Functional networks inference from rule-based machine learning models.

Area of Science:

• Bioinformatics
• Systems Biology
• Machine Learning in Biology

Background:

• Functional networks are crucial for understanding biological systems.
• Inferring these networks from high-throughput omics data is a key research area.
• Traditional similarity-based methods (e.g., gene co-expression) have limitations in capturing complex relationships.

Purpose of the Study:

• To propose and evaluate FuNeL, a new protocol for inferring functional networks from machine learning models.
• To demonstrate the biological relevance and complementary nature of FuNeL-inferred networks compared to similarity-based methods.
• To showcase FuNeL's potential in identifying disease-associated genes.

Main Methods:

• Developed FuNeL, a protocol that infers functional networks based on genes used together in rule-based machine learning models.
• Tested FuNeL on synthetic datasets and a suite of 8 real-world human cancer datasets.
• Compared FuNeL networks against gene co-expression networks using enriched biological terms and disease-associated gene topology.

Main Results:

• FuNeL successfully inferred biologically relevant functional networks.
• The inferred networks demonstrated a complementary character to similarity-based methods.
• FuNeL identified known disease associations as core network elements, particularly in a prostate cancer case study.

Conclusions:

• FuNeL offers a powerful alternative for functional network inference, capturing complex gene relationships.
• The method provides biologically relevant insights and aids in discovering disease-associated genes.
• FuNeL's implementation is publicly available, facilitating its use in biological research.