DiaPredict

This project presents a comprehensive machine learning pipeline for predicting Type 2 Diabetes (T2D) using gut microbiome data derived from 16S rRNA sequencing. The workflow integrates QIIME2 for sequence preprocessing, taxonomic profiling, and feature table generation, followed by functional and pathway inference using PICRUSt2. Various ML algorithms—including Random Forest, SVM, and XGBoost—are trained on taxonomic and functional features to classify diabetic and control samples. MLflow is employed to automate model tracking, hyperparameter tuning, and performance logging, enabling efficient experimentation. The project also includes biomarker discovery through feature importance analysis and robust visualization for both taxonomic and functional insights, offering a reproducible and interpretable approach to microbiome-based T2D prediction.