Sebastian Medina
BME PhD Proposal Presentation

Date: 2026-11-04
Time: 11:00 am ~ 12:00 pm
Location / Meeting Link: HSRB II N357, https://emory.zoom.us/j/97793833880?from=addon

Committee Members:
Anant Madabhushi, PhD (Advisor); Peng Qiu, PhD; Drew Williamson, MD; Ravi B. Parikh MD; Ahmet Coskun, PhD

Title: Computational Pathology Biomarkers in Androgen Pathway Modulator Sensitive Prostate Cancer

Abstract:
Prostate cancer accounted for an estimated 1.47 million new cases and 397,000 deaths worldwide in 2022. Although most men are diagnosed with localized disease that is potentially curable, the risk of death is concentrated in high-risk localized and metastatic disease, where androgen deprivation therapy (ADT) is intensified with an androgen receptor pathway inhibitor (ARPI), docetaxel, or radiotherapy. In metastatic androgen pathway modulator sensitive (APMS) prostate cancer, docetaxel is added to ADT and an ARPI according to the volume and timing of metastases, criteria that are prognostic but do not measure the tumor biology that determines response. Between 40% and 50% of men selected in this way derive no survival benefit from docetaxel while experiencing its toxicity. Molecular states associated with progression and treatment response can be measured by genomic profiling, an assay performed on additional tissue in a minority of patients. These states also alter tumor morphology on the hematoxylin and eosin (H&E) slide prepared at diagnosis, which in routine practice is assessed only through visual grading of glandular architecture. The objective of this dissertation is to determine whether the diagnostic H&E slide can provide quantitative measurements of tumor biology to guide treatment selection in prostate cancer. We hypothesize that this biology is expressed in H&E morphology and can be quantified computationally. In Aim 1, we will develop methods to quantify tumor-immune spatial architecture and nuclear morphologic heterogeneity, and to infer molecular states from tissue morphology by relating histologic appearance to spatially resolved gene expression and to tumor genomic profiles. In Aim 2, we will evaluate whether the morphological features predict benefit from docetaxel and enzalutamide through treatment-by-biomarker interaction analyses in three completed randomized phase 3 trials. In Aim 3, we will validate the inferred molecular states against genomic, protein, and spatial transcriptomic measurements, compare their prognostic value with that of the Decipher genomic classifier, and evaluate whether their distribution within the tumor is associated with treatment response and benefit. The expected outcomes are a histologic biomarker of treatment benefit validated in completed phase 3 trials and an estimate of molecular state, and of its distribution within the tumor, from routine histology. With these measurements, treatment selection in prostate cancer could be informed by tumor biology, including for patients without access to genomic testing.