The specificities of 28 commercially available compounds reported to be relatively selective inhibitors of particular serine/threonine-specific protein kinases have been examined against a large panel of protein kinases. The compounds KT 5720, Rottlerin and quercetin were found to inhibit many protein kinases, sometimes much more potently than their presumed targets, and conclusions drawn from their use in cell-based experiments are likely to be erroneous. Ro 318220 and related bisindoylmaleimides, as well as H89, HA1077 and Y 27632, were more selective inhibitors, but still inhibited two or more protein kinases with similar potency. LY 294002 was found to inhibit casein kinase-2 with similar potency to phosphoinositide (phosphatidylinositol) 3-kinase. The compounds with the most impressive selectivity profiles were KN62, PD 98059, U0126, PD 184352, rapamycin, wortmannin, SB 203580 and SB 202190. U0126 and PD 184352, like PD 98059, were found to block the mitogen-activated protein kinase (MAPK) cascade in cell-based assays by preventing the activation of MAPK kinase (MKK1), and not by inhibiting MKK1 activity directly. Apart from rapamycin and PD 184352, even the most selective inhibitors affected at least one additional protein kinase. Our results demonstrate that the specificities of protein kinase inhibitors cannot be assessed simply by studying their effect on kinases that are closely related in primary structure. We propose guidelines for the use of protein kinase inhibitors in cell-based assays.