PURPOSE

Recent advances in combinatorial chemistry and high throughput screens for pharmacologic activity have created an increasing demand for in vitro high throughput screens for toxicological evaluation in the early phases of drug discovery.

METHODS

To develop a strategy for such a screen, we have conducted a data mining study of the National Cancer Institute's Developmental Therapeutics Program (DTP) cytotoxicity database.

RESULTS

Using hierarchical cluster analysis, we confirmed that the different tissues of origin and individual cell lines showed differential sensitivity to compounds in the DTP Standard Agents database. Surprisingly, however, approaching the data globally, linear regression analysis showed that the differences were relatively minor. Comparison with the literature on acute toxicity in mice showed that the predictive power of growth inhibition was marginally superior to that of cell death.

CONCLUSIONS

This datamining study suggests that in designing a strategy for high throughput cytotoxicity screening: a single cell line, the choice of which may not be critical, can be used as a primary screen; a single end point may be an adequate measure and a cut off value for 50% growth inhibition between 10(-6) and 10(-8) M may be a reasonable starting point for accepting a cytotoxic compound for scale up and further study.