Biology subjects

Crawford, J.

Publications and source records attributed to Crawford, J..

At least 19 recordsLinked to original sources

bioRxiv2026

Peroxisome quality control is gated by a dynamic ATF6α-ABCD3 tether

The endoplasmic reticulum (ER) coordinates protein folding and lipid biosynthesis, while peroxisomes govern fatty-acid catabolism and redox balance. However, little is known about the mechanisms by which these organelles communicate. Here we show that activating transcription factor 6 (ATF6), a cytoprotective ER-resident sensor of the unfolded protein response (UPR), forms a molecular tether that zippers the ER and peroxisomes together at a 23-nm intermembrane distance. ATF6 achieves this by inserting a short segment of its unstructured N-terminal cytosolic tail into the solvated transmembrane cavity of the peroxisomal ABC-type lipid transporter ABCD3, trapping it in an off-state. This interaction is reinforced by Ceapin, a previously identified small molecule molecular-glue ligand, which promotes ER-peroxisome contact formation. The junctions serve as a scaffold that can initiate peroxisome turnover by p62-triggered autophagy (pexophagy), as enhanced lipid processing resulting from organelle zippering leads to peroxisome damage due to excessive reactive oxygen species (ROS) accumulation. These findings define ATF6 as a non-transcriptional regulator of organelle quality control and provide a template for engineering Ceapin analogues against ATF6/ABCD3-dependent colorectal and hepatocellular cancers and related diseases.

cell biology↗

bioRxiv2026

Replication fork directionality reveals how structural variants arise under replication stress

Structural variants (SVs) in cancer are associated with defects in DNA repair and replication stress, but the mechanisms generating common SV types remain unresolved. We propose that large (>100 kb) tandem duplications originate through a novel sister-fork breakage-fusion mechanism. To capture replication-related context beyond breakpoints, we developed an algorithm to characterize replication timing, origin density, and fork direction across SV-spanned regions, features that refine and differentiate previously defined SV signatures. Large tandem duplications frequently overlap replication origins from which forks proceed bidirectionally; combined with independent evidence from APOBEC strand asymmetry, this pattern is compatible uniquely with the proposed mechanism. Although tandem duplications in CCNE1-amplified and CDK12-mutant cancers also concentrate around origins and highly transcribed genes, they display distinct contexts: CDK12-mutant SVs arise near later-firing origins, whereas those in CCNE1-amplified tumors often coincide with genes in specific strand configurations, suggesting different causes of fork stalling. Incorporating replication features into signature analysis enabled the discovery of new SV signatures, which we used to build SVIG, a multi-class classifier of SV phenotypes. SV signatures attributed to replication stress may help guide therapies targeting this vulnerability.

bioinformatics↗

bioRxiv2026

Fully Phased Telomere-to-Telomere Assemblies for Thoroughbred Horse and Donkey Haplotypes derived from a Mule Illuminate the Peculiar Evolution of Equid Centromeres

We present telomere-to-telomere genome assemblies of a Thoroughbred horse and a donkey derived from their mule offspring. Now adopted and annotated by NCBI as reference genomes, these assemblies resolve previously inaccessible regions, including satellite arrays, duplications, and telomeres. Equids are known to exhibit an uncoupling between satellite DNA and centromeric function. The completeness of these assemblies enabled annotation of both satellite-based and satellite-free centromeres, as well as non-centromeric satellite loci, revealing notable centromeric plasticity. They also allowed detailed characterization of the variable binding domains of CENP-A--the epigenetic determinant of centromere identity--and CENP-B, whose association with CENP-A, previously considered typical based on a few model organisms, is absent in equids. Comparative analyses of satellite repeats and centromere positions provide new insights into the accelerated karyotypic reshuffling in equid evolution. These assemblies represent foundational resources for equid genomics and support ongoing initiatives such as the Equine Pangenome Project.

molecular biology↗

bioRxiv2026

Sequential, Chromosome-Specific Glutamine Synthetase Double Knockout with Cas-CLOVER Establishes Enhanced CHO Platforms for Cell Line Development

