Background: Mesenchymal neoplasia comprises a heterogeneous group of tumors with over 200 benign neoplasms and 100 sarcomas. Currently, tumors are classified using histologic and immunocytologic characteristics, with diagnostic error rates reported as high as 40% of cases. As a feasibility study, our goal was to generate a preliminary discriminatory gene list for selected mesenchymal tumors, including sarcomas. This technique may enable an eventual molecular classification schema based on expression profiles that can complement current clinical and pathologic diagnostic procedures in mesenchymal tumors.
Methods: cDNA microarray analyses were preformed on connective tissue tumors obtained at time of surgical resection or biopsy. Messenger RNA (mRNA) from four general tumor classes was competitively hybridized against a human dermal fibroblast cell line comparator and the resulting gene expression profiles processed by ANOVA and linear discriminate analysis.
Results: The tissue classification involved 18 patients with malignant peripheral nerve sheath tumors, giant cell containing tumors, benign spindle cell lesions, or Ewing's family of tumors. Lymph nodes from two patients served comparative purposes. Twenty-five differentially regulated genes considered most variable among the five tissue classes were identified. The tissues were segregated into five classes by linear discriminate analysis.
Conclusions: Linear discriminate analysis of cDNA gene expression profiles partitioned mesenchymal tumor classes, even when constrained by limited sample sizes.