PEARL: integrative multi-omics classification and omics feature discovery via deep graph learning.

Area of Science:

• Biomedical informatics
• Computational biology
• Genomics and Systems Biology

Background:

• Multi-omics data integration offers comprehensive biological insights but faces challenges like low reliability and generalizability.
• High dimensionality and low sample size in omics data hinder existing computational methods.
• Advancements in precision medicine require robust multi-omics integration techniques.

Purpose of the Study:

• Introduce PEARL (Pearson-Enhanced spectrAl gRaph convoLutional networks), a novel deep graph learning method.
• Address challenges in high-dimensional, low-sample-size multi-omics data integration.
• Improve biomedical classification and identify functionally important omics features.

Main Methods:

• Developed PEARL, a deep graph learning framework utilizing a Pearson-enhanced spectral graph convolutional network architecture.
• Applied PEARL to both synthetic and real-world biomedical datasets, including Alzheimer's disease multi-omics data.
• Evaluated PEARL's performance against state-of-the-art methods for classification and feature prioritization.

Main Results:

• PEARL demonstrated superior and robust performance in high-dimensional, low-sample-size multi-omics settings.
• PEARL significantly outperformed existing state-of-the-art methods on various datasets.
• Features prioritized by PEARL in Alzheimer's disease data were enriched in AD-related pathways, enhancing biological interpretability.

Conclusions:

• PEARL offers a powerful and reliable approach for multi-omics data integration and analysis.
• The method enhances biological interpretability, aiding in the understanding of complex diseases.
• PEARL shows practical utility in biomedical research and precision medicine applications.