Follow this innovation
[ via 
] [?]
Method to Predict Secreted Cancer Biomarkers
University of Georgia Research Foundation
posted on 06/01/2010
University of Georgia researchers have developed a novel sequence-based approach to classification of proteins. Computational methods have been developed to predict proteins encoded by abnormally and highly expressed genes in tumor cells are predicted to be blood-secreted proteins.
Using global and local characteristics of sequence-derived proteins, researchers have trained a Support Vector Machine (SVM)-based classifier to distinguish experimentally verified blood-secreted proteins from the rest of the human proteins. Prediction results of this classifier are highly promising, providing very useful information that can lead to the use of these proteins as serum biomarkers for various human diseases including cancers.
The researchers have taken their analysis a step further and identified a panel of biomarkers that are present in the bodily fluid of human patients suffering from gastric cancer. Patents are pending on both the method of discovering these secreted biomarkers and the gastric biomarkers themselves.
Suggested Uses
I. Identification of bloodstream biomarkers for cancer diagnosis, typing, and prognosis
II. Easy diagnosis of cancer using blood, urine, or other bodily fluids
Advantages
I. Immediate identification of bloodstream biomarkers from sequence data alone
II. Provides information regarding the location of secreted proteins, unlike previous studies that only show whether the protein is secreted outside the cell
III. Researchers used a Gaussian kernel, which consistently shows superior performance to other kernels used in SVM such as linear and polynomial kernels
IV. Classifiers were chosen to include as many previously unknown blood-secreted proteins as possible, while keeping the specificity high so classifier performance ranged from 87% to 98.6%
Detailed Description
Alterations in gene and protein expression provide important clues about the physiological states of a tissue or organ. During malignant transformation, genetic alterations in tumor cells can disrupt signaling networks, leading to the over-expression of some classes of proteins in ovarian, colorectal, breast, and prostate cancers suggesting that these proteins could be used as markers for cancer diagnosis, typing and staging. Yet, it is very difficult to experimentally detect such secreted proteins. It is thus desirable to investigate computational approaches to predict proteins that are both abnormally highly expressed in cancer tissues and can be secreted into the bloodstream, providing a target list for targeted proteomic work of human blood serum, and making the identification of such marker proteins more realistically solvable.
File Number: 1480, 1491
| Patent Number(s): | 2010017559 |
|---|
This innovation currently is not available for online licensing. Please contact Rachael Widener at University of Georgia Research Foundation for more information.
request more info
ProfileFind more innovations
