Authors

B Dankwa²; J Forde-Thomas²; K Lees²; A Coghlan³; M Berriman¹; K Hoffmann²;  ¹ University of Glasgow , UK;  ² Aberystwyth University, UK;  ³ Wellcome Sanger Institute, UK

Discussion

Abstract Schistosomiasis, a major neglected tropical disease, is responsible for thousands of human deaths annually. Currently, praziquantel is the only approved chemotherapy for treatment. However, concerns over praziquantel resistance, the need for repeat dosing and the lack of efficacy against juvenile worms threaten the long-term sustainable control of the disease. In the absence of an effective vaccine, there is the pressing need to develop new drugs to complement or replace the existing therapy. However, ab initio, whole-organism screening of druggable gene sets for early anti-schistosomal leads using highthroughput ex vivo phenotypic assays is expensive, laborious, and time-consuming. To fill this gap, we have developed an advanced bioinformatics/cheminformatics drug repurposing pipeline that has identified schistosome drug targets and hit to lead compounds based on associated human protein targets in the ChEMBL database. The approach combined top BLAST hits (E-value <= 1e-10) between Schistosoma mansoni protein coding genes (9896; version 10; PRJEA36577) and ChEMBL single protein targets, together with datasets comprised of bulk transcriptomes (RNA-seq), functional phenotypes (RNAi or gene knockout), compara families, multi-species homologues, protein structural information, sequence alignment conservation, toxicity targets, phylogenetics and chokepoint enzymatic predictions to rank 1921 probable targets (out of 9896). Focusing on this subset of probable targets, compounds linked to their counterpart ChEMBL proteins were extracted and screened through filters such as Lipinski rule of five, toxicity, safety warnings, pan-assay interference, binding affinities and ADMET properties to remove potentially risky compounds. A favourable library of 4916 purchasable compounds consisting of 11% phase IV drugs, 4% phase III compounds and 85% hit-tolead medicinal chemistry compounds was retained. Within-library similarity clustering (≥ 90%) using SkelSpheres descriptors as well as clustering with 2828 anthelminthic compounds curated from the literature identified closely related compounds that limited library diversity. Removing these and previously screened in-house compounds left us with 2077 diverse compounds against the 1921 S. mansoni targets awaiting experimental validation in our diverse ex vivo, whole organism assays.