Cas-CLOVER is an emerging high-fidelity genome editing system that enables precise and efficient cell engineering. In this study, we applied Cas-CLOVER to establish a robust, gene-edited platform in suspension-adapted CHO-K1 cells supporting cell line development (CLD) for biopharmaceutical production. An attractive strategy for high yield clone selection is the use of glutamine synthetase (GS) knockout CHO cells. The primary GS gene resides on chromosome 5 (GS5), while a recently identified GS pseudogene is located on chromosome 1 (GS1). To compare editing efficiency, we evaluated Cas-CLOVER and Cas9 at both GS loci using the Neon Transfection System. Cas-CLOVER achieved 84% editing at GS5 and 74% at GS1, markedly higher than Cas 9. Leveraging Cas-CLOVERs dual-guide RNA design, we generated a GS5 single knockout (GS5-SKO) and subsequently a double knockout (GS-DKO) line at both the GS5 and GS1 loci, both with none detected off-target mutations analyzed in 40 predictably off-target sites. For functional validation, these cell lines were engineered with the proprietary Harbor-IN transposase system to stably express trastuzumab. Using an optimized protocol, the resulting GS-DKO platform, termed CleanCut GS CHO, enabled stringent selection and yielded high-producing clones with cell-specific productivity exceeding 100 pg/cell/day and antibody titers greater than 5 g/L in 24 deep well-plate fed-batch cultures after 14 days. The antibody titer stability analysis showed consistency over 60 generations. Collectively, these findings establish Cas-CLOVER as a versatile genome editing tool for developing high-yield CHO host platforms in CLD.

bioengineering↗

bioRxiv2026

CAR-T-drug conjugate against solid tumor

Chimeric antigen receptor (CAR)-T cell therapy has been clinically successful in hematologic cancers but faces challenges in solid tumors, primarily due to limited tumor infiltration, immunosuppressive tumor microenvironments (TME), and antigen heterogeneity. While combining CAR-T cell therapy with chemotherapy can enhance antitumor activity, this often leads to substantial systemic toxicity. In this study, we introduce CAR-T-drug conjugate (CAR-T-D-C), a new class of dual-functional therapeutics that effectively addresses these obstacles by integrating click chemistry for precise conjugation of cytotoxic agents onto antigen-specific CAR-T cells. By transforming the hostile TME into an ally, this approach facilitates localized delivery of cytotoxic payload directly to the tumor site, enhancing the overall effectiveness of CAR-T therapy in solid tumors. CAR-T-D-Cs with different CAR-T cell binders exhibit robust antitumor activity across diverse solid tumor models, including both human tumor xenografts and syngeneic models. Spatial transcriptomic studies reveals that CAR-T-D-C achieves improved CAR-T tumor infiltration and functional activation within the TME. Compared to conventional CAR-T therapy, CAR-T-D-C markedly enhances immune cell infiltration, augments effector functions, promotes antigen spreading, amplifies systemic immune responses, and improves overall anti-tumor immunity. CAR-T-D-C represents a versatile therapeutic concept that combines the potency of small molecule drugs and the specificity of CAR-T cells as a 2-in-1 immunochemotherapy for treatment of solid tumors.

immunology↗

bioRxiv2025

UC Irvine Brain Initiative Cell Atlas Network (BICAN) Brain Procurement Program for the Center for Multiomic Human Brain Cell Atlas Project

High-quality neurotypical postmortem human brain tissue is essential but difficult to obtain for constructing comprehensive human brain cell atlases. Here we describe the establishment of UC Irvines Brain Procurement Program, a coordinated initiative to collect and process neurotypical donor brains for multiomic mapping studies within the NIH BRAIN Initiative Cell Atlas Network (BICAN) consortium. Through partnerships with the Orange County Coroners Office, the UC Irvine Willed Body Program, UCI Medical Center, and the Childrens Hospital of Orange County, we have developed standardized workflows encompassing donor identification, postmortem brain recovery and processing, region-of-interest dissection, neurotypical donor selection, and data management. Our experience demonstrates the feasibility of a community-based, multi-institutional procurement framework while highlighting challenges in recruiting neurotypical donors and ensuring demographic representation reflective of Southern California. We further identify opportunities to strengthen outreach and donation pathways. This program provides a scalable model for advancing population-reflective, high-quality human brain cell atlas efforts. HighlightsO_LIEstablish a multi-site pipeline for procuring high-quality neurotypical human brains. C_LIO_LIDemonstrate feasibility of donor collection across childhood to adulthood. C_LIO_LIIdentify barriers and propose strategies to improve broad donor recruitment. C_LI

