In silico data mining of large-scale databases for the virtual screening of human interleukin-2 inhibitors

Abstract

Interleukin-2 (IL-2) is involved in the activation and differentiation of T-helper cells. Uncontrolled activated T cells play a key role in the pathophysiology by stimulating inflammation and autoimmune diseases like arthritis, psoriasis and Crohn's disease. T cells activation can be suppressed either by preventing IL-2 production or blocking the IL-2 interaction with its receptor. Hence, IL-2 is now emerging as a target for novel therapeutic approaches in several autoimmune disorders. This study was carried out to set up an effective virtual screening (VS) pipeline for IL-2. Four docking/scoring approaches (FRED, MOE, GOLD and Surflex-Dock) were compared in the re-docking process to test their performance in producing correct binding modes of IL-2 inhibitors. Surflex-Dock and FRED were the best in predicting the native pose in its top-ranking position. Shapegauss and CGO scoring functions identified the known inhibitors of IL-2 in top 1, 5 and 10 % of library and differentiated binders from non-binders efficiently with average AUC of > 0.9 and > 0.7, resp. The applied docking protocol served as a basis for the VS of a large database that will lead to the identification of more active compounds against IL-2.