Research Papers:

Prostate cancer diagnosis using epigenetic biomarkers, 3D high-content imaging and probabilistic cell-by-cell classifiers

Darko Stefanovski, George Tang, Kolja Wawrowsky, Raymond C. Boston, Nils Lambrecht and Jian Tajbakhsh _

PDF  |  HTML  |  Supplementary Files  |  How to cite

Oncotarget. 2017; 8:57278-57301. https://doi.org/10.18632/oncotarget.18985

Metrics: PDF 1831 views  |   HTML 2299 views  |   ?  


Darko Stefanovski1, George Tang2, Kolja Wawrowsky3, Raymond C. Boston1, Nils Lambrecht4,5 and Jian Tajbakhsh2,6

1Department of Clinical Studies, School of Veterinary Medicine, University of Pennsylvania, Philadelphia, PA, USA

2Translational Cytomics Group, Department of Surgery, Cedars-Sinai Medical Center, Los Angeles, CA, USA

3Department of Biomedical Sciences, Cedars-Sinai Medical Center, Los Angeles, CA, USA

4Pathology and Laboratory Medicine Service, Veterans Affairs Medical Center, Long Beach, CA, USA

5Department of Pathology and Laboratory Medicine, University of California Irvine, Orange, CA, USA

6Samuel Oschin Comprehensive Cancer Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA

Correspondence to:

Jian Tajbakhsh, email: [email protected]

Keywords: prostate cancer, epigenetics, 3D high-content imaging, tissue diagnostics, cell heterogeneity

Received: September 07, 2016    Accepted: June 02, 2017    Published: July 05, 2017


Background: Prostate cancer (PCa) management can benefit from novel concepts/biomarkers for reducing the current 20-30% chance of false-negative diagnosis with standard histopathology of biopsied tissue.

Method: We explored the potential of selected epigenetic markers in combination with validated histopathological markers, 3D high-content imaging, cell-by-cell analysis, and probabilistic classification in generating novel detailed maps of biomarker heterogeneity in patient tissues, and PCa diagnosis. We used consecutive biopsies/radical prostatectomies from five patients for building a database of ~140,000 analyzed cells across all tissue compartments and for model development; and from five patients and the two well-characterized HPrEpiC primary and LNCaP cancer cell types for model validation.

Results: Principal component analysis presented highest covariability for the four biomarkers 4’,6-diamidino-2-phenylindole, 5-methylcytosine, 5-hydroxymethylcytosine, and alpha-methylacyl-CoA racemase in the epithelial tissue compartment. The panel also showed best performance in discriminating between normal and cancer-like cells in prostate tissues with a sensitivity and specificity of 85%, correctly classified 87% of HPrEpiC as healthy and 99% of LNCaP cells as cancer-like, identified a majority of aberrant cells within histopathologically benign tissues at baseline diagnosis of patients that were later diagnosed with adenocarcinoma. Using k-nearest neighbor classifier with cells from an initial patient biopsy, the biomarkers were able to predict cancer stage and grade of prostatic tissue that occurred at later prostatectomy with 79% accuracy.

Conclusion: Our approach showed favorable diagnostic values to identify the portion and pathological category of aberrant cells in a small subset of sampled tissue cells, correlating with the degree of malignancy beyond baseline.

Creative Commons License All site content, except where otherwise noted, is licensed under a Creative Commons Attribution 4.0 License.
PII: 18985