Multiview learning for understanding functional multiomics.

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning

Background:

• Complex biological systems' molecular mechanisms are often elusive.
• High-throughput multiomics data offer multifaceted insights but are challenging to integrate.
• Machine learning is increasingly used for multiomics data analysis.

Purpose of the Study:

• Introduce multiview learning as a powerful machine learning technique for multiomics data integration.
• Review recent multiview learning methods and unify them under the multiview empirical risk minimization (MV-ERM) framework.
• Explore the application of multiview learning to diverse biological data and systems.

Main Methods:

• Review and framework unification of multiview learning algorithms.
• Discussion of applications across genomics, transcriptomics, and epigenomics.
• Exploration of use cases in human diseases, plant biology, and single-cell analysis.

Main Results:

• Multiview learning demonstrates superior capability in handling data heterogeneity and cross-talk compared to traditional methods.
• Potential applications identified for uncovering functional and mechanistic insights across various omics data.
• Benefits and limitations of multiview learning in biological contexts are discussed.

Conclusions:

• Multiview learning holds significant promise for advancing multiomics data interpretation.
• This approach can reveal complex molecular mechanisms and functions in diverse biological systems.
• Further research and application of multiview learning in biology are encouraged.