Individualized dynamic risk assessment for multiple myeloma.

Area of Science:

• Oncology
• Genetics
• Machine Learning

Background:

• Accurate risk stratification is crucial for individualized treatment decisions in multiple myeloma (MM).
• Patient-specific genetic abnormalities and tumor microenvironment impact MM outcomes and therapy response.

Purpose of the Study:

• To develop a novel risk classification scheme for MM using machine learning.
• To improve prediction of progression-free survival (PFS) across diverse MM patient cohorts.

Main Methods:

• Trained a machine learning (ML) algorithm on the mmSYGNAL network, derived from multi-omics data of 881 MM patients.
• Evaluated the ML algorithm's performance against cytogenetics, International Staging System, and multi-gene panels.
• Validated the risk classification scheme across four independent patient cohorts.

Main Results:

• The mmSYGNAL risk classification significantly outperformed existing methods in predicting PFS.
• Demonstrated accurate prediction of MM progression risk at various disease stages: primary diagnosis, pre- and post-transplant, and after relapse.
• Highlighted the utility of mmSYGNAL for dynamic, individualized risk assessment throughout the disease trajectory.

Conclusions:

• mmSYGNAL offers enhanced, individualized risk stratification for MM patients.
• The model accounts for unique genetic abnormalities and monitors risk longitudinally.
• This approach supports dynamic management of MM based on changing disease characteristics.