Me riding the Pokemon Machamp, of course!
Title: Assistant Clinical Professor of Biostatistics
Department: Epidemiology and Biostatistics
Room Number: Biomolecular Sciences 3120D
Email: jrub@umd.edu
I am a Assistant Clinical Professor of Biostatistics at the University of Maryland, College Park School of Public Health and run the MACHine leArning for bioMarker discovery and Prediction (MACHAMP) Lab, in which me and my trainees focus on developing and applying machine learning methods for biomarker discovery and prediction of patient outcomes.
My trainees span diverse backgrounds including public health, epidemiology, cell and molecular biology, molecular pharmaceutics, data science, mathematical statistics, and computer science! We are scattered mainly across the US (with two international trainees - one in Sweden and one in Austrailia!).
We develop and apply machine learning techniques including unsupervised clustering, high-dimensional regression (especially lasso-based approaches), random forest, XGBoost, ensemble learning as well as feature/variable selection and data harmonization approaches.
Please feel free to reach me at jrub@umd.edu if you are interested in getting involved with research in the MACHAMP Lab!
I have primarily worked in the following application areas:
- Biomedical imaging (including renal histopathology, neuroimaging, and vascular imaging)
- Electronic health records (including transplant registry data)
I consider my machine learning specialties to be:
- Unsupervised clustering
- High-dimensional regression (especially lasso-based approaches)
- Random forest
- >XGBoost
- Ensemble learning
- >Feature/variable selection
- Data harmonization
Application and methodological areas I am looking to start new projects include (I welcome new collaborations with folks in these specialties!):
- Extensions of high-dimensional regression models with:
- Unsupervised clustering
- False discovery rate control
- Post-selection inference
- Multimodal data integration
- Conformal prediction
- Federated learning with imaging biomarkers
- Spatial transcriptomics and multiplexed protein imaging
- Multiorgan associations including associations of renal function with cognitive decline and cardiovascular health
Check out the projects in progress or completed by my trainees (or Champs as I sometimes call them to keep with the theme of the lab acronym)!
- Active Projects
- Cluster- and Histologic Object-Aware Pathomic Biomarker Identification for Kidney Function Using Machine Learning (Champ(s): Ketki)
- Scalar-on-Tensor Regression with Unbalanced Tensor Predictors (Champ(s): Alec)
- Integrating Tubule-Level Procurement Biopsy Pathomics and Clinical Factors for Machine Learning Prediction of Delayed Graft Function (Champ(s): Huiqian)
- Scalar-on-matrix Logistic Regression for Unbalanced Feature Matrices (Champ(s): Hedwig)
- Adapting Machine Learning Models Using Multisite Histopathology Data for Predicting Kidney Function (Champ(s): Ritesh, Advay, Raymond, Janelle)
- Flexible Ensemble Learning-based Classification of Post-Transplant Kidney Function Outcomes (Champ(s): Chris)
- Creating Interpretable User Interfaces for Machine Learning Predictions of Post-Transplant Kidney Function Outcomes with Donor Renal Histopathology and Clinical Data (Champ(s): Abby)
- Completed Projects
- A Pathomic-Ensemble Strategy for Exploring Histological Signatures of eGFR Decline in IgAN (Champ(s): Connie, Lylybell, Huiqian)
- A Pathomics-Integrated Approach Toward Improved Prediction of Kidney Survivability Up to 5 Years Post-Biopsy in IgA Nephropathy Patients (Champ(s): Lylybell, Connie)