Likun Hou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Likun Hou has a total of 157 co-authors. Here are the top co-authors by number of shared publications:
- Chunyan Wu
Department of Pathology Shanghai Pulmonary Hospital, Tongji University School of Medicine Shanghai People's Republic of China
14 shared publications - Jiawei Zou
Center for Excellence in Molecular Cell Science 1 Key Laboratory of Systems Biology Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences, Shanghai, China
13 shared publications - Jiazheng Pei
Center for Excellence in Molecular Cell Science 1 Key Laboratory of Systems Biology Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences, Shanghai, China
13 shared publications - Jing Xu
Center for Excellence in Molecular Cell Science 1 Key Laboratory of Systems Biology Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences, Shanghai, China
13 shared publications - Kai Xiao
Center for Excellence in Molecular Cell Science 1 Key Laboratory of Systems Biology Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences, Shanghai, China
13 shared publications - Luonan Chen
School of Mathematical Sciences and School of AI, Shanghai Jiao Tong University, Shanghai, China
13 shared publications
Latest Publications
- Open Access
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Figure S2 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Figure S2 The analysis of CarHE model training. A. The ablation performance of CarHE. B. Grad-CAM threshold sensitivity and convergence analyses. C. Convergence statistics when training CarHE. D. Data-scaling analysis demonstrating performance improvement with increased training-set size. E. Cross-platform and cross-tissue testing.
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S2 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Figure S2 The analysis of CarHE model training. A. The ablation performance of CarHE. B. Grad-CAM threshold sensitivity and convergence analyses. C. Convergence statistics when training CarHE. D. Data-scaling analysis demonstrating performance improvement with increased training-set size. E. Cross-platform and cross-tissue testing.
- Open Access
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Figure S3 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Figure S3. Grad-CAM alignment between histological morphology and cell-type distributions. A–D. Representative H&E regions with corresponding Grad-CAM activation maps showing spatial agreement between predicted and ground-truth distributions for tumor cells (A), Tregs (B), activated T cells (C), and cytotoxic CD8++ T cells (D).
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S3 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Figure S3. Grad-CAM alignment between histological morphology and cell-type distributions. A–D. Representative H&E regions with corresponding Grad-CAM activation maps showing spatial agreement between predicted and ground-truth distributions for tumor cells (A), Tregs (B), activated T cells (C), and cytotoxic CD8++ T cells (D).
- Open Access
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Data from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Spatial transcriptomic (ST) analyses provide spatially resolved gene expression data that can provide insights into complex biological processes. However, current ST approaches remain financially prohibitive and restricted in resolution, scalability, and gene coverage, limiting broader adoption for large-scale studies. In this study, we developed Contrastive Alignment of gene expRession for hematoxylin and eosin (H&E) images (CarHE), a multimodal pretraining framework that infers high-dimensional ST profiles from routine H&E-stained slides. By using contrastive learning to align cell type–specific transcriptomic information with histologic features, CarHE achieved high prediction accuracy across evaluated datasets and ST platforms. CarHE approximated spatially organized pathologic microenvironment features consistent with tertiary lymphoid structure–associated regions in breast cancer, lung cancer, melanoma, and clear-cell renal cell carcinoma. Additionally, CarHE inferred approximated three-dimensional (3D) ST context from two-dimensional (2D) images, providing more informative neighborhood context than 2D visualization. In a cohort of 880 patients with lung cancer, CarHE-derived features were associated with disease-free survival and outperformed current approaches. Overall, CarHE provides a cost-effective and scalable framework for H&E-based spatial inference, supporting further validation toward translational research applications. Significance:CarHE infers spatial transcriptomics landscapes from routine H&E slides that associate with patient outcomes, providing a powerful tool for localizing immune niches and microenvironmental programs within standard histological workflows.
