DeepMPTB

DeepMPTB utilizes vaginal microbiome data to predict preterm birth risk through a deep neural network achieving 84.10% accuracy and an AUROC of 0.875.

Key Features:

• Deep Neural Network Architecture: A deep neural network trained to classify term birth and preterm birth from vaginal microbiome data.
• Multi-Cohort Training: Trained on five independent study cohorts comprising 1290 vaginal samples from 561 pregnant women.
• Benchmarking Against Machine Learning Algorithms: Outperformed seven currently used machine learning algorithms during comparative evaluation.
• Diversity-Based Prediction: Incorporates overall diversity of the vaginal microbiota rather than specific species for preterm birth prediction.

Scientific Applications:

• Preterm Birth Risk Prediction: Predicts preterm birth risk to support clinicians in providing personalized patient assistance.
• Vaginal Microbiome Characterization: Analyzes vaginal metagenomics data for precise microbiome structure characterization.

Methodology:

Vaginal metagenomics data from five independent cohorts were analyzed for microbiome structure characterization, followed by training of a deep neural network to predict term birth and preterm birth.