Sabri Hamad1, Gianluca Adornetto1, J Jesús Naveja1,2,3, Aakash Chavan Ravindranath1, Johannes Raffler1, Mónica Campillos1,4. 1. Institute of Bioinformatics and Systems Biology, Helmholtz Zentrum München, Neuherberg, Germany. 2. Faculty of Medicine, PECEM, UNAM, Mexico City, Mexico. 3. Faculty of Chemistry, DIFACQUIM, UNAM, Mexico City, Mexico. 4. German Center for Diabetes Research, Neuherberg, Germany.
Abstract
MOTIVATION: The identification of protein targets of novel compounds is essential to understand compounds' mechanisms of action leading to biological effects. Experimental methods to determine these protein targets are usually slow, costly and time consuming. Computational tools have recently emerged as cheaper and faster alternatives that allow the prediction of targets for a large number of compounds. RESULTS: Here, we present HitPickV2, a novel ligand-based approach for the prediction of human druggable protein targets of multiple compounds. For each query compound, HitPickV2 predicts up to 10 targets out of 2739 human druggable proteins. To that aim, HitPickV2 identifies the closest, structurally similar compounds in a restricted space within a vast chemical-protein interaction area, until 10 distinct protein targets are found. Then, HitPickV2 scores these 10 targets based on three parameters of the targets in such space: the Tanimoto coefficient (Tc) between the query and the most similar compound interacting with the target, a target rank that considers Tc and Laplacian-modified naïve Bayesian target models scores and a novel parameter introduced in HitPickV2, the number of compounds interacting with each target (occur). We present the performance results of HitPickV2 in cross-validation as well as in an external dataset. AVAILABILITY AND IMPLEMENTATION: HitPickV2 is available in www.hitpickv2.com. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
MOTIVATION: The identification of protein targets of novel compounds is essential to understand compounds' mechanisms of action leading to biological effects. Experimental methods to determine these protein targets are usually slow, costly and time consuming. Computational tools have recently emerged as cheaper and faster alternatives that allow the prediction of targets for a large number of compounds. RESULTS: Here, we present HitPickV2, a novel ligand-based approach for the prediction of human druggable protein targets of multiple compounds. For each query compound, HitPickV2 predicts up to 10 targets out of 2739 human druggable proteins. To that aim, HitPickV2 identifies the closest, structurally similar compounds in a restricted space within a vast chemical-protein interaction area, until 10 distinct protein targets are found. Then, HitPickV2 scores these 10 targets based on three parameters of the targets in such space: the Tanimoto coefficient (Tc) between the query and the most similar compound interacting with the target, a target rank that considers Tc and Laplacian-modified naïve Bayesian target models scores and a novel parameter introduced in HitPickV2, the number of compounds interacting with each target (occur). We present the performance results of HitPickV2 in cross-validation as well as in an external dataset. AVAILABILITY AND IMPLEMENTATION: HitPickV2 is available in www.hitpickv2.com. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
Authors: Kerstin Hiesinger; Jan S Kramer; Janosch Achenbach; Daniel Moser; Julia Weber; Sandra K Wittmann; Christophe Morisseau; Carlo Angioni; Gerd Geisslinger; Astrid S Kahnt; Astrid Kaiser; Anna Proschak; Dieter Steinhilber; Denys Pogoryelov; Karen Wagner; Bruce D Hammock; Ewgenij Proschak Journal: ACS Med Chem Lett Date: 2019-05-29 Impact factor: 4.345