NeuroBiome AI

Microbiome and Metabolomics Explorer for Alzheimer's-Like Patterns

NeuroBiome AI analyzes metabolomics data from the tryptophan pathway and short-chain fatty acids to explore patterns associated with Alzheimer's disease. This research tool uses machine learning to predict how similar a sample's metabolite profile looks to known Alzheimer's patterns.

Research Use Only

This tool is for research and educational purposes only. It is not a medical test and must not be used for diagnosis or treatment decisions.

Upload or Enter Data

Choose your preferred method to analyze metabolomics data

Upload Metabolomics CSV

Upload a .csv file where each row is a sample and columns are metabolite intensities (e.g., Tryptophan, IPA, Kynurenine, SCFAs).

Drag and drop your CSV file here

or

Try a Single Sample

Enter a few key features from the TRP and SCFA pathways to simulate a sample prediction.

Tryptophan Pathway

SCFA Pathway

How It Works

A research-grade approach to metabolomics pattern analysis

1

Multi-cohort Training

Data from multiple research studies (ST000046, ST000047, ST000462, ST001152, ST001050) was combined, cleaned, and harmonized to create a comprehensive training dataset.

2

Pathway-aware Features

The model focuses on key metabolic pathways: tryptophan → indole derivatives and short-chain fatty acids (SCFAs), known to be dysregulated in Alzheimer's disease.

3

Machine Learning Model

A RandomForestClassifier from scikit-learn was trained with leave-one-study-out cross-validation to ensure robust, generalizable predictions across different metabolomics cohorts.

4

Research-Only Output

Predictions are designed for exploratory research and pattern discovery, not for clinical diagnosis. All outputs should be interpreted in context by trained researchers.

Model Performance & Limitations

Performance (Research Only)

Cross-cohort ROC AUC

0.8–0.9

Varies by dataset

Validation Method

Leave-one-study-out cross-validation

Performance metrics are based on research cohorts. Real-world application performance may vary depending on population characteristics and sample collection methods.

ROC Curve (Simulated)

1.00

Limitations

  • •Trained on specific research cohorts; may not generalize to all populations or geographic regions.
  • •Different labs and instruments can produce different metabolite distributions and detection limits.
  • •Not evaluated or approved for clinical use by regulatory agencies.
  • •Predictions should never be used for patient care decisions without professional medical interpretation.
  • •Model performance on rare populations or edge cases is unknown and requires further validation.
Built with v0