Investigating intratumoral and microenvironmental heterogeneity through computational pathology approaches
Principal Investigator
Name
Eliezer Van Allen
Degrees
M.D.
Institution
Dana-Farber Cancer Institute
Position Title
Chief, Division of Population Sciences, Dana-Farber Cancer Institute
Email
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCOI-1819
Initial CDAS Request Approval
Feb 10, 2025
Title
Investigating intratumoral and microenvironmental heterogeneity through computational pathology approaches
Summary
Localized prostate cancer (PCa) can have highly variable clinical trajectories. Patients with localized PCa are risk stratified using clinical and pathologic factors. However, conventional risk stratification systems have suboptimal performance, leading to undertreatment for some patients and overtreatment for others. Paralleling the highly variable clinical course of localized PCa, the genomic and histologic landscape also demonstrates significant variability.
While significant prior work has examined the relationship between these tumor-intrinsic features and clinical states, there is a nascent understanding of the roles that the tumor microenvironment (TME) plays in PCa. The localized PCa TME is generally thought to be immunosuppressive. However, some studies suggest that a proportion of localized PCa samples are more immunogenic and enriched in certain immune cell populations, with unclear clinical implications.
Investigating tumoral and microenvironmental heterogeneity may provide further insights into the biological and clinical behavior of aggressive localized PCa. Detailed analysis of PCa tumoral and microenvironmental heterogeneity is typically performed using genomic profiling methods or multiplex tissue imaging, which are resource-intensive.
Recently, advances in computational power and artificial intelligence (AI) have created new opportunities to improve our understanding of localized PCa. These include computational pathology, or the use of computational approaches to analyze digitized histopathology slides. Preliminary results from our lab, as well as published work from other groups, suggest that computational pathology methods may allow for systematic analysis of intratumoral and microenvironmental heterogeneity in clinically acquired specimens.
In this project, we propose to use the PLCO prostate cancer dataset to enable the development, validation, and application of computational pathology and AI approaches, paired with associated clinical data, to advance our understanding of tumoral and microenvironmental heterogeneity of localized PCa and its relationship with oncologic outcomes. The PLCO dataset is unique in being a large clinical set with long follow-up (of particular interest given the long natural history of localized PCa). The successful completion of this proposal would increase our understanding of PCa heterogeneity and further precision medicine for patients with PCa by identifying novel biomarker strategies to improve risk stratification.
While significant prior work has examined the relationship between these tumor-intrinsic features and clinical states, there is a nascent understanding of the roles that the tumor microenvironment (TME) plays in PCa. The localized PCa TME is generally thought to be immunosuppressive. However, some studies suggest that a proportion of localized PCa samples are more immunogenic and enriched in certain immune cell populations, with unclear clinical implications.
Investigating tumoral and microenvironmental heterogeneity may provide further insights into the biological and clinical behavior of aggressive localized PCa. Detailed analysis of PCa tumoral and microenvironmental heterogeneity is typically performed using genomic profiling methods or multiplex tissue imaging, which are resource-intensive.
Recently, advances in computational power and artificial intelligence (AI) have created new opportunities to improve our understanding of localized PCa. These include computational pathology, or the use of computational approaches to analyze digitized histopathology slides. Preliminary results from our lab, as well as published work from other groups, suggest that computational pathology methods may allow for systematic analysis of intratumoral and microenvironmental heterogeneity in clinically acquired specimens.
In this project, we propose to use the PLCO prostate cancer dataset to enable the development, validation, and application of computational pathology and AI approaches, paired with associated clinical data, to advance our understanding of tumoral and microenvironmental heterogeneity of localized PCa and its relationship with oncologic outcomes. The PLCO dataset is unique in being a large clinical set with long follow-up (of particular interest given the long natural history of localized PCa). The successful completion of this proposal would increase our understanding of PCa heterogeneity and further precision medicine for patients with PCa by identifying novel biomarker strategies to improve risk stratification.
Aims
Aim 1: Assess the generalizability and robustness of computational pathology methods across diverse cohorts.
Aim 2: Evaluate the relationship between spatial tumoral heterogeneity and oncologic outcomes in localized PCa.
Aim 3: Investigate the relationship between spatial cellular organization in the TME and oncologic outcomes in localized PCa.
Aim 4: Determine the relationship between latent histologic feature heterogeneity and oncologic outcomes in localized PCa.
Collaborators
David Yang, M.D. Dana-Farber Cancer Institute
Brendan Reardon Dana-Farber Cancer Institute