Personalized reference intervals
Setpoint Identification through Gaussian Mixture Modelling
Identify patient-specific setpoints for blood count data using Gaussian Mixture Models. This can be used to construct personalized reference intervals, to identify what is a 'normal' test result for each patient. Based on our 2025 Nature paper.
Tip: Enter at least 5 values for accurate variance estimates.
📈 Time Series Visualization
📊 Distribution & Model Fit
📖 About This Method
What it does: Gaussian Mixture Models (GMMs) identify patient-specific setpoints by separating normal baseline values
from outlier distributions. The algorithm automatically selects the optimal number of components (1-3) based
on Bayesian Information Criterion (BIC) and weight constraints. The resultant setpoint and variance estimate can be used to create a patient-specific reference interval for a given lab test, identifying
what range of test results are normal for them.
Note: This is a re-implementation of our original GMM code for the web app, and is for illustrative purposes only. For the most accurate
results, please use the model codebase published alongside our 2025 Nature Paper.
Learn more: Our Research | 2025 Nature Paper | Publications