Jianxin Wang

USP-ddG predicts mutational effects on protein-protein interactions by improving data efficiency and modeling conformational dynamics, outperforming existing methods on benchmark datasets.

RefLA improves genome assembly speed and quality by using a reference genome to guide long-read assembly, outperforming both de novo and fast assemblers.

panHiTE detects population-specific transposable elements in large-scale plant genomes, improving sensitivity and precision while reducing redundancy and enabling TE-gene associations.

PyrMol improves molecular property prediction by integrating expert knowledge into a hierarchical graph framework, outperforming existing methods on benchmark datasets.

TopoLa enhances cell representations in single-cell and spatial omics by using topology-encoded latent hyperbolic geometry, improving performance across multiple tasks.

The paper introduces scMustree, a tree-construction algorithm that improves clustering accuracy and reveals functional associations among cell populations in scRNA-seq data.

NeuralTE is a deep learning method for classifying transposable elements at the superfamily level, achieving high accuracy and robustness across diverse datasets.

This paper introduces a unified framework for drug-target interaction prediction using a semantic-guided meta-path method, achieving better performance and handling cold-start scenarios effectively.

This study examines the mitochondrial genome evolution in turbellarians, finding greater selection pressure in aquatic species and identifying NAD4 as a key gene under positive selection.

This study introduces DISHyper, a hypergraph neural network that integrates annotated gene sets to predict cancer genes, identifying 44 novel ones associated with multiple cancers.

PECAT improves diploid genome assembly from long noisy reads by correcting errors and resolving haplotypes, achieving high contiguity in cattle and human genomes.

This paper introduces scCAD, a method for identifying rare cell types in scRNA-seq data by iteratively decomposing clusters, and shows it outperforms existing methods across multiple datasets.

The paper introduces ccsmeth, a deep-learning method for detecting DNA 5mCpGs using PacBio CCS, achieving high accuracy and AUC, and presents ccsmethphase for haplotype-aware methylation detection.

HiTE improves TE detection accuracy by dynamically adjusting boundaries, identifying more full-length TEs than existing tools in rice.

The paper evaluates sequencing and assembly methods to create high-quality diploid human genome assemblies, finding that long reads and haplotype sorting improve accuracy and completeness.

This paper introduces CDHGNN, a model using edge-weighted graph attention and heterogeneous graph neural networks to predict circRNA-disease associations, outperforming existing methods.

This study determines the complete mitogenome of Lysmata vittata and identifies its phylogenetic position as an independent clade within Decapoda.

This study introduces a deep learning pipeline to detect 5mC in CpG and non-CpG contexts from Nanopore sequencing data in plants, achieving high accuracy compared to bisulfite sequencing.

The paper accelerates RepeatClassifier using Spark and a greedy algorithm with dynamic upper boundary, achieving 42-88x speedup while maintaining accuracy.

SSRE improves cell type identification in scRNA-seq by enhancing similarity measurement through sparse subspace representation and gene selection.

This study predicts Alzheimer's disease-associated genes using brain gene networks and identifies causal variants, providing a resource for therapeutic target discovery.

LoopPredictor predicts enhancer-promoter loops in cell types lacking long-range contact maps using ensemble machine learning and multi-omics features.

ClusterMine integrates biological knowledge to cluster samples based on gene set expression profiles, improving interpretability of clusters.

NECAT improves Nanopore read error correction and de novo assembly, achieving higher accuracy and faster performance for genome assembly.

DeepSignal detects DNA methylation sites from Nanopore sequencing reads using deep learning, achieving higher accuracy than HMM-based methods across multiple organisms and methylation types.

BPG is an R package that generates publication-quality, customizable plots and offers an interactive web tool for figure customization and integration with computational pipelines.

The paper introduces SIMMS, a method that identifies biomarkers by analyzing molecular data and biological networks, showing improved performance in predicting disease subtypes and outcomes.