Thyroid nodules are common. It would very helpful if genetic markers that can diagnose malignancy from fine needle aspiration samples were available. Few such markers has been thus identified and none are specific. Large panels of potential markers can be screened by microarray technology. Studies done to date have concentrated on single tumor types and thus provide no help in identifying tumor subtype specific markers. To that end we have studied gene profiles of 5 types of benign and malignant thyroid nodular tissue (multinodular goiter, follicular adenoma, papillary and follicular carcinomas). We have identified 195 genes whose differential expression clustered into clinically relevant groups. Twenty-eight genes were selected for further confirmation using real time quantitative polymerase chain reaction. Despite the differences in the microarray panels used, we confirmed the differential regulation of 12 genes previously reported in thyroid cancer, although we found the expression of several genes to be regulated in other histological tumor subtypes than originally described. We found, PCSK2, TRIB1, RAP1 GA1 to be specifically overexpressed in follicular cancer and S100A4 and GK2 in papillary carcinoma. SERP1, RNASE 2 and STATA5 were suppressed in papillary thyroid cancer. We have thus identified new potential markers specific to malignant thyroid tumors. It is apparent that a range of nodular thyroid tissue using large tumor sample numbers is necessary to establish robust markers for malignancy and to categorize tumors on the basis of small tumor samples.