MO-RiskVAE: A Multi-Omics Variational Autoencoder for Survival Risk Modeling in Multiple MyelomaMO-RiskVAE

2026-04-06

AI Summary

This work addresses the limitations of existing survival risk modeling approaches that integrate multi-omics and clinical data, where standard variational autoencoders (VAEs) often fail to preserve prognosis-relevant variation due to overly restrictive latent-space regularization, leading to unstable or suboptimal representations. Building upon the MyeVAE framework, the study systematically investigates how the strength of latent regularization, posterior geometry, and organization of latent variables influence multimodal survival prediction, proposing a novel model termed MO-RiskVAE. This model integrates multiple regularization mechanisms—including KL divergence, maximum mean discrepancy (MMD), and Hilbert–Schmidt independence criterion (HSIC)—and incorporates a continuous-discrete hybrid Gumbel-Softmax structure to enhance risk stratification capability. Experiments demonstrate that MO-RiskVAE significantly outperforms the original MyeVAE in risk stratification of multiple myeloma patients without requiring additional supervision or complex training strategies, while also revealing that survival performance is more sensitive to the magnitude and structure of regularization than to the specific choice of divergence measure.

Abstract

Multimodal variational autoencoders (VAEs) have emerged as a powerful framework for survival risk modeling in multiple myeloma by integrating heterogeneous omics and clinical data. However, when trained under survival supervision, standard latent regularization strategies often fail to preserve prognostically relevant variation, leading to unstable or overly constrained representations. Despite numerous proposed variants, it remains unclear which aspects of latent design fundamentally govern performance in this setting. In this work, we conduct a controlled investigation of latent modeling choices for multimodal survival prediction within a unified extension of the MyeVAE framework. By systematically isolating regularization scale, posterior geometry, and latent space structure under identical architectures and optimization protocols, we show that survival-driven training is primarily sensitive to the magnitude and structure of latent regularization rather than the specific divergence formulation. In particular, moderate relaxation of KL regularization consistently improves survival discrimination, while alternative divergence mechanisms such as MMD and HSIC provide limited benefit without appropriate scaling. We further demonstrate that structuring the latent space can improve alignment between learned representations and survival risk gradients. A hybrid continuous--discrete formulation based on Gumbel--Softmax enhances global risk ordering in the continuous latent subspace, even though stable discrete subtype discovery does not emerge under survival supervision. Guided by these findings, we instantiate a robust multimodal survival model, termed MO-RiskVAE, which consistently improves risk stratification over the original MyeVAE without introducing additional supervision or complex training heuristics.

Problem

• survival risk modeling
• multimodal variational autoencoder
• latent regularization
• multiple myeloma
• multi-omics integration

Innovation

• Multi-Omics Integration
• Variational Autoencoder
• Survival Risk Modeling
• Latent Space Regularization
• Gumbel-Softmax