neuroscience↗

bioRxiv2025

Allosteric Disordering of eIF2B Regulates the Integrated Stress Response

The ternary complex (TC), composed of translation initiation factor eIF2, GTP, and initiator methionyl tRNA, delivers the first amino acid to the ribosome to initiate protein synthesis. The activity of the decameric eukaryotic initiation factor 2B complex (eIF2B) initiates TC assembly by catalyzing GDP to GTP exchange on eIF2, thereby setting the TC levels in the cell. Stress-induced phosphorylation converts eIF2 from the substrate of the GDP/GTP exchange reaction into an inhibitor (eIF2-P) of eIF2B. This conversion reduces the cells TC levels and induces the widespread reprogramming of translation known as the Integrated Stress Response (ISR). Here, we chart an allosteric axis running through eIF2B, revealing the importance of a protrusive -helix in its {beta}-subunit, the latch-helix, that locks onto the -subunit to induce eIF2B activity. eIF2-P binding unhooks the latch-helix, opening eIF2B, which inhibits its GDP/GTP exchange activity. Distinct viral proteins have convergently evolved to bind to eIF2B and stabilize the latch-helix-bound active state. Using these insights, we generated ISR-ACTivating compounds, ISRACTs, that stabilize eIF2B in its inhibited, unlatched state. Our study thus highlights how state-transitions in eIF2B are regulated via long-range allostery.

biochemistry↗

bioRxiv2024

Cardiolipin Inhibits the Noncanonical Inflammasome by Preventing LPS Binding to Caspase 4/11 to Mitigate Endotoxemia in Vivo

In Gram-negative bacterial sepsis, excessive caspase 4/11 activation in response to circulating bacterial lipid LPS (endotoxemia) can cause organ damage and mortality. Current inhibitors of caspase 4/11 also block caspase 1 activity and are therefore not appealing clinical candidates for treating Gram-negative sepsis. Here, we identify double-unsaturated 18:2 cardiolipin as a selective inhibitor of caspase 4/11-dependent inflammatory cytokine secretion and pyroptosis, without affecting caspase-1 responses. Cardiolipin targets the CARD domain of caspase 4/11, impeding its interaction with LPS to restrain caspase 4/11 activation, thereby suppressing endotoxemia-induced systemic inflammation in vivo. Thus, we present cardiolipin as a promising candidate for preventing endotoxemia- induced sequelae in sepsis while preserving caspase-1-driven anti-microbial immune responses. By identifying cardiolipin as a specific caspase 4/11 inhibitor, we provide an urgently-needed tool for studying caspase 4/11 functions in inflammatory pathways, and open the way to studies of noncanonical inflammasome regulation by endogenous cardiolipin.

immunology↗

bioRxiv2024

The formation and propagation of human Robertsonian chromosomes

Robertsonian chromosomes are a type of variant chromosome found commonly in nature. Present in one in 800 humans, these chromosomes can underlie infertility, trisomies, and increased cancer incidence. Recognized cytogenetically for more than a century, their origins have remained mysterious. Recent advances in genomics allowed us to assemble three human Robertsonian chromosomes completely. We identify a common breakpoint and epigenetic changes in centromeres that provide insight into the formation and propagation of common Robertsonian translocations. Further investigation of the assembled genomes of chimpanzee and bonobo highlights the structural features of the human genome that uniquely enable the specific crossover event that creates these chromosomes. Resolving the structure and epigenetic features of human Robertsonian chromosomes at a molecular level paves the way to understanding how chromosomal structural variation occurs more generally, and how chromosomes evolve.

genomics↗

bioRxiv2024

CRISPR activation of Tfeb, a master regulator of autophagy and lysosomal biogenesis, in osteoblast lineage cells increases bone mass and strength

