This is the code repository of the paper Identifying Parkinson's disease subtypes with motor and non-motor symptoms via model-based multi-partition clustering (Nature Scientific Reports).
This project is organized in several folders:
- documentation. It contains information about the MDS-NMS and the MDS-UPDRS, as well as a Codebook with the meaning of the variables considered in the study.
- data. It contains the original data and the transformed data necessary for learning the clustering models.
- src. Main repository of source code. It contains the Java implementations as well as the code necessary for learning the clustering models.
- python-project. Secondary repository of source code. It contains the Python source code necessary for transforming the original data as well as for executing the article's comparative cluster analysis.
- results. It contains the results of the learning scripts (i.e., models in AMIDST *.bn format, models in GENIE *.xdsl format, JSON files with scores and times, and completed data for those algorithms that could work with missing data).
- best-model-genie. For ease of use, it contains the best model in GENIE format. This model is necessary for executing the article's probabilistic inference analysis.
The data provided by this repository should not be used for independent publications without the approval of the Movement disorders society.
- For any enquiries about the project, please email [email protected].
- For any enquieres about the data, please email [email protected].