Our initial approach to identify hits was two pronged

Our initial approach to identify hits was two-pronged and included a high-throughput screening (HTS) campaign of the OSI H 89 library as well as a virtual screening (VS) campaign, utilizing a publically available crystal structure of ACK1 (PDB code: ). A tolerance for protein flexibility was incorporated into the virtual screening design with the hope of identifying hits which would require diverse protein conformations for binding, since such potential hits might be rejected if a rigid protein structure was used. In order to confirm the potency of prioritized virtual screening hits, an Alpha-screen assay (ATP concentration of 100μM) was implemented where the low molecular weight (MW=356) imidazo[1,5-]pyrazine-derived compound () emerged as a potent hit (IC=0.19μM). Detailed computational studies were then performed on compound , with some of the major modeled interactions between this compound and ACK1 highlighted in . The C8-amino group and N7 of the imidazo[1,5-]pyrazine core of the compound form two hydrogen bonds with the hinge backbone amides. The 4-phenoxyphenyl moiety, termed the northern domain (NoD), substituted from the C1 position of the core, passes through a channel occupied by gatekeeper Thr205, with the ring A moiety further extending into a lipophilic pocket primarily formed by Val178 and Met181 of the αC-helix as well as Leu192, Met203, Thr205 and Phe271 (F of the DFG loop). Furthermore, modeling studies suggested a protein conformational change associated with this proposed binding mode, since the accommodation of ring A of required an enlargement of the back pocket from its original size which was not evident from the initial static view of the crystal structure (PDB code: ). The cyclobutyl moiety, termed the southern domain (SoD), emanating from C3 of the core, extends past the glycine rich P-loop and towards the solvent exposed region of the protein. To further explore and build SAR around compound , we first sought to analog around the NoD moiety and specifically ring A while keeping the rest of the molecule constant (). Various substitutions encompassing different sizes and electronic properties were walked around the ring. While C2-F (compound ) and C3-F (compound ) substitutions offered a slight increase in potency, all other substitutions resulted in a loss of potency. In the C4 position of ring A, substitution consistently resulted in a loss of potency. This observation was in agreement with modeling predictions which suggested that phenyl as ring A optimally fit into this hydrophobic pocket as there is not enough additional space available to accommodate further substituents. Pyridyl variant was also evaluated and found to be inferior to the phenyl group in terms of ACK1 potency, most likely due to an unfavorable desolvation penalty introduced by the pyridine nitrogen atom. Replacement of ring A with alkyl groups such as methyl, isopropyl and cyclohexyl groups abolished potency (data not shown), suggesting phenyl is an optimal pharmacophoric element at this position. We next sought to explore the SAR around the proximal phenyl ring (ring B) of the NoD moiety. While using phenyl as a template for ring B, modeling studies suggested that a hydrogen bond acceptor substituted off the C2′ position could form a hydrogen bond interaction with the side chain hydroxyl group of the gatekeeper Thr205 and thus may improve potency. As shown in , while both C2′-OMe (compound , a) and C2′-F (compound ) substitutions increased potency over , C2′-Cl (compound ) and C2′-Me (compound ) substitutions decreased potency. These results, when taken together, helped to further validate the model. Other hydrogen bond acceptor moieties evaluated at this position such as amides (compounds –) generally abolished potency as this was most likely caused by steric clashes with the protein due to limited binding space in this region. Interestingly, pyridone analog was found to be inactive although there were no steric issues in binding as suggested by modeling, and the pyridone carbonyl is a strong hydrogen bond acceptor based on its p value. To further clarify this apparent discrepancy, we performed quantum chemistry calculations on binding interactions of compounds , and . As shown in , compound indeed formed a stronger hydrogen bond interaction with Thr205 compared to as indicated by a more favorable interaction energy gain (ΔΔ) but suffered a much greater energy loss due to an increased desolvation penalty (ΔΔ) and adverse conformational changes (ΔΔ). As a result, compound had a overall energetically unfavorable interaction (ΔΔ) which rendered it inactive.