Autophagy is a recycling pathway in which damaged or dysfunctional proteins, protein aggregates, and organelles are delivered to lysosomes for degradation. Insufficiency of autophagy is thought to contribute to several age-related diseases including osteoporosis. Consistent with this, elimination of autophagy from the osteoblast lineage reduces bone formation and causes low bone mass. However, whether increasing autophagy would benefit bone health is unknown. Here, we increased expression of the endogenous Transcription Factor EB gene (Tfeb) in osteoblast lineage cells in vivo via CRISPR activation. Tfeb overexpression stimulated autophagy and lysosomal biogenesis in osteoblasts. Tfeb overexpressing male mice displayed a robust increase in femoral and vertebral cortical thickness at 4.5 months of age. Histomorphometric analysis revealed that the increase in femoral cortical thickness was due to increased bone formation at the periosteal surface. Tfeb overexpression also increased femoral trabecular bone volume. Consistent with these results, bone strength was increased in Tfeb overexpressing mice. Female Tfeb overexpressing mice also displayed a progressive increase in bone mass over time and at 12 months of age had high cortical thickness and trabecular bone volume. This increase in vertebral trabecular bone volume was due to elevated bone formation. Osteoblastic cultures showed that Tfeb overexpression increased proliferation and osteoblast formation. Overall, these results demonstrate that stimulation of autophagy in osteoblast lineage cells promotes bone formation and strength and may represent an effective approach to combat osteoporosis.

physiology↗

bioRxiv2024

Novel high throughput 3D ECM remodeling assay identifies MEK as key driver of fibrotic fibroblast activity

In fibrotic tissues, activated fibroblasts remodel the collagen-rich extracellular matrix (ECM). Intervening with this process represents a candidate therapeutic strategy to attenuate disease progression. Models that generate quantitative data on 3D fibroblast-mediated ECM remodeling with the reproducibility and throughput needed for drug testing are lacking. Here, we develop a model that fits this purpose and produces combined quantitative information on drug efficacy and cytotoxicity. We use microinjection robotics to design patterns of fibrillar collagen-embedded fibroblast clusters and apply automated microscopy and image analysis to quantify ECM remodeling between-, and cell viability within clusters of TGF{beta}-activated primary human skin or lung fibroblasts. We apply this assay to compound screening and reveal actionable targets to suppress fibrotic ECM remodeling. Strikingly, we find that after an initial phase of fibroblast activation by TGF{beta}, canonical TGF{beta} signaling is dispensable and, instead, non-canonical activation of MEK-ERK signaling drives ECM remodeling. Moreover, we reveal that higher concentrations of two TGF{beta} receptor inhibitors while blocking canonical TGF{beta} signaling, in fact stimulate this MEK-mediated profibrotic ECM remodeling activity.

cell biology↗

bioRxiv2024

Complete sequencing of ape genomes

We present haplotype-resolved reference genomes and comparative analyses of six ape species, namely: chimpanzee, bonobo, gorilla, Bornean orangutan, Sumatran orangutan, and siamang. We achieve chromosome-level contiguity with unparalleled sequence accuracy (<1 error in 500,000 base pairs), completely sequencing 215 gapless chromosomes telomere-to-telomere. We resolve challenging regions, such as the major histocompatibility complex and immunoglobulin loci, providing more in-depth evolutionary insights. Comparative analyses, including human, allow us to investigate the evolution and diversity of regions previously uncharacterized or incompletely studied without bias from mapping to the human reference. This includes newly minted gene families within lineage-specific segmental duplications, centromeric DNA, acrocentric chromosomes, and subterminal heterochromatin. This resource should serve as a definitive baseline for all future evolutionary studies of humans and our closest living ape relatives.

genomics↗

bioRxiv2024

Latrophilin-2 mediates fluid shear stress mechanotransduction at endothelial junctions

Endothelial cell responses to fluid shear stress from blood flow are crucial for vascular development, function and disease. A complex of PECAM-1, VE-cadherin, VEGF receptors (VEGFRs) and PlexinD1 located at cell-cell junctions mediates many of these events. But available evidence suggests that another mechanosensor upstream of PECAM-1 initiates signaling. Hypothesizing that GPCR and G proteins may serve this role, we performed siRNA screening of G subunits and found that Gi2 and Gq/11 are required for activation of the junctional complex. We then developed a new activation assay, which showed that these G proteins are activated by flow. We next mapped the G residues required for activation and developed an affinity purification method that used this information to identify latrophilin-2 (Lphn-2/ADGRL2) as the upstream GPCR. Latrophilin-2 is required for all PECAM-1 downstream events tested. In both mice and zebrafish, latrophilin-2 is required for flow-dependent angiogenesis and artery remodeling. Furthermore, endothelial specific knockout demonstrates that latrophilin plays a role in flow-dependent artery remodeling. Human genetic data reveal a correlation between the latrophilin-2-encoding Adgrl2 gene and cardiovascular disease. Together, these results define a pathway that connects latrophilin-dependent G protein activation to subsequent endothelial signaling, vascular physiology and disease.

