Dong He Xu

This paper introduces EmitGCL, a deep-learning framework that predicts future metastasis in cancer with high accuracy, identifies biomarkers, and validates therapeutic targets.

The paper investigates DNA methylation-expression relationships in rice internode development using RNA-seq and bisulfite sequencing, finding specific methylation contexts correlate with gene expression patterns.

This study shows that SPL in lung epithelial cells reduces influenza severity by enhancing innate defenses and altering cell type proportions.

This paper introduces DPLM, a dynamics-aware protein language model that integrates molecular dynamics simulations with sequence data to improve predictions of protein properties.

This study introduces ProtLoc-GRPO, a reinforcement learning method that improves subcellular localization prediction by optimizing PPI networks, achieving a 7% macro-F1 score improvement.

This study shows that SphK2 deficiency in neutrophils reduces immune suppression, enhances T cell responses, and decreases viral persistence during LCMV infection.

This paper introduces scPEFT, a parameter-efficient fine-tuning framework for scLLMs that improves task adaptation with reduced computational costs and enhances performance in disease-specific and cross-species analyses.

This paper presents a framework for privacy-preserving genomic data sharing using AES-256-GCM, Curve25519, and blockchain for dynamic consent and access control.

This study uses TRAP-seq to analyze phloem-specific translational dynamics during soybean nodulation, identifying key regulators and pathways involved in the process.

This paper introduces GCP-VQVAE, a geometry-complete tokenizer for protein structures, achieving state-of-the-art performance on benchmark datasets and demonstrating robust generalization to unseen proteins.

This paper benchmarks sequence-based models for predicting protein-ligand binding sites across different ligand availability regimes, showing improved performance with ligand conditioning and chemical language model integration.

CryoFSL improves protein particle picking in cryo-EM micrographs with minimal annotations, achieving high accuracy and efficiency in low-SNR conditions.

TrimNN identifies spatial cell organization patterns from spatial transcriptomics and proteomics data to reveal associations between cell-type arrangements and phenotypes.

This study identifies over 2500 S-acylation sites in Arabidopsis membrane proteins, revealing potential substrates for PAT5 and PAT9 and highlighting their roles in plant immunity and signaling.

The paper shows that fatty acid synthesis and breakdown occur simultaneously during plant oil synthesis, challenging previous assumptions and explaining underperformance in engineered oil-producing plants.

The paper introduces two sequence-based models to predict kinase-specific phosphorylation sites, showing improved performance and zero-shot capability.

This paper introduces a unified model combining a protein language model with an autoregressive transformer to predict protein-ligand binding sites, achieving broad ligand coverage and improved performance for underrepresented ligands.

This paper extends Prot2Token to align protein language models for diverse prediction tasks, improving performance across multiple applications through pre-training and integration with ESM-2.

This paper introduces S-PLM, a 3D structure-aware protein language model that improves prediction tasks by integrating structural information through efficient fine-tuning.

S-PLM improves protein language models by integrating 3D structure information through contrastive learning, outperforming sequence-only models in clustering and classification tasks.

This study evaluates AlphaFold-Multimer, AlphaFold3, and ColabFold for predicting protein-peptide complex structures using Template-Based and Template-Free methods, finding AlphaFold-Multimer excels in Template-Based predictions.

G2PDeep-v2 is a web-based deep learning platform for phenotype prediction and biomarker discovery using multi-omics data across organisms, offering model training and visualization tools.

PEFT-SP improves signal peptide prediction using parameter-efficient fine-tuning on large protein language models, achieving higher MCC2 gains with limited training data.

This study improves gene relationship prediction using iterative prompt refinement with ChatGPT, achieving better performance for complex gene interactions.

Prot2Token is a multi-task framework that uses autoregressive language modeling for protein prediction tasks, achieving performance comparable to specialized models and enabling integration with large language models.

scBSP identifies spatially variable genes from spatial transcriptomic data efficiently and accurately across different sequencing techniques and resolutions.

CellDiffusion improves scRNA-seq data integration by accurately inferring cell-cell associations and correcting batch effects across diverse datasets.

This paper introduces parameter-efficient fine-tuning strategies to improve cell type identification in single-cell models, achieving better performance with reduced computational cost.

This paper evaluates 21 large language models for mining gene interactions and pathway knowledge, finding significant performance differences between API-based and open-source models.