Proteomics for cancer biomarker discovery

Abstract

Background: If we are to successfully treat cancer, we must understand the biologic underpinnings in conjunction with early diagnosis. Genome-wide expression studies have advanced the research of many cancers. Nevertheless, understanding which genes are expressed in a tumor is not equivalent to knowing which proteins are being produced. Proteomics hold great promise for careful examination of the proteins in complex biologic fluids and tissues, and it may be possible to detect disease from a patient's serum, long before it would otherwise be clinically evident. Although there have been steady advances in all the steps of a proteomic analysis, much remains to be standardized. Because of some high-profile problems with the initial analysis of ovarian cancer proteomic data, early exuberance has now been tempered and replaced by a more methodical approach to these studies. Hypothesis: My hypothesis in this thesis is that proteomics is a valuable tool in the diagnosis and study of cancer, as will be demonstrated in several steps. Methods: First, I describe the current field of proteomics, specifically as it applies to early detection of cancer and biomarker discovery.

(cont.) I lay out the current state-of-the-art technologies for preparing samples and enumerating the proteins in complex fluids and tissues, giving special treatment to the main threats to validity-chance and bias. I also describe the bioinformatic tools necessary for analyzing the large amounts of data produced. Through the example of a mouse model of colorectal carcinoma, I demonstrate the steps involved in a proteomic study, from procuring samples to peptide and protein determination to bioinformatic analysis. Finally, I discuss these findings in light of the proteomic considerations discussed earlier. Results: From this work, I discovered that proteomic profiling can describe the proteins in serum from mice both with and without colon cancer. Furthermore, I developed a naive Bayes classifier that could distinguish between the serum of mice with colorectal carcinoma and their normal litter-mates. Contributions: Through this work, I have contributed the following. I described the field of proteomics with special emphasis on cancer biomarker discovery and early detection. I enumerated the challenges and pitfalls to developing early detection schemes for cancer based on high-dimensional proteomic analyses.

(cont.) I described a set of experiments on mice harboring a gene mutation that predisposes them to colorectal carcinoma. I detailed the bioinformatic analysis of this data, including the development of a naive Bayes classifier to differentiate the cancerous state from the normal state. Finally, I discussed the caveats of the current work, in reference to the initial discussion on the challenges and pitfalls of early detection schemes and cancer biomarker discovery.