Identification of Key Taxonomic and Metabolic Players in the Gut Metagenome of T2D Patients

This repository presents a comprehensive bioinformatics analysis of gut metagenomic data from Chinese Type 2 Diabetes (T2D) patients. The study investigates microbial taxonomic composition, functional potential, gene/protein annotations (including CAZyme families), and applies machine learning to associate microbial patterns with clinical metadata.

---

📊 Project Overview

• Study Design: Longitudinal analysis of gut microbiota in T2D patients pre- and post-treatment.
• Samples: 162 fecal samples from 91 patients, treated with either Acarbose or Vildagliptin.
• Goals:
  • Reconstruct metagenomes from sequencing data
  • Identify microbial taxa and metabolic functions
  • Annotate CAZyme families
  • Analyze species and functions using statistical and ML approaches

---

🧬 Materials and Methods

• Data Acquisition: Retrieved from NCBI BioProject.
• Preprocessing: Quality control (FastQC, fastp), host read removal (BBDuk).
• Assembly: De novo assembly using metaSPAdes via KBase.
• Taxonomic Profiling: MetaPhlAn v4 with Bowtie2.
• Functional Profiling: HUMAnN v3.9 using UniRef50_EC.
• Gene Prediction: Prodigal with meta mode.
• Non-redundant Catalog: Clustered genes/proteins using MMseqs2 (90% identity, 80% coverage).
• CAZyme Annotation: DIAMOND BLASTP against CAZyme DB.
• Quantification: CoverM used to estimate gene counts.
• Machine Learning: Random Forest, SVM, XGBoost, etc. applied to species/protein metadata.
• Statistical Analyses:
  • HALLA: Mutual information-based association analysis
  • MaAsLin2: Linear model-based multivariate association testing

Ready to apply this analysis to your own data? All the necessary code and scripts to replicate these steps are available in the Scripts folder of this repository.

---

📈 Results and Discussion (Highlights)

• Significant variation in microbial taxa after drug treatment (Acarbose vs. Vildagliptin).
• Identification of key metabolic pathways, especially carbohydrate metabolism.
• CAZyme families (e.g., GH13, GH2) were enriched in high responders.
• ML models (e.g., XGBoost, Random Forest) achieved high accuracy in predicting drug response based on microbiome profiles.
• HALLA and MaAsLin2 revealed associations between specific microbes and clinical markers (GLP-1, BMI, etc.)

Full visualizations and detailed results are available in the Results and Output_Graphs directories.

---

▶️ Getting Started

1. Clone the repository:

   git clone https://github.com/yourusername/Identification-of-key-taxonomic-and-metabolic-players-in-the-gut-metagenome-of-T2D-Patients.git
   cd Identification-of-key-taxonomic-and-metabolic-players

2. Ensure required tools are installed:

   • FastQC
   • fastp
   • BBMap (BBDuk)
   • metaSPAdes / KBase
   • MetaPhlAn
   • HUMAnN
   • Prodigal
   • MMseqs2
   • DIAMOND
   • CoverM
   • Python, R (for ML and stats analysis)

---

📜 License

This project is open-source and available under the MIT License. You're free to use and adapt the code, data, and results with proper attribution.

---

🙏 Acknowledgments

• Zhang et al. (2022) for the dataset
• Developers of all the open-source tools used
• Community resources and collaborators who supported this project

---

🔬 Designed and analyzed by a Bioinformatics MS graduate with expertise in metagenomics and ML applications in microbiome research.