Evaluation of quantitative variation in gene expression

We investigate the behaviour of the gene-expression rate as a statistical variable using autoradiographic data for 39 transcripts from a heterogeneous set of 80 breast-tissue cultures. Despite standardization, the data distributions of all transcripts showed intervals of normality and intervals of systematic departure from normality which most frequently resulted in a significant skewness and/or kurtosis. Non-normal shapes are attributed to modulation of gene expression. This statistical particularity creates difficulties in the evaluation of differences among specimens. Using classical parametric and non-parametric procedures for normal and non-normal variation, respectively, we demonstrate that large differences in optical density are neither necessary nor sufficient for associating expression rates with biological factors. The transcripts coding for the metalloprotease stromelysin-3 (ST3) and for the receptor to insulin-like growth factors (IGFR) are used as examples and their variation is presented in detail. ST3 expression appeared to be specifically associated with mammary stroma fibroblasts derived from post-radiation fibrosis lesions. IGFR was expressed at higher rates in mammary gland and skin fibroblasts than in mammary epithelial cells and was subject to frequent and strong modulation.