Related Experiment Video
Updated: Jan 15, 2026
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
7.4K
Machine learning risk stratification strategy for multiple myeloma: Insights from the EMN-HARMONY Alliance platform
Adrian Mosquera Orgueira1, Marta Sonia Gonzalez Perez1, Mattia D'Agostino2,3
1Department of Hematology, IDIS University Hospital of Santiago de Compostela Santiago de Compostela Spain.
Hemasphere
|October 13, 2025
Summary
Machine learning models offer improved risk stratification for multiple myeloma (MM) patients. These novel prognostic scores enhance prediction accuracy over traditional methods for newly diagnosed MM (NDMM).
Area of Science:
- Hematology
- Computational Biology
- Oncology
Background:
- Traditional risk stratification for multiple myeloma (MM) using clinical and cytogenetic data has limited predictive accuracy.
- Machine learning (ML) presents a novel approach to enhance prognostic capabilities by analyzing complex interactions within large datasets.
Purpose of the Study:
- To develop and validate novel ML-driven prognostic scores for newly diagnosed MM (NDMM).
- To improve upon the predictive accuracy of existing risk stratification systems for MM.
Main Methods:
- Analysis of the EMN-HARMONY MM cohort (14,345 patients, including 10,843 NDMM patients).
- Development of three ML models: comprehensive (20 variables), reduced (6 key variables), and cytogenetics-free.
- Internal validation (cross-validation) and external validation (Myeloma XI trial).
- Performance evaluation using concordance index (C-index) and time-dependent ROC-AUC.
Main Results:
- ML models demonstrated improved prognostic accuracy for overall survival (OS) and progression-free survival (PFS) compared to ISS, R-ISS, and R2-ISS.
- The comprehensive model achieved C-indices of 0.666 (OS training) and 0.620 (PFS training).
- Reduced and cytogenetics-free models also showed competitive performance, with enhanced accuracy when treatment data was included.
Conclusions:
- ML-based risk stratification provides individualized predictions for MM patients, outperforming traditional group-based methods.
- The developed models are reproducible across diverse MM subgroups, including real-world and relapsed/refractory populations.
- An online risk calculator is available for practical application of these advanced prognostic tools.
