Contrasting patterns of transcript abundance in tumour tissue and cancer cell lines

Comparison of data on transcript abundance in ovarian, prostate and colon tumours with the corresponding cancer cell lines was used to assess the similarities of expression profiles. Although transcript abundances in tumours and cell lines were positively correlated, there were substantial differences with respect to the overall expression pattern. Compared with tumours, cancer cell lines showed more variable patterns of transcript abundance among tissue types. In the ovary and colon, cancer cell lines showed greater overall transcript abundance than normal tissue; this increase was much more marked in the case of the colon. However, in the prostate, cancer cell lines showed overall reduced transcript abundance when compared with normal tissue. Principal component analyses, applied separately to each tissue type, showed that approximately 80% of the variance was explained by overall expression level differences, which were maintained across normal tissue, tumour tissue and cancer cell lines. The remaining variance ( approximately 20%) could be attributed to contrasts in expression pattern among normal tissue, tumour tissue and cancer cell lines. In each dataset and in a combined dataset of transcripts shared among the three datasets, principal components revealed both contrasts in expression pattern between tumour tissue and cancer cell lines, and common features in the expression pattern of cancer cell lines that were distinct from those of tumour tissue and were shared across the different tissue types. These results imply that data on gene expression in cancer cell lines should be used with caution in inferring gene expression of in vivo tumours.