Unsupervised machine learning approaches for patient stratification/clustering by phenotype.
Although this is largely an exploratory project rather than an actual package, one may install the development version of PhenoClustR from this repository:
devtools::install_github("Ryan-Laird/PhenoClustR")Disclaimer: This is for research purposes only and does not claim to aid diagnostics or medical decision making. If you are working with sensitive data take precaution to protect PII/PHI. Please refer to the Code of Conduct for user/community standards.
This is an experimental package I’ve developed in an attempt to better
stratify a cohort of patients with undifferentiated autoinflammatory
disease. It leans heavily on the Human Phenotype
Ontology,
ontologyX
R package suite, and HPO2Vec (and
therefore Node2Vec).
Please see bibliography for related research.
The initial findings of this work are to be presented during NIH
Postbac poster
day, 2020. My
poster is available in this repository and was
generated with the wonderful R package
posterdown.
For the most part this will remain an exploratory project, however I do plan on including additional functionality and resources.
To do’s include:
-
Create synthetic autoinflammatory phenotype data for reproducible examples
-
Provide script for my initial analysis (poster)
-
Provide script for generating an enriched HPO node embedding