conducted an in-depth characterization of the cellular composition and molecular features of the GC TME, with a particular focus on tumor-associated macrophages (TAMs) subsets and their mediated intercellular communication. Investigating the effect of C1q on the polarisation of THP-1 induced macrophages through migration and invasion experiments ResultsTranscriptomic analysis of 56,151 single cells identified a key TAM subset - C1QA{square} TAMs. This subset is characterized by high expression of genes including C1QA, C1QB, C1QC and FN1, exhibiting a distinct M2-like macrophage phenotype. We developed a LASSO-based predictive model, C1Q-LASSO, to accurately stratify patients based on survival outcomes and chemotherapy responses, independently of established prognostic parameters. In vitro functional experiments further confirmed that C1q directly promotes malignant phenotypes in GC cells. Additionally, C1q significantly enhances the malignant progression of gastric cancer induced by M2 macrophages. ConclusionsUtilizing scRNA-seq technology, this study systematically delineated the heterogeneous landscape of the GC immune microenvironment and its complex cellular interaction network at single-cell resolution. We first identified and functionally validated the C1QA{square} TAM subset, demonstrating its association with poor prognosis and pro-tumorigenic functions.

cancer biology↗

bioRxiv2025

YAP/TEAD4-reagulated placental alkaline phosphatases ALPP and ALPPL2 are immunosuppressive ectonucleotidases modulated by MAPK inhibitors

Placental alkaline phosphatases ALPP and ALPPL2 (ALPP/L2) are promising clinical targets for therapies such as antibody-drug conjugates (ADCs). However, their regulation and biological functions remain unclear. Here, we identify the YAP/TEAD4 signaling axis as the primary regulator of ALPP/L2 expression in tumor cells. Importantly, we demonstrate that inhibitors targeting KRAS, MEK, and ERK upregulate ALPP/L2 expression in vitro and in vivo, thereby enhancing the efficacy of antibody-drug conjugates (ADCs) when used in combination. Moreover, our work reveals a novel immunomodulatory function of ALPP/L2 as cell surface ectonucleotidases that hydrolyze extracellular ATP, ADP, and AMP to adenosine, promoting regulatory T cell infiltration into the tumor microenvironment. Together, these findings reveal a biological function of ALPP/L2 and identify druggable regulators to enhance ALPP/L2-targeted therapies. SIGNIFICANCEThis study uncovers the YAP/TEAD pathway as a positive regulator of the clinically relevant tumor-associated antigens ALPP and ALPPL2. Kras, MEK, or ERK inhibitors enhance ALPP/L2 expression in tumor cells, thereby increasing the efficacy of ALPP/L2-targeted antibody-drug conjugates. These findings offer a promising combination strategy to overcome resistance to ALPP/L2-based therapies caused by low target expression.

cancer biology↗

bioRxiv2025

A Population-scale Single-cell Spatial Transcriptomic Atlas of the Human Cortex

The genetic and spatial determinants of cell type diversity in human cerebral cortices remain poorly defined. Here, we present a population-level single-cell spatial transcriptomic atlas of human cortices from 71 donors across the lifespan. We identified 906 layer-specific genes showing conserved and divergent laminar expression patterns between humans and other species. Spatial analysis revealed neuronal vulnerability and glial activation during aging, together with a decline in the proportion of superficial SST neurons and their interactions with other cells. Disease-associated genes exhibited high cell-type and layer-specific expression, implicating the pathogenic role of spatially specific gene expression. Spatial cis-eQTL analysis identified regulatory variants linked to genes related to diseases like Tourette syndrome. Cross-species comparison demonstrated glial expansion in the human cortex, accompanied by enhanced neuron-glia communication via the neuregulin signaling. Together, we provide a comprehensive single-cell atlas of the human cortex that is essential for understanding aging, evolution, and disease pathogenesis.

neuroscience↗

bioRxiv2025

Non-Carbohydrate Inhibitors of Sialic Acid-binding Immunomodulatory-type Lectin-7 (Siglec-7) Discovered from Genetically Encoded Bicyclic Peptide Libraries

