Learning Explainable Imaging-Genetics Associations Related to a Neurological Disorder

2025-08-22

AI Summary

Traditional neuroimaging genetics approaches are limited by restrictive linear assumptions or lack of interpretability, hindering the characterization of nonlinear, biologically interpretable associations between brain structural phenotypes and genetic variants in neuropsychiatric disorders. To address this, we propose NeuroPathX—a novel interpretable deep learning framework that integrates MRI-derived features and pathway-level genomic information via early fusion and models their dynamic interplay using a cross-attention mechanism. NeuroPathX further introduces two custom losses: a sparsity loss to capture individual-specific effects and a pathway similarity loss to enforce biological consistency across populations. Evaluated on autism spectrum disorder and Alzheimer’s disease cohorts, NeuroPathX significantly outperforms state-of-the-art methods, identifying reproducible, literature-supported imaging-genetic associations. The framework establishes a new paradigm for mechanistic insight and biomarker discovery in neurogenetics.

Technology Category

• Machine Learning: Neuro-Symbolic Learning
• Humans and AI: Brain-Sensing and Analysis
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

While imaging-genetics holds great promise for unraveling the complex interplay between brain structure and genetic variation in neurological disorders, traditional methods are limited to simplistic linear models or to black-box techniques that lack interpretability. In this paper, we present NeuroPathX, an explainable deep learning framework that uses an early fusion strategy powered by cross-attention mechanisms to capture meaningful interactions between structural variations in the brain derived from MRI and established biological pathways derived from genetics data. To enhance interpretability and robustness, we introduce two loss functions over the attention matrix - a sparsity loss that focuses on the most salient interactions and a pathway similarity loss that enforces consistent representations across the cohort. We validate NeuroPathX on both autism spectrum disorder and Alzheimer's disease. Our results demonstrate that NeuroPathX outperforms competing baseline approaches and reveals biologically plausible associations linked to the disorder. These findings underscore the potential of NeuroPathX to advance our understanding of complex brain disorders. Code is available at https://github.com/jueqiw/NeuroPathX .

Problem

• Learning explainable imaging-genetics associations for neurological disorders
• Capturing meaningful interactions between brain structure and genetics
• Overcoming limitations of simplistic linear or black-box models

Innovation

• Explainable deep learning with cross-attention mechanisms
• Early fusion strategy integrating MRI and genetics data
• Sparsity and pathway similarity loss functions for interpretability