Statistical and blind neural analyses of the expression of the metastasis-associated genes h-mts1 (S100A4) and Nm23 in carcinoma of the breast

The murine 18A2/mts1 and its human homologue h-mts1 (S100A4) encoding a Ca/sup 2+/-binding protein belonging to the S-100 family have been previously shown to be associated with high invasive and metastatic potential of murine tumours, human tumour cell lines in vitro and of human tumours growth as xenografts. Nm23 is a putative metastasis suppressor gene whose expression has been found to correlate inversely with metastatic potential of some forms of human cancer. In view of the equivocal association of nm23 with the metastatic potential of human cancer, we suggested that the relative expression of h-mts1 and nm23 might reflect tumour progression more accurately than either of them alone. We show here that high h-mts1 expression is associated with metastatic spread to the regional lymph nodes. The expression of nm23 on its own did not show a statistically significant inverse correlation with nodal spread. However, the expression status of the two genes, taken together, correlated strongly with the occurrence of nodal metastases. The clinical data together with the state of expression of steroid receptors and the expression levels of h-mts1 and Nm23 genes were analysed using artificial neural networks for accuracy of prediction of nodal spread of the carcinomas.