Ensemble learning based on matrix completion improves microbe-disease association prediction.

Area of Science:

• Microbiology
• Computational Biology
• Bioinformatics

Background:

• Microbes significantly impact human health, but identifying disease-associated microbes is challenging.
• Current experimental methods are labor-intensive, limiting the discovery of microbe-disease links.
• Accurate computational tools are essential for efficient prediction and screening of potential microbe-disease associations.

Purpose of the Study:

• To develop an advanced computational framework, SABMDA, for improved microbe-disease association inference.
• To integrate multi-source data from microbes and diseases for enhanced prediction accuracy.
• To validate the efficacy of SABMDA against existing methods and in real-world scenarios.

Main Methods:

• Integration of multi-source information from microbial and disease data.
• Development and application of two successive matrix completion algorithms for association prediction.
• Utilizing an ensemble learning approach within the SABMDA framework.

Main Results:

• Ablation tests confirmed that combining the two matrix completion algorithms enhances prediction performance.
• SABMDA significantly outperformed seven recent baseline methods in comprehensive cross-validation and independent tests.
• The framework demonstrated remarkable prediction ability when applied to identify microbes associated with three specific diseases.

Conclusions:

• SABMDA offers a robust and accurate computational approach for predicting microbe-disease associations.
• The findings support the utility of SABMDA in accelerating the discovery of disease-related microbes.
• This framework has the potential to guide future drug development and disease treatment strategies.