Robust Multi-Omics Integration from Incomplete Modalities Significantly Improves Prediction of Alzheimer's Disease

2025-09-25

AI Summary

Incomplete multi-omics data—characterized by frequent modality-wise missingness—severely hinder accurate modeling and biological interpretation of Alzheimer’s disease (AD). To address this, we propose MOIRA, a robust integrative framework for incomplete multi-omics data. MOIRA first aligns heterogeneous modalities into a shared latent embedding space via representation alignment; then employs a learnable-weight early-fusion mechanism for adaptive feature aggregation; and finally incorporates feature importance analysis to identify disease-relevant biomarkers. Evaluated on the ROSMAP cohort, MOIRA significantly outperforms state-of-the-art methods (p < 0.01) in both AD staging and cognitive decline prediction. Biologically, it successfully recapitulates established AD risk genes—including APOE and TREM2—as well as synaptic function pathways, demonstrating strong interpretability and clinical translational potential.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Multi-omics data capture complex biomolecular interactions and provide insights into metabolism and disease. However, missing modalities hinder integrative analysis across heterogeneous omics. To address this, we present MOIRA (Multi-Omics Integration with Robustness to Absent modalities), an early integration method enabling robust learning from incomplete omics data via representation alignment and adaptive aggregation. MOIRA leverages all samples, including those with missing modalities, by projecting each omics dataset onto a shared embedding space where a learnable weighting mechanism fuses them. Evaluated on the Religious Order Study and Memory and Aging Project (ROSMAP) dataset for Alzheimer's Disease (AD), MOIRA outperformed existing approaches, and further ablation studies confirmed modality-wise contributions. Feature importance analysis revealed AD-related biomarkers consistent with prior literature, highlighting the biological relevance of our approach.

Problem

• Integrating multi-omics data with missing modalities for disease prediction
• Addressing incomplete omics data hindering robust integrative analysis
• Improving Alzheimer's Disease prediction using incomplete multi-omics datasets

Innovation

• Robust multi-omics integration with incomplete data
• Uses representation alignment and adaptive aggregation
• Projects data onto shared embedding space for fusion