Proteomic Analysis of the Human Serum Proteome for Population Screening, Diagnosis and Biomarker Discovery

A joint project between University College London, the Ludwig Institute for Cancer Research and the CLRC

Project Main Goals

• Identify and analyse disease-specific biomarkers via proteomic technology and mass spectrometry (MS)
• Asses whether genomic and proteomic profiles correlate in predicting disease risk and clinical outcome in ovarian and breast cancer, and coronary artery disease
• Identify common low-moderate risk epithelial ovarian cancer (EOC) susceptibility genes
• Establish the significance of proteomic profiles and high-risk gene mutation status in ovarian and breast cancer
• Establish the association between proteomic bio-markers in EOC development and metastatic progression

Primary Participating Organizations

Key Issues

There is broad agreement that analysis of the human serum proteome has great potential for diagnosis and early detection of human disease. The challenges are immense given the complexity of the human proteome and the broad dynamic range in abundance of individual proteins. The key to unlocking this potential is the development of reproducible, sensitive and specific technology for proteomic analysis. Recent advances in technology suggest that this may now be feasible.

Several groups, including our own, have demonstrated the potential power of proteomic techniques for the discovery of novel disease biomarkers from complex body fluids. In the past two years, a number of studies have investigated the use of mass spectrometry based methods for early cancer detection from human serum through "proteomic pattern diagnostics." These studies have been criticised because of inconsistencies between the biomarkers identified in similar studies and when compared with classical immunological testing as well as the choice of cancer and control samples. It is also clear that there is insufficient data about the effect of different methods of sample handling on this technology as well as its analytical reproducibility, sensitivity and specificity. Finally, most of the work to date has focused on pattern recognition with little effort directed at establishing the identity of specific biomarkers.

Technical Approach

Subjects and Samples
The primary source fo samples for this research programme will be the unique serum bank collected via the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS). This trial involves 200,000 apparently healthy postmenopausal women aged 50-74 years of age. All participants provide a serum sample at registration and 50,000 participants give additional samples annually for 6 years. Large numbers of samples will be available from women who have developed each of the common cancers (eg. colorectal, lung and ovary) and common non-malignant disorders (eg. cardiac disease, stroke).

Proteomic Technology
Several serum preparation methods will be investigated during Phase I of this study. Among them are surface-enhanced laser desorption/ionisation (SELDI, Ciphergen Biosystems), reproducible microfluidic sample clean-up by the Gyrolab worksation using chromatographic microstructures and electrospray ionisation (ESI). We will mainly use existing proteomic technologies and methods to define the analytical sensitivity and variability of these approaches for optimal diagnostic pattern analysis of human serum. However, emerging improvements will only be implemented if their implimentation is obvious, facile and fast.

Bioinformatics
The scale of data collection and dimensionality generated by proteomic technology requires non-convential data analysis. We plan to use the Support Vector Machine (SVM) and Confidence Machines along with the development of other learning algorithms for "proteomic pattern diagnostics."

Expected Achievments/Impact

A series of studies will be performed to asses and define:

• The 'normal' proteomic pattern in healthy postmenopausal women
• Variations in the pattern related to physiological factors such as age and years post menopause
• Variability over time in serial samples from the same healthy postmenopausal women
• The effect of differences in time to serum separation and ambient temperature
• The reproducibility in terms of intra/inter assay precision fo the proteomic technology
• The optimal data handling and processing of the proteomic data for subsequent pattern analysis

Data Analysis Algorithms

The project has resulted in the development of a number of algorithms and programs. When using the algorithms and programs for research or teaching purposes, all documents must be properly cited. The main reference is:

A.Gammerman et al. "Serum Proteomic Abnormality Predating Screen Detection of Ovarian Cancer", The Computer Journal Advance Access first published online on April 4, 2008. This version published online on April 15, 2008 The Computer Journal, doi:10.1093/comjnl/bxn021

• System Documentation, pdf (Pilot study)
• Matlab system code (Pilot study)

Reports

• Seven bio-markers report
• UKOPS report
• UKOPS Supplementary results 1
• UKOPS Supplementary results 2
• UKCTOCS Breast Cancer report
• UKCTOCS Heart Disease report
• UKCTOCS Ovarian Cancer report

Slides

• Ebeads
• UKOPS October 16th 2008