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Open AccessData from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Spatial transcriptomic (ST) analyses provide spatially resolved gene expression data that can provide insights into complex biological processes. However, current ST approaches remain financially prohibitive and restricted in resolution, scalability, and gene coverage, limiting broader adoption for large-scale studies. In this study, we developed Contrastive Alignment of gene expRession for hematoxylin and eosin (H&E) images (CarHE), a multimodal pretraining framework that infers high-dimensional ST profiles from routine H&E-stained slides. By using contrastive learning to align cell type–specific transcriptomic information with histologic features, CarHE achieved high prediction accuracy across evaluated datasets and ST platforms. CarHE approximated spatially organized pathologic microenvironment features consistent with tertiary lymphoid structure–associated regions in breast cancer, lung cancer, melanoma, and clear-cell renal cell carcinoma. Additionally, CarHE inferred approximated three-dimensional (3D) ST context from two-dimensional (2D) images, providing more informative neighborhood context than 2D visualization. In a cohort of 880 patients with lung cancer, CarHE-derived features were associated with disease-free survival and outperformed current approaches. Overall, CarHE provides a cost-effective and scalable framework for H&E-based spatial inference, supporting further validation toward translational research applications. Significance:CarHE infers spatial transcriptomics landscapes from routine H&E slides that associate with patient outcomes, providing a powerful tool for localizing immune niches and microenvironmental programs within standard histological workflows.
- Open Access
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Figure S1 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Figure S1. The Umap of cell embedding using scGPT fine-tuned. A. The Umap of the UMAP colored by the curated main cell types, major lineages (e.g., epithelial, endothelial, immune, stromal, and parenchymal populations). B, C. The umap of organ and cell cluster distribution after embedding using fine-tuned scGPT. D. The relationship between cell cluster and cell types.
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S1 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Figure S1. The Umap of cell embedding using scGPT fine-tuned. A. The Umap of the UMAP colored by the curated main cell types, major lineages (e.g., epithelial, endothelial, immune, stromal, and parenchymal populations). B, C. The umap of organ and cell cluster distribution after embedding using fine-tuned scGPT. D. The relationship between cell cluster and cell types.
- Open Access
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Figure S4 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Figure S4. The prediction of CarHE on ccrcc patient 2. A. The H&E image of the FFPF tumor. B. Immune cell abundance predicted by CarHE and ground truth. C. Average expression of the non-switched IGHM gene, the switched immunoglobulin genes IGHG1 and IGHA1, and the fibroblast marker COL1A1.
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S4 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Figure S4. The prediction of CarHE on ccrcc patient 2. A. The H&E image of the FFPF tumor. B. Immune cell abundance predicted by CarHE and ground truth. C. Average expression of the non-switched IGHM gene, the switched immunoglobulin genes IGHG1 and IGHA1, and the fibroblast marker COL1A1.
- Open Access
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Figure S5 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Figure S5 The analysis of related factor related with DFS risk predicted by CarHE in lung cancer. A. The evaluation of CarHE and npj-DM2025, WSS-CNN performance with 6 to 24 monthe threshold for DFS prediction. B. The DE gene analysis by wilcox-test between high-risk and low risk samples. C. The GO pathwasy enrichment in high/low risk samples. G. The correlation bewteen pathway scores and CarHE risk. D. The bar plot of CarHE risk of different TMN groups. F. The correlation between CarHE risk and Macrophage/lymphocyte ritiao. G. The correlation between CarHE risk and lymphocyte/all cell ritiao.
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S5 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Figure S5 The analysis of related factor related with DFS risk predicted by CarHE in lung cancer. A. The evaluation of CarHE and npj-DM2025, WSS-CNN performance with 6 to 24 monthe threshold for DFS prediction. B. The DE gene analysis by wilcox-test between high-risk and low risk samples. C. The GO pathwasy enrichment in high/low risk samples. G. The correlation bewteen pathway scores and CarHE risk. D. The bar plot of CarHE risk of different TMN groups. F. The correlation between CarHE risk and Macrophage/lymphocyte ritiao. G. The correlation between CarHE risk and lymphocyte/all cell ritiao.
- Open Access
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Table S1 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Table S1. The metadata of dataset used in this study.
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Open AccessTable S1 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Table S1. The metadata of dataset used in this study.
- Open Access
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Table S2 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Table S2. The comprehensive table of training hyperparameters.
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Open AccessTable S2 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Table S2. The comprehensive table of training hyperparameters.
- Open Access
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Table S3 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Table S3. The top marker gene profile of cell type annotation before training.
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Open AccessTable S3 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Table S3. The top marker gene profile of cell type annotation before training.
- Open Access
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Table S6 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Table S6. Prognostic performance of CarHE beyond clinicopathologic variables for disease-free survival.
- 16 November 2026
- Published by American Association for Cancer Research (AACR)
Open AccessTable S6 from The Multimodal Pretraining Framework CarHE Predicts Spatial Transcriptomics in Tumors from Routine Pathology Images
Table S6. Prognostic performance of CarHE beyond clinicopathologic variables for disease-free survival.