Learning from failures: anatomy and post-mortem analysis of an unsuccessful search for inhibitors of SARS-CoV2 papain-like protease

Details of failed computer-aided molecular design campaigns are lacking in the published literature due to the prevalent preference for the publication of positive results and successful research narratives. The lessons learned from failed endeavors therefore remain circumscribed to the gray literature and to the institutional memories of research groups, which may prevent the identification of common pitfalls or the sharing of lessons learned from these failures. In this report we describe a promising inhibitor- design campaign which unexpectedly failed in the experimental test phase, and the likely source for that failure. The target protein for this research was the papain-like protease (PLPro) from SARS-CoV2, which cleaves the host's interferon-stimulated gene 15 protein and leads to the weakening of type 1 interferon response. In this work, we have performed a QSAR-based search of the PubChem database to find molecules with higher affinity to PLPro and, consequently, better inhibitory effect than previously-confirmed inhibitor GRL-0167. The 50 detected candidates were docked to the structure of PLPro, and the complexes of the three molecules bound to the experimentally-confirmed S3-S4 pocket were subjected to triplicate molecular dynamics simulations to ascertain the stability of the binding mode. The simulations showed that one of the candidates remained stably bound to the inhibitory pocket. Analysis of binding energies showed that this molecule was at least as good a binder as previously-confirmed inhibitors. Despite such promising results, these candidates failed experimental validation, and a simple, newly-developed, computational protocol identified self-aggregation as the likely reason for this lack of success. Widespread adoption of this simple protocol will highlight ligands prone to this unhelpful behavior, and hopefully increase the success rate of future molecular design campaigns.