Scuba: scalable kernel-based gene prioritization.

Area of Science:

• Genomics
• Computational Biology
• Precision Medicine

Background:

• Identifying genes associated with human diseases is crucial but challenging due to numerous candidate genes and data heterogeneity.
• Existing computational methods, particularly kernel-based approaches, struggle with scalability when integrating diverse biological knowledge.

Purpose of the Study:

• To introduce Scuba, a scalable kernel-based method for prioritizing candidate genes.
• To address the limitations of existing methods in handling large-scale genomic data and heterogeneous information sources.

Main Methods:

• Scuba utilizes a novel multiple kernel learning approach within a semi-supervised framework.
• The method is optimized for unbalanced datasets and efficiently manages a large number of candidate genes and data sources.
• An integrated strategy for selecting optimal kernel parameters enhances efficiency.

Main Results:

• Scuba demonstrates enhanced scalability compared to current kernel-based methods for genomic data analysis.
• Cross-validation and simulated experiments show Scuba outperforms various state-of-the-art gene prioritization techniques.
• The method effectively handles large candidate gene sets and heterogeneous data sources.

Conclusions:

• Scuba offers state-of-the-art performance and improved scalability for gene prioritization.
• This tool is valuable for identifying disease-associated genes, especially with large candidate pools or complex data.
• The Scuba software is publicly available for research use.