A machine learning approach for multimodal data fusion for survival prediction in cancer patients
Problem
Cancer patient survival prediction lacked a comprehensive machine-learning pipeline for comparing multimodal data fusion methods.
Why it matters
Better survival predictions are stated to support precision oncology for cancer patients.
What the AI did
A machine-learning pipeline was built on TCGA data fusing transcripts, proteins, metabolites and clinical factors, comparing feature extraction and fusion strategies including late fusion.
Results
Late fusion models consistently outperformed single-modality approaches on TCGA lung, breast and pan-cancer datasets
Impact
Evidence supports improved prediction accuracy on TCGA datasets; clinical adoption is not stated.
Record
| Venue | npj Precision Oncology |
|---|---|
| Published | 2025 |
| Work type | article |
| Authors | Νικόλαος Νικολάου, Domingo Salazar, Harish RaviPrakash, Miguel Gonçalves, Rob Mulla, Nikolay Burlutskiy, Natasha Markuzon, Etai Jacob |
| Citations | 35 (OpenAlex, at import time) |
| References | 77 |
| Model or method | late fusion models (specific architecture not stated in source) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41698-025-00917-6 |
| Open access | gold |
| Keywords | multimodal fusion, cancer survival, machine learning, TCGA, late fusion, precision oncology |
Importance rating
significant — thousands to millions affected
Cancer survival prediction supports precision oncology, a field affecting large populations, but abstract only demonstrates a research pipeline on TCGA datasets; no scale or clinical adoption stated.