Glycan-binding proteins (GBP) are among the most difficult to drug targets. This deficiency delays clinical progress for therapeutically important GBPs. We employed bicyclic genetically encoded libraries (BiGELs), produced by chemical modification of phage-displayed libraries of peptides with two-fold symmetric linchpins, to discover inhibitors of therapeutically relevant Siglec-7:GD3 interactions. Next-generation sequencing (NGS) analysis of panning of BiGEL against Siglec-7 yielded 815 candidates from which 23 hits yielded KD = 1-100 {micro}M as determined by surface plasmon resonance (SPR). Competitive enzyme-linked immunosorbent assays (ELISA) identified a subset of leads that disrupted the Siglec-7:GD3 interaction with IC50= 3-300 {micro}M. Machine learning models trained on NGS datasets identified additional inhibitors with equivalent potency. Alanine scans of 8c (SWCRPATVNC, IC50 = 3.8 {micro}M) and 12c (SFCHYPTHVC, IC50= 11 {micro}M), identified key residues as crucial for activity. Ring reshaping studies of compound 8c highlighted the critical role of bicyclic topology produced by analogue 46e (SAAAAAWCRPATVNC, IC50= 9.5 {micro}M). Multivalent display of the lead bicycles alongside [~]100 glycans in Liquid glycan Array (LiGA), made it possible to compare the binding of bicycles and glycans to Siglec-7 expressed on CHO, Jurkat, and Raji cells. LiGA assays confirmed binding of the bicycles to Siglec-7 but revealed considerable non-specific interactions with receptor-negative cells. Saturation transfer difference nuclear magnetic resonance (STD-NMR) revealed 46e binds to Siglec-7 at a site distinct from the V-Ig domain, suggesting it might inhibit binding of glycans to the glycan-binding site of Siglec-7 via an allosteric site. Together these results demonstrate that BiGEL enables the discovery of bicyclic peptides for undruggable Siglec targets but highlights future challenges in molecular discoveries that aim to identify small, non-carbohydrate inhibitors of GBPs.

biochemistry↗

bioRxiv2025

FlexRibbon: Joint Sequence and Structure Pretraining for Protein Modeling

AO_SCPLOWBSTRACTC_SCPLOWProtein foundation models have advanced rapidly, with most approaches falling into two dominant paradigms. Sequence-based language models (e.g., ESM-2) capture sequence semantics at scale, and a number of recent works incorporate structural signals into sequence encoders. MSA-based predictors (e.g., AlphaFold 2/3) achieve accurate folding by exploiting evolutionary couplings, but their reliance on homologous sequences makes them less reliable in highly mutated or alignment-sparse regimes. We present FlexRibbon{ddagger}, a pretrained protein model that jointly learns from amino acid sequences and three-dimensional structures. Our pretraining strategy combines masked language modeling with diffusion-based denoising, enabling bidirectional sequence-structure learning without requiring MSAs. Trained on both experimentally resolved structures and AlphaFold 2 predictions, FlexRibbon captures global folds as well as flexible conformations critical for biological function. Evaluated across diverse tasks spanning interface design, intermolecular interaction prediction, and protein function prediction, FlexRibbon establishes new state-of-the-art performance on 12 different tasks, with particularly strong gains in mutation-rich settings where MSA-based methods often struggle.

molecular biology↗

bioRxiv2025

Single-cell lineage tracing identifies hemogenic endothelial cells in the adult mouse bone marrow

During mouse development, hematopoietic stem and progenitor cells (HSPC) originate from hemogenic endothelial cells (ECs) through a process of endothelial-to-hematopoietic transition. These HSPC are thought to fully sustain adult hematopoiesis. However, it remains unknown whether adult ECs retain hemogenic potential. Here we used in vivo genetic lineage tracking at population and single-cell (sc) levels, scRNA sequencing, and bone marrow transplantation to detect hemogenic ECs in adult mice. We identify and characterize bone marrow-resident, adult Cdh5/VE-Cadherin+ ECs that produce hematopoietic cell-progeny in vitro and in mice. These adult hemogenic ECs and their hematopoietic cell progeny give rise to hematopoietic cells following adoptive transfer into adult mice. Furthermore, blood cells generated from adult and developmental ECs comparably home to peripheral tissues, where they similarly contribute to inflammatory responses. Thus, our results identify previously unrecognized bone marrow-derived adult hemogenic ECs that generate HSPC and functional mature blood cells.

cell biology↗

bioRxiv2025

GatorAffinity: Boosting Protein-Ligand Binding Affinity Prediction with Large-Scale Synthetic Structural Data

Protein-ligand binding affinity prediction is a fundamental task in computational drug discovery. Although substantial efforts have been made to enhance prediction accuracy using data-driven approaches, progress remains limited by persistent data scarcity. The widely used PDBbind dataset, for example, contains fewer than 20, 000 experimental structures with annotated binding affinities, while a vast number of affinity measurements remain underutilized due to missing structural data. Here, we investigate this untapped potential by curating more than 450, 000 synthetic protein-ligand complexes annotated with Kd and Ki values using the Boltz-1 structure prediction model. Building on this unprecedented scale of synthetic data, further augmented with over 1 million synthetic complexes from the recently released SAIR database annotated with IC50 values, we develop GatorAffinity, a geometric deep learning-based scoring function pretrained on large-scale synthetic data and fine-tuned using high-quality experimental structures from PDBbind. Extensive evaluation on a leak-proof benchmark demonstrates that GatorAffinity significantly outperforms state-of-the-art affinity prediction methods, offering superior accuracy and generalizability. Our findings show that augmenting available experimental data with synthetic complexes can effectively address the data scarcity challenge while maintaining strong predictive reliability. By releasing the pretrained GatorAffinity model and the large-scale synthetic dataset GatorAffinity-DB, we provide a scalable and reproducible foundation for affinity prediction, virtual screening, and broader structure-based drug design applications (https://github.com/AIDD-LiLab/GatorAffinity).

bioinformatics↗