Software
Open-source software, statistical methods, and reproducibility code developed by the Shyr Lab.
Clinical AI and Biomedical Informatics
RARE-PHENIX
RARE disease PHENotyping with Intelligent eXtraction is an end-to-end artificial intelligence framework for rare disease phenotyping from clinical notes. The workflow extracts phenotype mentions, standardizes them to Human Phenotype Ontology (HPO) terms, and prioritizes diagnostically informative phenotypes.
llmReproScore
A Python package for quantifying the repeatability and reproducibility of stochastic large language model outputs. The package implements semantic and token-distribution-based metrics and includes tools for statistical comparisons across models, prompts, and experimental conditions.
EHR-Based Growth Curves and Centile Charts
An R-based framework for modeling longitudinal growth trajectories and constructing growth centile charts for genetic disorders using electronic health record data. The workflow includes model selection, diagnostic evaluation, and parallelized analyses across genetic conditions.
Causal and Multi-Study Machine Learning
Multi-Study R-Learner
A causal machine learning framework for estimating heterogeneous treatment effects across multiple studies while accounting for between-study heterogeneity. The method can integrate randomized trials, observational studies, or combinations of both.
Multi-Study Boosting
Methods and accompanying R code for studying when heterogeneous datasets should be merged versus modeled separately and ensembled. The framework provides theoretical guidance for multi-study prediction using boosting algorithms.
Research Communication
AI-Generated Lay Summaries
Code for retrieving scientific abstracts and generating lay-language summaries with large language models. This work was developed to improve the accessibility and transparency of research results for study participants and the public.