PhenoProfiler: Advancing Morphology Representations for Image-based Drug Discovery

Code, WebServer

Bo Li, Bob Zhang, …, Qianqian Song

Under review in Nature Communications (Major Revision)

Abstract: Achieving end-to-end cell image encoding for the first time in Image-based Phenotypic Drug Discovery tasks, with performance comparable to non-end-to-end encoding while reducing inference time by approximately 40 times.

SpaIM: Single-cell Spatial Transcriptomics Imputation via Style Transfer

Code

Bo Li, Ziyang Tang, …, Qianqian Song

Nature Communications 2025 (In Press)

Abstract: Predicting unmeasured gene expression in spatial transcriptomics (ST) using single-cell RNA sequencing (scRNA-seq) and achieving state-of-the-art (SOTA) performance in cross-modal imputation based on style transfer concepts.

Gene Expression Prediction from Histology Images via Hypergraph Neural Networks

Code

Bo Li, Yong Zhang, …, Qianqian Song

Briefings in Bioinformatics 2024

Abstract: Constructing a hypergraph via Euclidean distance and adjacent position weighting to exploit local correlations in cell images, enabling gene expression prediction for cells in Whole Slide Image (WSI) pathology images.

Multi-scale Hypergraph-based Feature Alignment Network for Cell Localization

Code

Bo Li, Yong Zhang, …, Baocai Yin

Pattern Recognition 2024

Abstract: Rethinking the cell localization task from the perspective of feature alignment for the first time and proposing a multi-scale hypergraph module that significantly improves accuracy by adaptively aggregating multi-level features.

Exponential Distance Transform Maps for Cell Localization

Code

Bo Li, Jie Chen, …, and Hong Bu

Engineering Applications of Artificial Intelligence 2024

Abstract: To address the challenge of existing density maps losing cell location information in dense regions, a new exponential distance transform map is proposed to provide accurate cell location information with reasonable gradients.

Lite-UNet: A Lightweight and Efficient Network for Cell Localization

Code

Bo Li, Yong Zhang, …, Baocai Yin

Engineering Applications of Artificial Intelligence 2024

Abstract: A lightweight and efficient cell localization model called Lite-UNet is proposed for quickly and accurately localizing cells in images to further improve the efficiency of computer-aided medicine.

Hypergraph Association Weakly Supervised Crowd Counting

Code

Bo Li, Yong Zhang, …, Baocai Yin

ACM Transactions on Multimedia Computing, Communications and Applications 2023

Abstract: A novel hypergraph association module is proposed to solve the problem of uneven distribution of crowd density by encoding higher-order associations among features, which opens a new direction to solve this problem.