Early diagnosis of cancer has been shown to substantially improve 5-year survival rates for many cancer types. With current methodologies early diagnosis has proven difficult for cancers of deep tissues, such as the pancreas and lung. However, whole peripheral blood has been demonstrated to be a promising non-invasive surrogate tissue for the detection of many types of cancer. Blood samples were collected in EDTA blood collection tubes from 2,485 human volunteers and gene expression profiled with the goal to develop a classification model that differentiations patients with cancer from those without cancer. Overall design: A set of 1,013 unique patient blood samples spanning 11 types of cancer or pre-cancer (colorectal polyps) of interest and 1,832 unique control samples, including those with autoimmune and cardiovascular disease, were collected and profiled on Affymetrix U133 Plus 2.0 GeneChips. A multi-cancer classification model was generated using logistic regression to classify sample profiles into the 11 different cancer classes.