Expanding the Accessible Chemical Space of SIRT2 Inhibitors through Exploration of Binding Pocket Dynamics

Abstract

Considerations of binding pocket dynamics are one of the crucial aspects of the rational design of binders. Identification of alternative conformational states or cryptic subpockets could lead to the discovery of completely novel groups of the ligands. However, experimental characterization of pocket dynamics, besides being expensive, may not be able to elucidate all of the conformational states relevant for drug discovery projects. In this study, we propose the protocol for computational simulations of sirtuin 2 (SIRT2) binding pocket dynamics and its integration into the structure-based virtual screening (SBVS) pipeline. Initially, unbiased molecular dynamics simulations of SIRT2:inhibitor complexes were performed using optimized force field parameters of SIRT2 inhibitors. Time-lagged independent component analysis (tICA) was used to design pocket-related collective variables (CVs) for enhanced sampling of SIRT2 pocket dynamics. Metadynamics simulations in the tICA eigenvector space revealed alternative conformational states of the SIRT2 binding pocket and the existence of a cryptic subpocket. Newly identified SIRT2 conformational states outperformed experimentally resolved states in retrospective SBVS validation. After performing prospective SBVS, compounds from the under-represented portions of the SIRT2 inhibitor chemical space were selected for in vitro evaluation. Two compounds, NDJ18 and NDJ85, were identified as potent and selective SIRT2 inhibitors, which validated the in silico protocol and opened up the possibility for generalization and broadening of its application. The anticancer effects of the most potent compound NDJ18 were examined on the triple-negative breast cancer cell line. Results indicated that NDJ18 represents a promising structure suitable for further evaluation.

SIRT2i_Predictor: A Machine Learning-Based Tool to Facilitate the Discovery of Novel SIRT2 Inhibitors

“…Early recognition is the direct reflection of the models’ ability to rank active molecules very early in an ordered list. Herein, we used ROC EF 0.5%, 1%, 2%, and 5%, which quantified the area covered by the curve at 0.5%, 1%, 2%, and 5% of the screened false positives, respectively [ 20 , 46 ]. With the dataset containing a significantly larger number of chemically diverse inactive compounds, the RF:ECFP4 binary model stood out as the model with the greatest predictive power ( Table 4 and Figure 5 ).…”

“…Recently described pharmacological advantages of selective SIRT2 inhibition over non-selective inhibition of other isoforms of the sirtuin family, particularly sirtuin 1 (SIRT1) and sirtuin 3 (SIRT3), positioned selectivity as one of the most important objectives in development of novel SIRT2 inhibitors [ 19 ]. Furthermore, a recent study indicated that the complex conformational behavior of SIRT2 in interaction with inhibitors represents one of the major obstacles in the discovery of novel inhibitors through structure-based computer-aided drug-design (CADD) approaches [ 20 ]. However, years of searching for novel SIRT2 inhibitors resulted in large and diverse datasets that could greatly benefit ligand-based CADD approaches relying on specific machine-learning techniques.…”

SIRT2i_Predictor: A machine learning-based tool to facilitate the discovery of novel SIRT2 inhibitors

“…Herein, we used ROC EF 0.5%, 1%, 2% and 5% which quantify the area covered by the curve at the 0.5%, 1%, 2% and 5% of the screened false positives. 44,52 With dataset containing significantly larger amount of chemically diverse inactive compounds, RF:ECFP4 binary model stood out as the model with the greatest predictive power (Table 4 and Figure 4). In the heavily disbalanced decoys set, RF:ECFP4 binary model displayed better sensitivity, specificity, precision and robustness but also better early recognition.…”

“…The novel structure-based virtual screening (SBVS) approach relying on alternative conformational states of SIRT2 discovered through computationally intensive simulations of the binding pocket dynamics was recently published by our group. 44 Utilization of alternative binding pocket conformational states, besides showing significant improvements in validation metrics compared with single structure approach, resulted in expansion of chemical space coverage of virtual hits. In order to test SIRT2i_Predictor's ability to expand the chemical space of virtual hits, we have repeated prospective SBVS campaign from the mentioned paper encompassing around 200000 compounds from the SPECS database.…”

“…41 For interpretation of atomic contributions and graphical interpretation of structure-activity relationships, similarity maps were build according to the RDKit implementation of the approach proposed by Riniker et al 42 Chemical space projections of virtual hits from virtual screening of SPECS database 43 were calculated using self-organizing maps as described in previous work. 44 In vitro enzymatic evaluation of potency for in-house library of compounds was performed using the fluorometric assay as described elsewhere. 45 Percent of inhibition for all compounds were evaluated in triplicates at concentration of 200 µM.…”