Multi-modal data Integration for Metastatic Risk Prediction of AJCC Stage I/II Melanomas: Results from the Population-based Dutch Early-stage Melanoma Study (D-ESMEL)

Intermediate integration of genomics and transcriptomics in AJCC stage I/II melanoma case-control sets achieved the highest balanced accuracy (mean 0.62; 95% CI 0.57–0.66) in the D-ESMEL discovery set.

• Why it matters: Prognostic models for early-stage melanoma often use single data types or one integration approach, leaving uncertainty about which multi-modal combinations best capture metastatic progression.
• What they did: Using D-ESMEL multi-modal data (clinicopathological, genomic, transcriptomic, pathomics), the study compared early, intermediate, mixed, and late integration via repeated (n=5) nested 3-fold cross-validation in 221 matched discovery sets, then tested top models in 154 matched validation sets.
• The result: In validation, early integration, late integration, and unimodal transcriptomics each reached balanced accuracy 0.60; early integration remained independently prognostic (OR 1.75; p<0.01) and improved weighted C-index to 0.85 (p<0.01) when population-weighted.

AI summary of the abstract · may contain errors · not medical advice · read the paper.