RoBERTA-based sequence classifier modeling on a novel natural language dataset achieving a macro-averaged F1 score of 0.88 with an AUC/ROC of .99.
The full repository with notebooks can be found on my GitHub repo.
Ordinal regression to predict used car purchase/appraisal prices in an obfuscated and highly incomplete dataset using iterative imputation methods and tree-based regression methods. Here I achieved an R^2 of 0.91 and concordance score of 0.95 without excluding any incomplete samples.
The full repository with notebooks can be found on my GitHub repo.
I performed bulk and single-cell transcriptomic analyses of KSHV infected lymphatic endothelial cells, which are the best available de novo infection for modeling Kaposi’s sarcoma. The results of this analysis are still largely confidential due to the competitive nature of this topic right now. A publication is in preparation pending some molecular imaging and mechanistic follow-up studies based on these results. Click above for a short discussion on some of the challenges this dataset presented and some of the methods employed to overcome them.
My PhD dissertation (get PDF copy here), this work is largely published in Cell Death and Differentiation in which I make use of genome-wide synthetic CRISPR rescue screens alongside traditional low-throughput biological methods to identify novel regulators of cFLIP and ligand-independent TRAIL-R1-mediated cell death.
See my publications section for more details on this project.