CV
Education
- Ph.D., Pattern Recognition & Intelligence Systems, Beihang University, 2020.9 – present
- B.S., Control Science & Engineering, Beihang University, 2016.9 – 2020.6
Research Interest
- Probabilistic and generative modeling for high-dimensional biological data
- Continuous and structured representations of spatial and single-cell systems
- Scalable learning frameworks for integrating multi-modal and partially observed data
Publications and Manuscripts
Primary Research
-
Bohan Li†, Yue Deng*, Qionghai Dai*, et al. Tissue characterization at an enhanced resolution across spatial omics platforms with deep generative model
Nature Communications, 15:6541 (2024)
https://doi.org/10.1038/s41467-024-50837-5 -
Bohan Li, Feng Bao*, Yue Deng*, et al.
Reconstructing isotropic-resolution spatial transcriptomics from serial sections by modeling tissue continuity
Under review at Nature Methods
https://doi.org/10.1101/2025.08.15.670472
Collaborative Research
- M. Clevenger†, Bohan Li†, Feng Bao*, N. Chevrier*, et al.
Whole-body molecular and cellular mapping of the laboratory mouse
Cell (2026)
https://doi.org/10.1016/j.cell.2026.03.006
† Co-first author; * Corresponding author
Research Experience
My research develops generative modeling frameworks to enhance the effective resolution of spatial omics data, addressing both lateral and depth-wise sparsity by reconstructing continuous molecular landscapes from discrete measurements.
I. Lateral resolution enhancement for spatial omics (soScope)
Advisor: Yue Deng, Feng Bao
Proposed soScope, a generative framework that models spatial omics measurements as aggregated observations of latent cellular states.
- Formulated spatial aggregation within a variational inference framework
- Designed model architecture and learning objective
- Integrated histological morphology with molecular measurements in a unified probabilistic model
https://github.com/deng-ai-lab/soScope
II. Depth-wise resolution enhancement (isoST)
Advisor: Yue Deng, Feng Bao
Developed isoST, a generative framework that models spatial gene expression as a continuous process across tissue depth using stochastic differential equations.
- Developed continuous modeling framework based on stochastic differential equations
- Designed learning objectives for reconstruction from sparse sections
- Implemented scalable inference and reconstruction pipeline
https://github.com/deng-ai-lab/isoST
III. Cross-scale annotation transfer (LABEL)
Advisor: Feng Bao, Nicolas Chevrier
Developed LABEL, a framework for transferring molecular annotations from spatial transcriptomics data to histological images.
- Developed the LABEL module for cross-scale annotation transfer
- Contributed to model design and system integration
- Participated in large-scale data processing and validation
https://github.com/chevrierlab/WMST-paper
Academic Service
- Invited reviewer for Briefings in Bioinformatics