POEMS: Product of Experts for Interpretable Multi-omic Integration using Sparse Decoding

2025-11-05

AI Summary

Multi-omics integration faces a fundamental trade-off between interpretability and nonlinear modeling. Method: We propose an unsupervised deep generative model that jointly leverages sparse decoding and Product-of-Experts (PoE) to address this challenge. Specifically: (i) structured sparse connections map raw omics features to shared latent factors, ensuring decoder interpretability; (ii) a gating network dynamically weights each omics modality’s contribution to the latent space; and (iii) the PoE framework enables nonlinear, probabilistic cross-omics fusion. Contribution/Results: Unlike linear decoders or black-box models that sacrifice interpretability, our approach preserves strong nonlinear expressivity while enabling biomarker identification and biologically meaningful cross-omics association analysis. On cancer subtype classification, the model achieves state-of-the-art clustering and classification performance, yielding interpretable, biologically plausible insights.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Multi-modal Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Integrating different molecular layers, i.e., multiomics data, is crucial for unraveling the complexity of diseases; yet, most deep generative models either prioritize predictive performance at the expense of interpretability or enforce interpretability by linearizing the decoder, thereby weakening the network's nonlinear expressiveness. To overcome this tradeoff, we introduce POEMS: Product Of Experts for Interpretable Multiomics Integration using Sparse Decoding, an unsupervised probabilistic framework that preserves predictive performance while providing interpretability. POEMS provides interpretability without linearizing any part of the network by 1) mapping features to latent factors using sparse connections, which directly translates to biomarker discovery, 2) allowing for cross-omic associations through a shared latent space using product of experts model, and 3) reporting contributions of each omic by a gating network that adaptively computes their influence in the representation learning. Additionally, we present an efficient sparse decoder. In a cancer subtyping case study, POEMS achieves competitive clustering and classification performance while offering our novel set of interpretations, demonstrating that biomarker based insight and predictive accuracy can coexist in multiomics representation learning.

Problem

• Overcoming tradeoff between predictive performance and interpretability in multiomics integration
• Enabling nonlinear expressiveness while maintaining biomarker discovery capabilities
• Providing cross-omic associations through shared latent space without linearization

Innovation

• Sparse connections mapping features to latent factors
• Product of experts model for cross-omic associations
• Gating network adaptively computes omic contributions