A Generative Multi-State Deep Learning Framework for Dynamic Virtual-Twin Modeling of Multiple Myeloma

Abstract Multiple myeloma (MM) is a heterogeneous and incurable cancer of plasma cells in which patients experience repeated disease relapses and widely variable survival outcomes. Existing risk prediction models focus primarily on patients at initial diagnosis, leaving a critical gap for individuals with relapsed or refractory disease, where previous treatments could potentially influence future outcomes. Computational virtual twins (VTs) represent a promising approach for personalized medicine by continuously updating patient-specific predictions as new clinical information becomes available. However, enabling such personalized prediction requires artificial intelligence (AI) models that can jointly represent disease trajectories and heterogeneous, longitudinal patient data. We present DG-MSM, a generative multi-state deep learning framework designed as the computational foundation for VT modeling. The model jointly learns disease-state transitions, event timing, and heterogeneous longitudinal patient characteristics, incorporating the full patient history into the risk assessment. Beyond dynamic risk prediction, this generative structure enables uncertainty-aware, patient-specific simulation of future disease trajectories, allowing outcomes to be examined as full predictive distributions rather than single risk scores. DG-MSM achieves competitive performance relative to dedicated time-to-event models, while providing long-term survival predictions that grow more accurate as a patient's treatment history accumulates, an advantage most pronounced in later treatment lines. These findings demonstrate the potential of DG-MSM to enable predictive virtual twins in MM and other chronic diseases, providing dynamically updated prognosis, individualized risk assessment, and patient-specific trajectory simulation to support treatment planning.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-13
DOI
https://doi.org/10.21203/rs.3.rs-11204103/v1
Primary Topic
Multiple Myeloma Research and Treatments
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

A Generative Multi-State Deep Learning Framework for Dynamic Virtual-Twin Modeling of Multiple Myeloma

Holger Fröhlich, Yasamin Salimi, Shammi More, Guillaume Azarias et al.
5 links
Research Square
Multiple Myeloma Research and Treatments
preprint

A Generative Multi-State Deep Learning Framework for Dynamic Virtual-Twin Modeling of Multiple Myeloma

Holger Fröhlich, Yasamin Salimi, Shammi More, Guillaume Azarias, Sebastian Schwick
preprint en

Abstract

Abstract Multiple myeloma (MM) is a heterogeneous and incurable cancer of plasma cells in which patients experience repeated disease relapses and widely variable survival outcomes. Existing risk prediction models focus primarily on patients at initial diagnosis, leaving a critical gap for individuals with relapsed or refractory disease, where previous treatments could potentially influence future outcomes. Computational virtual twins (VTs) represent a promising approach for personalized medicine by continuously updating patient-specific predictions as new clinical information becomes available. However, enabling such personalized prediction requires artificial intelligence (AI) models that can jointly represent disease trajectories and heterogeneous, longitudinal patient data. We present DG-MSM, a generative multi-state deep learning framework designed as the computational foundation for VT modeling. The model jointly learns disease-state transitions, event timing, and heterogeneous longitudinal patient characteristics, incorporating the full patient history into the risk assessment. Beyond dynamic risk prediction, this generative structure enables uncertainty-aware, patient-specific simulation of future disease trajectories, allowing outcomes to be examined as full predictive distributions rather than single risk scores. DG-MSM achieves competitive performance relative to dedicated time-to-event models, while providing long-term survival predictions that grow more accurate as a patient's treatment history accumulates, an advantage most pronounced in later treatment lines. These findings demonstrate the potential of DG-MSM to enable predictive virtual twins in MM and other chronic diseases, providing dynamically updated prognosis, individualized risk assessment, and patient-specific trajectory simulation to support treatment planning.

Research Square
Fraunhofer Institute for Algorithms and Scientific Computing (DE)
Multiple Myeloma Research Foundation, European Commission, European Health and Digital Executive Agency
Good health and well-being
Multiple Myeloma Research and Treatments
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon