Identifying Critical Phases for Disease Onset with Sparse Haematological Biomarkers

2025-03-18

AI Summary

Existing interpolation methods for sparse, irregularly sampled clinical blood biomarker data often introduce signal distortion, prediction bias, and lack biological interpretability. To address this, we propose the Graph Neural Additive Network (GNAN), which models irregular sampling events as a time-weighted directed graph and employs additive decomposition to explicitly disentangle feature-specific contributions from temporal dynamics—bypassing interpolation entirely and directly learning biologically plausible trajectory patterns. Evaluated on real-world blood test data, GNAN accurately identifies critical pre-symptomatic time windows for disease onset, yielding unbiased predictions with mechanistic interpretability. It significantly outperforms interpolation-based baselines (p < 0.01) in both predictive accuracy and clinical interpretability, enabling early intervention and facilitating hypothesis generation for underlying pathophysiological mechanisms.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• 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

Abstract

Routinely collected clinical blood tests are an emerging molecular data source for large-scale biomedical research but inherently feature irregular sampling and informative observation. Traditional approaches rely on imputation, which can distort learning signals and bias predictions while lacking biological interpretability. We propose a novel methodology using Graph Neural Additive Networks (GNAN) to model biomarker trajectories as time-weighted directed graphs, where nodes represent sampling events and edges encode the time delta between events. GNAN's additive structure enables the explicit decomposition of feature and temporal contributions, allowing the detection of critical disease-associated time points. Unlike conventional imputation-based approaches, our method preserves the temporal structure of sparse data without introducing artificial biases and provides inherently interpretable predictions by decomposing contributions from each biomarker and time interval. This makes our model clinically applicable, as well as allowing it to discover biologically meaningful disease signatures.

Problem

• Identify critical disease phases using sparse biomarkers.
• Model biomarker trajectories with time-weighted directed graphs.
• Provide interpretable predictions without imputation biases.

Innovation

• Graph Neural Additive Networks model biomarker trajectories
• Time-weighted directed graphs preserve temporal structure
• Decomposes feature and temporal contributions for interpretability