High Throughput Assessment of Biomarkers in Tissue Microarrays Using Artificial Intelligence: PTEN Loss as a Proof-of-Principle in Multi-Center Prostate Cancer Cohorts - Beyond the Abstract

New milestone in assessment of tissue-based biomarkers: PTEN loss as a proof-of principle in improved assessment of disease prognostication using novel biomarker quantification methods and high-throughput automated detection on digital images using artificial intelligence.

Despite prostate cancer being one of the most commonly diagnosed diseases in men, the road to finding the cure is still hindered by several challenges. Successful management of post-operative patients is directly related to the accurate diagnosis and prognostication of prostate cancer. Over the years, many researchers around the world have put forward numerous molecular markers to aid in this process. However, none of these markers have been translated into clinical practice. Lack of conclusive data and subjectivity in the assessment of these molecular markers are some of the main reasons behind this.

Decades of prostate cancer research has shown that phosphatase and tensin homolog (PTEN), a tumor suppressor gene, is a very promising candidate for prognostication and risk assessment of aggressive prostate cancer, and it may aid to augment effective treatment regimens for postoperative patients. Loss of PTEN protein is frequent in aggressive prostate cancer and is associated with unfavorable patient outcomes after definitive local therapy. To demonstrate the utility and usability of PTEN in clinical scenarios, Dr. Jamaspishvili and Dr. Patel in Dr. David Berman’s lab from Queen's University, Canada led a multi-institutional retrospective study involving >500 prostate cancer patients. For the first time, they developed a quantitative method for the objective assessment of PTEN loss and demonstrated its prognostic significance in helping identify prostate cancer patients who may benefit from additional therapy after surgical intervention. The results of this study have been published in the Journal of National Cancer Institute (JNCI).1

However, the research team saw that the approach was too laborious requiring high-quality expert pathologists’ annotations of digital images. Therefore, the researchers developed a less labor-intensive method of analyzing the images that would be easy to implement in clinical practice in the future. This novel study was led by Dr. Jamaspishvili and team (Queen’s University, Canada) in collaboration with Dr. Harmon and colleagues from the National Cancer Institute (NCI) who developed an automated system for PTEN loss detection and localization on digital images using cutting edge artificial intelligence and machine learning technologies. The results from this study were recently published in the Journal of Modern Pathology by Nature.2

The collaborative team of researchers from Queen’s University and national cancer institutes dedicated to incorporating state of the art machine learning and artificial intelligence technology in everyday pathology practices for better and more informative risk stratification and management of cancer patients.

Written by: Tamara Jamaspishvili, MD, PhD, Department of Pathology & Molecular Medicine, Division of Cancer Biology & Genetics, Cancer Research Institute, Queen's University, Kingston, Ontario, Canada


  1. Jamaspishvili, Tamara, Palak G. Patel, Yi Niu, Thiago Vidotto, Isabelle Caven, Rachel Livergant, Winnie Fu et al. "Risk stratification of prostate cancer through quantitative assessment of PTEN loss (qPTEN)." JNCI: Journal of the National Cancer Institute (2020).
  2. Harmon, Stephanie A., Palak G. Patel, Thomas H. Sanford, Isabelle Caven, Rachael Iseman, Thiago Vidotto, Clarissa Picanço et al. "High throughput assessment of biomarkers in tissue microarrays using artificial intelligence: PTEN loss as a proof-of-principle in multi-center prostate cancer cohorts." Modern Pathology (2020): 1-12.
Read the Abstract