Motivation and method: Small-molecule inhibitors targeting the ATP binding pocket of the catalytic domain of protein kinases have potential to become drugs devoid of (major) side-effects, particularly if they bind selectively. Here, the sequences of the 518 human kinases are first mapped onto the structural alignment of 116 kinases of known three-dimensional structure. The multiple-structure alignment is then used to encode the known strategies for developing selective inhibitors into a fingerprint. Finally, a network analysis is used to partition the kinases into clusters according to similarity of their fingerprints, i.e., physico-chemical characteristics of the residues responsible for selective binding. RESULTS: For each kinase the network analysis reveals the likelihood to find selective inhibitors targeting the ATP binding site. Systematic guidelines are proposed to develop selective inhibitors. Importantly, the network analysis suggests that the tyrosine kinase EphB4 has high selectivity potential, which is consistent with the selectivity profile of two novel EphB4 inhibitors. CONTACT: email@example.com, firstname.lastname@example.org.