SGUQ: Staged Graph Convolution Neural Network for Alzheimer's Disease Diagnosis using Multi-Omics Data.

Area of Science:

• Neuroscience
• Computational Biology
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) is a leading cause of dementia, posing significant global health and economic challenges.
• High-throughput omics technologies have advanced AD molecular understanding, but conventional AI models require complete data upfront, proving inefficient.
• Current diagnostic approaches for AD using multi-omics data can be costly and may involve unnecessary tests.

Purpose of the Study:

• To develop a cost-effective and accurate method for Alzheimer's disease diagnosis using multi-omics data.
• To propose a novel staged graph convolutional network with uncertainty quantification (SGUQ) that selectively incorporates omics data.
• To reduce the financial burden and improve the efficiency of AD diagnosis.

Main Methods:

• A staged graph convolutional network with uncertainty quantification (SGUQ) was developed.
• SGUQ initiates diagnosis with mRNA data and progressively integrates DNA methylation and miRNA data as needed.
• The model was evaluated on the ROSMAP dataset for AD diagnosis accuracy.

Main Results:

• SGUQ achieved a diagnostic accuracy of 0.858 on the ROSMAP dataset, outperforming existing methods.
• 46.23% of samples were reliably predicted using only mRNA data (single-modal).
• An additional 16.04% of samples achieved reliable predictions with combined mRNA and DNA methylation data (two-modal).

Conclusions:

• The proposed SGUQ model offers a significant advancement in Alzheimer's disease diagnosis through efficient multi-omics data utilization.
• SGUQ reduces clinical costs and improves diagnostic accuracy by incorporating omics data selectively.
• The SGUQ framework has potential applications beyond AD diagnosis, including clinical decision-making with multi-viewed data.