Classification of Prostate Cancer using Serum at the John Van Geest Cancer Research Centre, Nottingham, UK.

Tumour Trace technology classifies healthy/cancerous prostate serum.

Research Study was conducted by

John van Geest Research Centre Team: Prof. R. Rees, Prof. G.Ball, Dr D. Boocock, Dr. M. Ahmed and C. Coveney

Nottingham Trent University: Dr.G. Shahtahmassebi,

Tumour Trace Ltd Team: Prof. Dr Djuro Koruga, Chief Scientist, Aleksandra Dragicevic, MSc, research assistant.

Period of study

March 21-23rd 2015


John van Geest Research Centre

In 2008, the John and Lucille van Geest Foundation endowed Nottingham Trent University with a research grant of £8 million to establish a dedicated cancer research facility. Nottingham has been a focus of cancer research for over 40 years and our centre cements that legacy.

The opening of this purpose-built premises at NTU’s Clifton campus in 2010 allowed the University to focus on two key approaches to the treatment of patients with cancer.

  • Improving the diagnosis and management of breast and prostate cancers.
  • Developing effective vaccines and immunotherapies that will significantly improve the survival rates and quality of life for cancer sufferers.

The team of dedicated and world class scientists’ work with state-of-the art equipment and facilities; enabling us them successfully meet the significant challenges of modern day cancer research.

Existing Methods

Existing diagnostic accuracy of prostate cancer using serum, based on free PSA (Prostate-Specific Antigen, is 63 % [1])


67 Prostate serum samples were collected from UK hospitals or prepared in the JvG Centre [1] and stored in refrigerators on -800C. Samples were separated into three main groups: Healthy (H), Benign (B) and Cancer (C). Next numbers were separated into sub-groups; training set (TS) and predicting set (PS) for Naïve Bayes classifier and for Monte Carlo Cross validation classifications.

  Total TS PS
Healthy 30 24 6
Benign 9 6 3
Cancer 28 24 4


Classifier Accuracy Notes Analyser
Naïve Bayes


88.00% Based on training set (TS) and predicted set (PS):

TS – 85-90%, PS -10-15%. Data are randomized

Nottingham Trent University (NTU), Department of mathematics
Monte Carlo

Cross Validation



Based on training set (TS), stopping set (SS) and predicted set (PS). TS-60%, SS-20%, PS-20%. John van Geest Cancer Research Centre (JvG CRC)
  1. Tumour Trace algorithm classified Healthy and Cancer samples with an accuracy of 72% using Monte Carlo Cross Validation algorithm, while Naïve Bayes classifier with randomised data has accuracy 88%.
  2. Tumour Trace results were better than standard PSA technique by 10%, according to the Monte Carlo Cross Validation algorithm, or 25% according to the Naïve Bayes classifier.
  3. This initial study strongly indicates that the Tumour Trace method for screening, monitoring and diagnosis of prostate cancer can be used but more research should be done in order introduce this method into clinical practice.

1] Partin WA et al, Analysis of procent free prostate-specific antigen (PSA) for prostate cancer detection: Influence of total PSA, prostate volume, and age, Urology, 48 (6A):  pp.55-61, 1996.

Download a pdf of this case study