molecular biology↗

bioRxiv2024

Protein Translation Dysregulation and Immune Cell Evasion Define Metastatic Clones in HPV-related Cancer of the Oropharynx

Head and neck cancers, representing the seventh most common malignancy globally, have seen a shift in causative factors from traditional smoking and alcohol use to human papillomavirus (HPV) infection, now accounting for up to 80% of oropharyngeal cancers. We identify the cellular and clonal mechanisms underlying immune avoidance and metastasis by analysing single-cell and spatial genomic data from primary and metastatic cancers. We first map the clonal evolution of malignant cells based on the accumulation of mutations. We identify metastasising clones based on mutational similarity scores between cells in the primary and lymph node metastasis. Genomic analysis of metastasising and non-metastasising clones identified virally mediated protein translation relief (P=4.24x10-24) pathway underlying metastatic expansion. We show that in metastatic clones, this process is driven through upregulation of transition-initiating factors, EIF4E (P=1.5x10-13) and EIFG1 (P<2.22x10-16), and suppression of regulatory kinases EIF4EBP1 (P=2.1x10), EIF2AK2 (P<2.22x10-16), and EIF2S1 (P<2.22x10-16). We subsequently identify that metastatic clones have a corresponding downregulation of the JAK/STAT pathway and immunoproteasome genes PSMB8 (P<2.22x10- 16) and PSMB9 (P<2.22x10-16), suggesting these clones escape immune surveillance through decreased INF inflammatory response and antigen presentation. We validate these results using spatial RNA-seq data, where metastatic cancer clones show decreased cell-to-cell interactions with CD4 T-effector memory cells (CD4TEM) (P=0.0077), CD8 T-exhausted cells (CD8Ex) (P=0.0191), and innate lymphoid cells (ILC) (P=0.04). Finally, we demonstrate that the upregulation of cap-independent translational drives cell proliferation in metastatic clones through the expression of translation initiation factors (EIF4G1: P<2.22x10-16). Our results provide evidence of the mechanisms by which virally induced cancer clones lead to advanced disease and poor prognosis in patients.

cancer biology↗

bioRxiv2023

The effect of clinically relevant beta-lactam, aminoglycoside, and quinolone antibiotics on bacterial extracellular vesicle release from E. coli

Sepsis, a leading cause of death in hospitals, can be defined as a dysregulated host inflammatory response to infection, which can lead to tissue damage, organ failure, and cardiovascular complications. Although there is no cure for sepsis, the condition is typically managed with broad spectrum antibiotics to eliminate any potential bacterial source of infection. However, a potential side-effect of antibiotic treatment is the enhanced release of bacterial extracellular vesicles (BEVs). BEVs are membrane-bound nanoparticles produced by a variety of mechanisms, one of which includes the pinching-off of the outer membrane (in Gram-negative bacteria) to enclose proteins and other biological molecules for transport and intercellular communication. Some of the Gram-negative EV cargo, including Peptidoglycan associated lipoprotein (Pal) and Outer membrane protein A (OmpA), have been shown to induce both acute and chronic inflammation in host tissue. We hypothesize that antibiotic concentration and its mechanism of action can have an effect on the amount of released BEVs, which could potentially exacerbate the host inflammatory response. In this study, we evaluated nine clinically relevant antibiotics for their effect on EV release from Escherichia coli. EVs were characterized using immunoblotting, nanoparticle tracking analysis, and transmission electron microscopy. Several beta-lactam antibiotics caused significantly more EV release, while quinolone and aminoglycosides caused relatively less vesiculation. Further study is warranted to corroborate the correlation between an antibiotics mechanism of action and its effect on EV release, but these results underline the importance of antibiotic choice when treating sepsis patients.

microbiology↗

bioRxiv2023

Targeting the Hippo pathway in cancers via ubiquitination dependent TEAD degradation

