Multi-modal intermediate integrative methods in neuropsychiatric disorders: A review

Area of Science:

• Computational biology and bioinformatics
• Systems biology approaches to neuropsychiatric disorders

Background:

• Neuropsychiatric disorders arise from complex biological processes across genomics, transcriptomics, epigenetics, proteomics, and metabolomics.
• High-throughput technologies and large datasets enable multi-omics data integration, crucial for systems biology.
• Integrating heterogeneous multi-omics data presents significant computational challenges.

Purpose of the Study:

• To review multi-modal intermediate integration techniques for multi-omics data in neuropsychiatric research.
• To explore how intermediate integration captures complementary information and reveals cross-omics interactions.
• To assess the strengths and weaknesses of various intermediate integration methods.

Main Methods:

• Review of intermediate integration techniques including component analysis, matrix factorization, similarity networks, multiple kernel learning, Bayesian networks, artificial neural networks, and graph transformation.
• Analysis of applications of these techniques within neuropsychiatric domains.
• Comparative assessment of the reviewed methods.

Main Results:

• Intermediate integration transforms individual omics data into shared representations, facilitating cross-omics analysis.
• This approach highlights novel interactions and complementary information across different omics layers.
• Various component analysis, matrix factorization, and network-based methods show promise for multi-omics integration.

Conclusions:

• Intermediate integration strategies offer advantages over early and late integration for multi-omics data.
• The reviewed techniques provide valuable tools for computational scientists studying neuropsychiatric disorders.
• Findings aid researchers in transforming and integrating multi-omics data for deeper insights into disease etiology.