Feature Synthesis and Interpretation for Multi-Omics Data Integration

With growing omics data being generated, powerful data-driven computational methods are urgently needed. Machine learning has been widely applied in analyzing these biological data, including sequencing data and interaction network data. How to generate and extract discriminant features is crucial for subsequent tasks. In addition, most machine learning involves black-box methods. Interpretability is crucial for analyzing biological data and understanding biological processes. Recently, deep feature synthesis and network embedding methods have been widely used. Considering that genes generally function through interacting with others, it bears important practical utility for learning representation from the interaction data for follow-up tasks and end-to-end graph neural network for biological tasks, and it also calls for robust model interpretability to explain how a model makes a prediction.

This Research Topic will bring together the state-of-the-art research contributions that include feature synthesis from multi-omics data and interpretable rule learning for explaining how a model makes a prediction in bioinformatics application. All submitted articles will be peer-reviewed and selected on the basis of their quality and relevance to the theme of this collection.

The subtopics of interest include, but are not limited to:
• Deep feature synthesis for multi-omics data.
• Integration of biological motivation into network embedding algorithms.
• Unsupervised network embedding on the interaction network.
• Graph convolution network for biological data analysis.
• Rule learning on large-scale biological data
• Model regulatory networks using multi-omics data.

Topic Editor Xi Wang is employed by The BASF Corporation (Ghent, Belgium). All other Topic Editors declare no competing interests with regard to the Research Topic subject.