Method for improving disease diagnosis using measured analytes

A method for preparing a training set model capable of diagnosing a disease, comprising the steps of:

• a) determining concentrations of at least five predetermined analytes in a training set of samples from a group of subjects, wherein at least one of the predetermined analytes is a low abundance protein;
• b) selecting a meta-variable associated with the subjects, which varies in a population associated with the subjects for members of the population who are known either to have or not have the disease, wherein the disease is breast cancer or ovarian cancer;
• c) transforming the concentrations of the analytes as a function of one or more population distribution characteristics and the meta-variable to compute a pseudo-concentration that represents each analyte, wherein the pseudo-concentration is computed as a unitless, clinically predictive measure of each analyte;
• d) preparing the training set model from the pseudo-concentrations determined for members of the populations who are known either to have or not have the disease, wherein the training set model diagnoses the disease by scoring the pseudo-concentrations against the training set model by plotting each pseudo-concentration in a bi-marker plane with data points of the training set model and determining the distance from each pseudo-concentration to the nearest measured data points of the training set model, wherein the training set model utilizes a multidimensional analysis comprising at least five orthogonal axes, wherein false negative and false positive performance of the training set model is at least better than 90%;
• e) diagnosing the disease on the basis of the training set model;
• and
• f) administering one or more anti-cancer drugs capable of treating the indicated disease.