GBDR: a Bayesian model for precise prediction of pathogenic microorganisms using 16S rRNA gene sequences

Area of Science:

• Microbiology
• Computational Biology
• Bioinformatics

Background:

• Human microorganisms play crucial roles in biological activities, and their interactions with the host can lead to complex disorders.
• Identifying disease-specific microbes is challenging and resource-intensive using traditional laboratory methods.
• Advancements in sequencing and omics necessitate computational approaches for large-scale microbe-disease association prediction.

Purpose of the Study:

• To develop a computational model for predicting microbe-disease associations.
• To prioritize potential microbial biomarkers for various human diseases.
• To leverage existing knowledge on microbe-disease relationships for predictive modeling.

Main Methods:

• A group-based computational model using Bayesian disease-oriented ranking was developed.
• Microbe-microbe similarity was measured using gene sequence information via BLAST+ and MEGA 7.
• Disease-disease similarity was calculated using hierarchy information from Medical Subject Headings (MeSH) data.

Main Results:

• The model demonstrated high performance with Area Under the ROC Curve (AUC) values of 0.9456 (leave-one-out) and 0.8866 (five-fold cross-validation).
• In a case study of colorectal carcinoma, 80% of the top-20 predicted microbes were experimentally confirmed in published literature.
• The model effectively prioritizes microbes associated with human diseases based on similarity measures.

Conclusions:

• The proposed computational model accurately and effectively predicts microbe-disease associations.
• This approach facilitates the identification of potential microbial biomarkers.
• The findings support the assumption that functionally similar microbes share similar interaction patterns with human diseases.