The tumor classification is based on the primary organ of the cancer development and, in cases of metastatic tumors, finding the primary origin site is mandatory, once each cancer treatment and therapy is totally dependent of its the primary origin site. Immunohistochemistry and imaging workflows are able to classify around 95% of metastasis cases but the subjective interpretation among pathologists introduces a non-trackable bias in the final results. The other 5% of cases are classified as Cancer of Unknown or Uncertain Primary Origin (ICD10 C76-C80), the 4th more lethal and 8th more frequent cancer type. In this work, we present the development of an algorithm which uses a Gene Expression Profiling approach to classify the tumor types. We analyzed 6544 microarray samples files and built a database of 64 known cancer types and subtypes (primary and metastatic).
展开▼
