Paul Bernard Romesser

This paper introduces Kodiak, a discrete representation learning method for scientific imaging that improves representation quality and enables interpretable, reusable concepts across domains.

P-selectin-targeted fucoidan nanoparticles carrying Navitoclax or dBET6 deplete a senescent-like, lipid-associated macrophage subset, reverse mouse liver and lung fibrosis with less toxicity, and restore checkpoint therapy response in fibrotic tumors.

This study identifies a dietary lipid axis driven by linoleic acid and FADS2 that suppresses antitumor immunity in liver cancer, offering a therapeutic target for immunotherapy resistance.

In mouse pancreatic cancer models, p53, CDKN2A and SMAD4 programs converge on a rare Kras-mutant progenitor-like state whose niche KRAS inhibition collapses and p53 loss expands.

The paper develops Mosaic, a machine learning system that classifies tumors into detailed subtypes and identifies molecular phenotypes using histopathologic and genomic data from a large patient cohort.

The paper investigates how aging-dependent dysregulation of EXOSC2 affects cellular traits and finds that PPP2R5E, a PP2A phosphatase subunit, is a key target of EXOSC2, with this regulation maintained in stem cells and cancer.

This study integrates DNA methylation profiles from tumors and organoids to identify CpGs predictive of chemotherapy response in rectal cancer, finding a methylation risk score significantly predicts progression-free survival.