The Hippo pathway is among the most frequently altered key signaling pathways in cancer. TEAD1-4 are essential transcription factors and key downstream effectors in the Hippo pathway in human cells. Here, we identified RNF146 as a ubiquitin ligase (E3) of TEADs, which negatively regulates their stability in cells through proteasome-mediated degradation. We show that RNF146-mediated TEAD ubiquitination is dependent on the TEAD PARylation state. We further validated the genetic interaction between RNF146 and the Hippo pathway in cancer cell lines and the model organism Drosophila melanogaster. Despite the RNF146 and proteasome-mediated degradation mechanisms, TEADs are stable proteins with a long half-life in cells. We demonstrate that degradation of TEADs can be greatly enhanced pharmacologically with heterobifunctional chemical inducers of protein degradation (CIDEs). These TEAD-CIDEs can effectively suppress activation of YAP/TAZ target genes in a dose-dependent manner and exhibit significant anti-proliferative effects in YAP/TAZ-dependent tumor cells, thus phenocopying the effect of genetic ablation of TEAD protein. Collectively, this study demonstrates that the ubiquitin-proteasome system plays an important role in regulating TEAD functions and provides a proof-of-concept demonstration that pharmacologically induced TEAD ubiquitination could be leveraged to target YAP/TAZ-driven cancers.

cancer biology↗

bioRxiv2023

Optimizer's dilemma: optimization strongly influences model selection in transcriptomic prediction

MotivationMost models can be fit to data using various optimization approaches. While model choice is frequently reported in machine-learning-based research, optimizers are not often noted. We applied two different implementations of LASSO logistic regression implemented in Pythons scikit-learn package, using two different optimization approaches (coordinate descent and stochastic gradient descent), to predict driver mutation presence or absence from gene expression across 84 pan-cancer driver genes. Across varying levels of regularization, we compared performance and model sparsity between optimizers. ResultsAfter model selection and tuning, we found that coordinate descent (implemented in the liblinear library) and SGD tended to perform comparably. liblinear models required more extensive tuning of regularization strength, performing best for high model sparsities (more nonzero coefficients), but did not require selection of a learning rate parameter. SGD models required tuning of the learning rate to perform well, but generally performed more robustly across different model sparsities as regularization strength decreased. Given these tradeoffs, we believe that the choice of optimizers should be clearly reported as a part of the model selection and validation process, to allow readers and reviewers to better understand the context in which results have been generated. Availability and implementationThe code used to carry out the analyses in this study is available at https://github.com/greenelab/pancancer-evaluation/tree/master/01_stratified_classification. Performance/regularization strength curves for all genes in the Vogelstein et al. 2013 dataset are available at https://doi.org/10.6084/m9.figshare.22728644.

bioinformatics↗

bioRxiv2023

An Ensemble Learning Approach to perform Link Prediction on Large Scale Biomedical Knowledge Graphs for Drug Repurposing and Discovery

Generating knowledge graph embeddings (KGEs) to represent entities (nodes) and relations (edges) in large scale knowledge graph datasets has been a challenging problem in representation learning. This is primarily because the embeddings / vector representations that are required to encode the full scope of data in a large heterogeneous graph needs to have a high dimensionality. The orientation of a large number of vectors requires a lot of space which is achieved by projecting the embeddings to higher dimensions. This is not a scalable solution especially when we expect the knowledge graph to grow in size in order to incorporate more data. Any efforts to constrain the embeddings to lower number of dimensions could be problematic as insufficient space to spatially orient the large number of embeddings / vector representations within limited number of dimensions could lead to poor inferencing on downstream tasks such as link prediction which leverage these embeddings to predict the likelihood of existence of a link between two or more entities in a knowledge graph. This is especially the case with large biomedical knowledge graphs which relate several diverse entities such as genes, diseases, signaling pathways, biological functions etc. that are clinically relevant for the application of KGs to drug discovery. The size of the biomedical knowledge graphs are therefore much larger compared to typical benchmark knowledge graph datasets. This poses a huge challenge in generating embeddings / vector representations of good quality to represent the latent semantic structure of the graph. Attempts to circumvent this challenge by increasing the dimensionality of the embeddings often render hardware limitations as generating high dimensional embeddings is computationally expensive and often times infeasible. To practically deal with representing the latent structure of such large scale knowledge graphs (KGs), our work proposes an ensemble learning model in which the full knowledge graph is sampled into several smaller subgraphs and KGE models generate embeddings for each individual subgraph. The results of link prediction from the KGE models trained on each subgraph are then aggregated to generate a consolidated set of link predictions across the full knowledge graph. The experimental results demonstrated significant improvement in rank-based evaluation metrics on task specific link predictions as well as general link predictions on four open-sourced biomedical knowledge graph datasets.

bioinformatics↗