To the editor:

We read with interest the article by Fabian et al. in the March 2005 issue (Nat. Biotechnol. 23, 329−336, 2005) entitled “A small molecule–kinase interaction map for clinical kinase inhibitors.” Although the study is an elegant attempt at rapidly determining kinase-specificity profiles of small-molecule inhibitors, we wish to raise some concerns with the paper that may compromise its broad utility.

Our most serious concern is that the method might not be sufficiently quantitative and statistically validated to support some of the conclusions made. For example, the authors report a commercially performed enzyme assay for validation, but this was only for two kinases (epidermal growth factor receptor (EGFR) and vascular endothelial growth factor receptor (VEGFR-2)) and one compound ZD6474. The fact that they obtained results that were contrary to literature reports with this kinase pair and ZD6474 should have prompted more extensive comparative assays using more compounds against more kinases, but apparently they did not. To state that there is a good correlation between their results and the literature kinase assay results (Supplementary Table 4 online in their paper) is not good enough. The correlation coefficient is not given. Furthermore, there are no error bars for any of the values presented that will give a sense of precision, and it is said that experiments were done at least two times. This constitutes a woeful lack of statistical validation. The authors made other attempts at validation, but did not provide unequivocal evidence on sensitivity and precision. This is a major flaw for such a potentially far-reaching study.

In view of these concerns, to convey the impression to readers that the higher sensitivity of inhibition of the EGFR variants that confer clinical efficacy as reported in the literature1,2 has no dependence on binding affinity changes at the ATP site may be misleading. In contrast to this notion, the literature reports1,2 show about tenfold higher sensitivity of some of the mutants, especially the L858R mutant towards inhibition by gefitinib (Iressa) or erlotinib (Tarceva) than the wild-type EGFR. It appears that the authors' assay does not have the ability to discriminate inhibitor sensitivity differences less than 50-fold. The authors admit that in their assay the inhibitors are binding the non-activated forms of the kinases, which will have different binding affinities compared with binding at activated kinase ATP sites.