Xuefeng Wang

The paper introduces multiScaleAC, a method to analyze cell-cell interactions across spatial scales using a modified Moran's I with varying kernel bandwidths, showing improved statistical power and reduced bias.

This study establishes patient-derived organoids (PDOs) for HNSCC drug response testing, integrating AI-driven histopathology and multi-omics analysis to create a scalable preclinical model.

This study uses single-cell RNA-sequencing to show that aging promotes KRAS-driven lung adenocarcinoma evolution toward a stem-like state through age-related changes in the tumor microenvironment.

GPS-Net identifies prognostic pathway modules using network regularized kernel learning to improve biomarker accuracy and reduce computational complexity in cancer immunotherapy studies.

The paper develops ROICellTrack, a deep learning framework that integrates cellular imaging with spatial transcriptomics to analyze tumor tissues, revealing cancer-immune mixtures and spatial clustering patterns.

CGPA provides a platform for multi-context gene-centric prognostic analysis, enabling customized exploration of gene relationships and biomarker discovery in cancer research.

This paper analyzes 11,473 tumor samples across 33 cancer types to study gamma delta T cell receptor diversity and its prognostic potential, revealing its significance as a biomarker and therapeutic target.

SmartImpute improves scRNA-seq imputation by targeting marker genes, enhancing accuracy and efficiency while reducing computational burden.

The paper identifies differentially methylated regions in cfDNA from tumor and normal tissues using paired-sample analysis to discover potential biomarkers for head and neck cancers.

This study compares plasma cfDNA methylation profiles pre- and post-surgery in OCSCC patients to identify prognostic and diagnostic biomarkers.

The paper proposes a weighted sparse network learning method for high-dimensional genomics data to improve prediction accuracy and model interpretability, showing robustness and good performance in simulations and a melanoma gene expression case study.

This paper introduces Xsurv, an R package using XGB and LGB for survival analysis, showing it outperforms traditional methods in biomarker discovery for cancer prognosis.

This paper introduces a dual form of the Cox partial likelihood loss using Fenchel duality, enabling efficient kernel learning for survival analysis with censored outcomes.

This paper applies wide deep learning to cell type classification, showing improved accuracy when models are trained on one cancer type and tested on another.

This paper introduces an R package for multiple kernel learning, showing improved classification by combining multiple data sources in ovarian cancer.

lncDIFF detects differentially expressed long non-coding RNAs with higher power and lower false discovery rate compared to existing methods.

This study uses DNA methylation data to accurately predict immune cell infiltration in solid tumors, particularly in melanoma, and provides a simple model for clinical application.