Improved detection of human diseases by statistical evaluation of autoantibody profiles.
For many human diseases, including cancer, no accurate diagnostic tools exist. To perform a diagnosis, molecular tumor markers that are secreted as molecules into body fluids are widely used. A popular example for such molecular markers are tumor antigens that can be found in patients blood sera. Perhaps the most popular tumor antigen is the prostate specific antigen (PSA) that is used to detect prostate cancer. However, even PSA shows an enormous lack of specificity.
We are developing for various cancer entities diagnostic tests with high sensitivity and specificity. Our approach relies on autoantibody profiles rather than on single antigen markers.
Usually, blood of a set of patients and a control group is screened for reactivity with a panel of disease related antigens. The measured autoantibody profiles are statistically evaluated and classified. First, the most relevant antigens are selected by feature subset selection methods and the relevant features are used to perform the diagnosis.
Our results show that our diagnostic framework outperforms classical diagnostic test such that specificity and sensitivity values above the 90% barrier can be well achieved.
The projet's website can be found