Abstract 4850: Combination of effector mimicking with in silico fragment screens is an effective approach for the development of Ras inhibitors

Targeting Ras proteins directly has proven to be exceptionally challenging because they lack deep pockets on the surface that can provide for effective interactions with small molecules. Several shallow pockets have been identified on Ras surface. However, their positioning does not allow for simultaneous targeting with a single small molecule. To generate high affinity Ras binders, we have combined fragment-based drug design and in silico screening of fragment libraries with mimicking of Ras effectors. We have chosen Ras effector NORE1A for mimicking because it has one of the most compact continuous Ras-binding interfaces. Peptide mimetic of the central part of the interface was chemically rigidified and optimized to expand into additional pockets adjacent to Ras's NORE1A binding interface. These included one previously described pocket (Sun et.al 2012, Maurer et al, 2012) along with two newly defined ones. In particular, we have identified a “new” pocket between helix 2 and strands 2 and 3 of Ras protein. The pocket is not highly populated and is present in only 9 out of 71 analyzed Ras structures. The apparent binding region appears disordered in seven published structures suggesting that the fold of the pocket may be environmentally regulated and is likely affected by crystallization conditions. In silico screening of specially constructed virtual libraries consisting of millions of compounds allowed for identification of several potential binders. Two fragment-like molecules have been tested for Ras binding to verify the predictions. Binding assay that utilized microscale thermophoresis revealed that: 1) both compounds bind to WT, G12D and G12V mutants of K-Ras; 2) one of the compounds has higher affinity towards GTP-bound than to GDP -loaded wild type and G12V mutant of K-Ras (KD ≈0.9 mM and 3.2 mM respectively), but binds with approximately same KD≈ 3.3 mM to GDP and GTP-bound G12D K-Ras. We also utilized for screening a novel virtual library of synthetically accessible molecules that can be prepared in one step using well established chemistries and easily obtainable reagents. This resulted in additional potential ligands for the pocket. Two compounds were prepared and proven to bind with affinities in low millimolar range. The use of the ligands identified from virtual screens allowed for the design of a NORE1A mimetic with MW< 1000. The data along with previously published studies show that in silico screens for Ras binders are challenging becaus