Hi all,
Excited to share PIE Workbench with the Data Modality & Methodology Task Force and the MJFF community. It’s a desktop app that brings classical statistical analysis and state-of-the-art machine learning together under one roof, with no code required.
The goal is to give clinical researchers and data scientists working in PD a fast path from raw PPMI data to publishable results, without bouncing between tools or writing Python. In one place, you can:
- Load and inspect PPMI data with automatic modality detection and missingness heatmaps
- Run t-tests, ANOVA, chi-square, correlations, and survival analysis (Kaplan-Meier + log-rank)
- Build end-to-end ML pipelines: target selection, feature selection, leakage controls, model comparison
- Get interactive ROC curves, confusion matrices, and feature importance
Under the hood, PIE is now powered by endgame, my ML library, which means PIE inherits a serious model zoo, calibration, leakage-aware evaluation, and a consistent reporting layer for free, and improvements to endgame flow straight through to PIE.
I’d love for folks to take it for a spin and send feedback: what works, what’s confusing, what’s missing. Issues on GitHub or via email are welcome.
What’s next:
- Bringing imaging data into the supported modalities
- Putting PIE to work inside an actual PD research study, end-to-end
Thanks,
Cameron