In Development
DeSurv Platform
Efficient survival estimators pairing de-biased gradients with flexible machine-learning bases for biomarker-rich trials.
ADAPT Statistical Hub
evolveTrial and BATON packages for adaptive monitoring and automated operating-characteristic tuning for the ARPA-H ADAPT platform.
TrialMatch LLM
Fine-tuned language model translating pancreatic cancer eligibility criteria into bilingual summaries with transparent inclusion/exclusion logic.
Pilot Deployment
dlGLM Package
Deep-learning generalized linear models for non-ignorable missing data used in ADAPT analytics and shared-resource consults.
DeCAF Classifier
Transcriptomic classifier of pancreatic tumor and stromal interaction states informing biomarker analyses for PDAC trials.
NIMIWAE Package
Deep-learning imputation methods for EHR labs and longitudinal clinical data supporting Lineberger analytics.
Deployed Software
epigraHMM
Multi-condition epigenomic discovery pipeline used for ATAC/ChIP analyses at UNC Lineberger.
FSCseq + CompDTUReg
Subtype discovery and differential transcript usage suite with reproducible vignettes, used in SPORE core analyses.
PurIST Clinical Pipeline
Single-sample PDAC classifier developed with the Yeh laboratory, with CLIA reporting used in UNC Molecular Diagnostics and PANCREAS trial evaluations.
Collaboration & Access
Projects are co-developed with Gillings trainees, UNC Lineberger oncologists, and cooperative group collaborators. See repositories for code, or email naim@unc.edu about pilot access. Gillings students can participate through BIOS 690 capstone projects and Lineberger internships.