A Swiss study in prostate cancer, published in PNAS, presents a new two-part framework for the rational discovery of biomarkers.

There is recognised to be an urgent and critical medical need to identify novel molecular and genetic biomarkers to help both predict the people who will develop cancer and to detect the disease at an early stage. In prostate cancer the current standard biomarker for early detection is PSA, which is controversial due to a lack of sensitivity and specificity resulting in considerable over diagnosis and treatment. The discovery of serum biomarkers has proved technically challenging for reasons including the variable composition of the serum proteome and it enormous complexity.

“As the genetic alterations that cause cancer are becoming better understood, one strategy to overcome the limitations of traditional serum proteome comparisons is to use the knowledge about specific cancer-causing mutations and the underlying disrupted signalling pathways to guide the discovery of novel cancer serum biomarkers,” write the authors, Wilhelm Krek and colleagues from Eidgenössische Technische Hochschule (Zurich, Switzerland).

In the first stage of the study, Krek and colleagues undertook a comparison of serum proteins taken from wild-type mice and mice in which the phosphate and tensin homolog (PTEN) tumour suppressor gene had been inactivated. PTEN, which is commonly inactivated in many human cancers, is known to be a biomarker of prostate cancer.  An established consequence of PTEN inactivation is the activation of the PI3K signalling pathway which drives uncontrolled cell growth, proliferation and survival. 

From the mice studies the team were able to show that PTEN inactivation leads to measurable differences of the prostate and serum glycoproteome between the two groups.  The candidate biomarker list, which was further narrowed using PTEN dependency, prostate specificity and serum detection ability criteria, was 126 proteins.

In the second verification phase of the study, the investigators tested whether PTEN inactivation in human prostate cancer was associated with a specific serum signature.  In the study, serum and prostate tissue samples were collected from 143 patients with prostate cancer and 66 controls with benign prostatic hyperplasia.  Then using machine learning algorithms the team were able to identify 39 ortholog human proteins.

Additional bioinformation prioritisation of the potential signatures led to a four-protein diagnostic signature that showed 79% specificity and 85% sensitivity in distinguishing between patients with prostate cancer and those with benign prostatic hyperplasia.

“The ability of our four-protein signature for prostate cancer diagnosis to distinguish accurately between locPCa and BPH makes it potentially suited for screening tests by reducing false-positives and therefore avoiding anxiety and biopsies in men who have an elevated PSA but do not harbour cancer,” write the authors.

Reference

I Cima, R Schiess, P Wild. Cancer genetics-guided discovery of serum biomarker signatures for diagnosis and prognosis of prostate cancer. PNAS.  Doi: 10.1073/pnas.1013699108.