1We have upgraded our Computational Analysis of Novel Drug Opportunities (CANDO) 2 platform for shotgun drug repurposing to include ligand-based, data fusion, and decision tree 3 pipelines. The first version of CANDO implemented a structure-based pipeline that mod-4 eled interactions between compounds and proteins on a large scale, generating compound-5 proteome interaction signatures used to infer similarity of drug behavior; the new pipelines 6 accomplish this by incorporating molecular fingerprints and the Tanimoto coefficient. We 7 obtain improved benchmarking performance with the new pipelines across all three evalua-8 tion metrics used: average indication accuracy, pairwise accuracy, and coverage. The best 9 performing pipeline achieves an average indication accuracy of 19.0% at the top10 cutoff, 10 compared to 11.7% for v1, and 2.2% for a random control. Our results demonstrate that the 11 CANDO drug recovery accuracy is substantially improved by integrating multiple pipelines, 12 thereby enhancing our ability to generate putative therapeutic repurposing candidates, and 13 increasing drug discovery efficiency. 14 Introduction 15 Drug repurposing 16 Bringing a new drug to the market may costs hundreds of millions of dollars and takes years 17 of work. 1 Drug repurposing is the process of discovering a new use for an existing drug. 2,3 18This process may take advantage of existing data on safety and pharmacokinetic properties 19 from previous trials and clinical use to reduce costs and time associated with traditional drug 20 discovery. Classic examples of drug repurposing include sildenafil (Viagra) and thalidomide, 21 which initially were developed to treat chest pain and morning sickness, but were repurposed 22 to treat erectile dysfunction and erythema nodosum leprosum respectively. 2,4,5 Drugs which 23 have already been repurposed once are being researched for even more novel uses. For exam-24 ple, raloxifene was originally indicated for prevention of osteoporosis and was subsequently 25 approved for risk reduction in the development of breast cancer. 6 More recently, raloxifene 26 has been suggested as a possible treatment for Ebola virus disease 7-9 '. These examples of 27 putative and/or successful drug repurposing underlies the diverse mechanisms through which 28 a single compound may treat a variety of disease types. 10,11 High throughput, target-based, 29 and phenotypic screening of compounds can be used to generate putative candidates for re-30 purposing. 12 For example, potential treatments for Zika virus infection were identified using 31 a phenotypic screen. 13 32 Computational drug discovery and repurposing 33 Finding new drugs or new uses for existing drugs computationally takes advantage of the 34 growing amount of data generated from wet lab experiments accessible on the Internet, 35 increased computational power, and higher fidelity of computational models to reality. Ap-36 proaches to computational drug discovery and repurposing have been classified as structure-37 b...