Back

Transforming Histology into Virtual Multiplex Immunofluorescence to Decode Prognostic Spatial Immunity in Hepatocellular Carcinoma

Cai, L.; Jiang, S.; Liang, J.; Liu, F.; Zhang, B.; Reitsam, N. G.; Zeng, Q.; Ma, Y.; Li, Z.; Feng, S.; Hu, M.; Zhang, X.; Zhang, J.; Kather, J. N.; Zhang, Y.; Liang, W.

2026-02-26 pathology
10.64898/2026.02.25.707931 bioRxiv
Show abstract

The spatial organization of the tumor immune microenvironment (TIME) drives hepatocellular carcinoma (HCC) prognosis but remains unquantifiable on routine H&E slides. Here, we present HCCExplorer, a deep learning framework that translates H&E into virtual multiplex immunofluorescence (mIF) and uses multi-modal graph learning to decode spatial survival signals. Trained on 30 H&E-mIF slide pairs, HCCExplorer evaluated a 1,813-slide multi-center cohort. It achieved superior overall survival stratification over clinical indices and state-of-the-art pathology and protein foundation models, yielding a concordance index of 0.71 and a Hazard Ratio (HR) of 15.46 (P < 0.001), maintaining stability across three external cohorts. Beyond risk stratification, interpretation of model features identified M1 macrophage infiltration as a protective determinant (HR=0.40, P < 0.05). Furthermore, it uncovered a protective "Containment Niche" at the invasion frontier (HR=0.02, P < 0.01), featuring macrophages co-localizing with Foxp3+ Tregs and CD4+ T cells. Ultimately, HCCExplorer provides actionable, spatially-resolved biomarkers from conventional histology for precision HCC management.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

1
Nature Communications
○ 5641 papers in training set
Top 7%
19.2%
Similar papers in this journal
2
Nature Methods
○ 385 papers in training set
Top 0.3%
19.2%
Similar papers in this journal
3
Nature
○ 645 papers in training set
Top 0.8%
15.7%
Similar papers in this journal
50% of probability mass above
4
Nature Machine Intelligence
○ 70 papers in training set
Top 0.2%
8.2%
Similar papers in this journal
5
Cancer Discovery
● 66 papers in training set
Top 0.6%
4.5%
Similar papers in this journal
6
Nature Medicine
○ 125 papers in training set
Top 0.7%
3.6%
7
Advanced Science
○ 286 papers in training set
Top 2%
3.2%
8
npj Digital Medicine
○ 118 papers in training set
Top 2%
1.8%
9
Cell Discovery
○ 57 papers in training set
Top 0.4%
1.8%
10
Cell Reports
○ 1498 papers in training set
Top 18%
1.8%
11
ACS Nano
● 113 papers in training set
Top 1%
1.6%
12
Nature Aging
○ 60 papers in training set
Top 1%
1.2%
13
Cell Reports Medicine
○ 153 papers in training set
Top 4%
0.9%
14
Science Advances
● 1243 papers in training set
Top 29%
0.9%
15
Light: Science & Applications
○ 16 papers in training set
Top 0.3%
0.9%
16
National Science Review
◐ 21 papers in training set
Top 0.2%
0.9%
17
eLife
● 5828 papers in training set
Top 65%
0.9%
18
Proceedings of the National Academy of Sciences
● 2444 papers in training set
Top 40%
0.9%
19
Modern Pathology
○ 22 papers in training set
Top 0.4%
0.6%
20
Nature Biomedical Engineering
○ 47 papers in training set
Top 2%
0.6%
21
Advanced Intelligent Systems
○ 11 papers in training set
Top 0.3%
0.6%
22
Communications Medicine
○ 113 papers in training set
Top 4%
0.6%
23
Communications Biology
○ 993 papers in training set
Top 33%
0.6%
24
iScience
○ 1154 papers in training set
Top 44%
0.5%
25
Cell Reports Methods
○ 165 papers in training set
Top 5%
0.5%

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.