Interactions between Activin-Like Kinase 5 (ALK5) receptor and its inhibitors and the construction of a Docking Descriptor-Based QSAR model

Abstract

Neste estudo, simulações de dinâmica molecular (MD) e docking foram realizadas a fim de investigar o sítio de ligação e as interações de 61 inibidores com o receptor de kinase-5 do tipo ativina (ALK5). Uma simulação MD foi realizada sobre o receptor para obter sua conformação em água. A análise de docking revelou que interações do tipo ligação hidrogênio e hidrofóbicas desempenham papéis importantes no complexo inibidor-ALK5. Estas interações foram confirmadas por cristalografia de raios X. Além disso, para estudar a estabilidade da conformação do receptor, uma segunda simulação MD sobre o complexo foi realizada em um ambiente aquoso. O raio de giro para o complexo mostrou que a conformação de ALK5 não muda na presença do inibidor. Foram calculados 134 descritores obtidos da estrutura molecular e docking; os mais viáveis foram usados em relações quantitativas entre estrutura química e atividade biológica (QSAR). A regressão de vetores-suporte baseada em mínimos quadrados (LS-SVR) apresentou um modelo confiável com Q 2 = 0,837 e R = 0,917. Finalmente, os tipos de interações e propriedades dos descritores foram usados para propor novos inibidores.In this study, molecular docking and molecular dynamics (MD) simulations were conducted to investigate both the binding site and interactions of 61 inhibitors with activin-like kinase-5 (ALK5) receptor. A MD simulation was performed on the receptor to obtain receptor conformation in a water environment. Docking analysis revealed that hydrophobic and hydrogen bonding interactions play important roles in ALK5-inhibitor complex. These interactions were confirmed by X-ray crystallography. Furthermore, to study receptor conformation stability, a second MD simulation on complex was performed in an aqueous environment. Radius of gyration for complex showed that the ALK5 conformation did not change in the presence of the inhibitor. 134 descriptors emerging from docking and molecular structure were calculated and the most feasible ones were used in quantitative structure -activity relationships (QSAR). The LS-SVR (least squares support vector regression) gave reliable model with Q 2 = 0.837 and R = 0.917. Finally, the types of interactions and properties of the descriptors were used to propose new inhibitors.

Investigation of the interaction of anthraquinones of Cassia occidentalis seeds with bovine serum albumin by molecular docking and spectroscopic analysis: Correlation to their in vitro cytotoxic potential

“…spacing of 0.375Å (Chinnadurai et al, 2015;Ebrahimi et al, 2012;Cohen et al, 2011). This was done in such a way so as to enclose the active site residues where the ligands could rotate around flexibly.…”

Performance of smoothly clipped absolute deviation as a variable selection method in the artificial neural network‐based QSAR studies

“…LASSO shrinks the coefficients of the insignificant variables to zero, and the most effective variables are selected according to the cross‐validation prediction error. Recently, LASSO has been considered in QSAR/QSPR studies as a modeling approach 17–20 . There is only one report on the application of LASSO as variable selection before nonlinear modeling by support vector machine (LASSO‐SVM), in which the performance of LASSO has been compared with random forest (RF) as a variable selection method 21 .…”

LM-ANN-based QSAR model for the prediction of pEC50 for a set of potent NNRTI using the mixture of ligand–receptor interaction information and drug-like indexes

“…The results from the established QSAR models help researchers to find a quantitative relationship between structures and biological activities that leads to the design of new compounds with remarkable biological activities with no need for any experimental studies (Muhammad et al 2018;Ojha Lokendra et al 2013). In recent years, the three-dimensional (3D) structures for a wide variety of receptors have become available and also the computational methods have greatly improved; thus, the use of descriptors containing information about the interactions of ligands with the active site of receptors has been highly suggested in the QSAR studies (Amini et al 2016;Chakraborty et al 2014;Chen and Chen 2012;Coi and Bianucci 2013;Davood and Iman 2011;Ebrahimi and Khayamian 2014;Ebrahimi et al 2012;Garg et al 2010;Gharaghani et al 2013;Rasouli and Davood 2018;Safarizadeh and Garkani-Nejad 2019;Sheikhpour et al 2017;Singla et al 2011;Zheng et al 2014). Among the computational chemistry methods, molecular docking is a powerful tool that provides the LR interaction information (Gharaghani et al 2013) through the different computational software such as Dock and AutoDock (Kramer et al 1999;Morris et al 1998).…”

“…The descriptors computed in this way are in fact the same as the structural descriptors; however, because they are calculated from the modified structure of the docked ligand, they have to some extent the interaction information but do not fully reflect the LR interactions. The other way is the extraction of MDDs as the interactions information from the best conformation of LR complex after successful docking of each ligand in the active site of the receptor (Amini et al 2016;Chakraborty et al 2014;Chen and Chen 2012;Coi and Bianucci 2013;Davood and Iman 2011;Ebrahimi and Khayamian 2014;Ebrahimi et al 2012;Garg et al 2010;Gharaghani et al 2013;Rasouli and Davood 2018;Safarizadeh and Garkani-Nejad 2019;Sheikhpour et al 2017;Singla et al 2011;Zheng et al 2014). The MDDs computed in this way are of the LR binding energy and enter the LR interaction information into the QSAR models, successfully.…”

“…The low prediction ability of such QSAR models could be attributed to this fact that although MDDs contain valuable information about the interactions, they do not have the structural features of the ligands. Therefore, MDDs are not a complete representation of the studied ligands and adding some structural descriptors to the MDDs has been suggested to improve the predictability of the QSAR models (Chakraborty et al 2014;Coi and Bianucci 2013;Ebrahimi and Khayamian 2014;Ebrahimi et al 2012;Gharaghani et al 2013;Safarizadeh and Garkani-Nejad 2019;Singla et al 2011). It is expected that the use of a mixture of MDDs and structural descriptors improves the performance of the presented QSAR models, but when MDDs are combined with a large number of structural descriptors, MDDs are not appeared in the final QSAR model (Amini et al 2016;Davood and Iman 2011;Sheikhpour et al 2017).…”