PM-CNN: microbiome status recognition and disease detection model based on phylogeny and multi-path neural network

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• The human microbiome significantly impacts health and disease.
• Disparities in microbiome composition are observed between healthy individuals and those with diseases.
• Current machine learning models for microbiome analysis often neglect crucial microbial relationships, limiting their predictive power.

Purpose of the Study:

• To introduce a novel phylogeny-based neural network model, PM-CNN, for enhanced microbiome data analysis.
• To improve multi-status classification and disease detection using microbial phylogenetic relationships.
• To overcome the limitations of existing machine learning approaches in microbiome research.

Main Methods:

• Developed PM-CNN (Phylogenetic Multi-path Convolutional Neural Network), a neural network model that leverages microbial phylogenetic relationships.
• Organized microbes based on their evolutionary history to extract relevant features.
• Employed a multi-path convolutional neural network architecture combined with ensemble learning for robust classification.

Main Results:

• PM-CNN demonstrated superior performance in classifying human microbiome data for status and disease detection compared to existing machine learning models.
• The model effectively utilizes phylogenetic information for feature extraction and classification.
• Achieved significant improvements in accuracy for microbiome-based health state recognition.

Conclusions:

• PM-CNN offers a powerful new approach for analyzing microbiome data, integrating phylogenetic information for improved accuracy.
• The findings establish a strong foundation for microbiome-based disease prediction and health monitoring.
• The developed software is publicly available, facilitating further research and application in the field.