Vinardo: A Scoring Function Based on Autodock Vina Improves Scoring, Docking, and Virtual Screening.
Quiroga R, Villarreal MA. · 2016
Read Vinardo: A Scoring Function Based on Autodock Vina Improves Scoring, Docking, and Virtual Screening. by Quiroga R, Villarreal MA. (2016) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4865195.txt.
Introduction
Rodrigo Quiroga and Marcos A. Villarreal presented Vinardo in 2016, a new scoring function for the popular docking program AutoDock Vina. Docking software predicts how a drug-like molecule fits a protein pocket, and a scoring function rates the fit. Standard training tunes scores to match measured binding strength, which the pair found does not give the best docking.
Vinardo was tuned instead with scoring, shape settling, and re-docking rounds on carefully cleaned datasets, and its atomic size values were optimized. Across all tests and datasets, Vinardo beat Vina at scoring, docking, and screening many compounds. The free tool gives drug hunters a sharper first filter before costly lab work.
Catalog record
| CREATOR | Quiroga R, Villarreal MA. |
|---|---|
| YEAR | 2016 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Vinardo: A Scoring Function Based on Autodock Vina Improves Scoring, Docking, and Virtual Screening By Rodrigo Quiroga, Marcos A. Villarreal, Heinrich Sticht Published in: PLoS ONE DOI: 10.1371/journal.pone.0155183 ABSTRACT Autodock Vina is a very popular, and highly cited, open source docking program. Here we present a scoring function which we call Vinardo (Vina RaDii Optimized). Vinardo is based on Vina, and was trained through a novel approach, on state of the art datasets. We show that the traditional approach to train empirical scoring functions, using linear regression to optimize the correlation of predicted and experimental binding affinities, does not result in a function with optimal docking capabilities. On the other hand, a combination of scoring, minimization, and re-docking on carefully curated training datasets allowed us to develop a simplified scoring function with optimum docking performance. This article provides an overview of the development of the Vinardo scoring function, highlights its differences with Vina, and compares the performance of the two scoring functions in scoring, docking and virtual screening applications. Vinardo outperforms Vina in all tests performed, for all datasets analyzed. The Vinardo scoring function is available as an option within Smina, a fork of Vina, which is freely available under the GNU Public License v2.0 from http://smina.sf.net. Precompiled binaries, source code, documentation and a tutorial for using Smina to run the Vinardo scoring function are available at the same address. 1 INTRODUCTION Protein-ligand docking is a computational method which attempts to predict the most probable position, orientation and conformation with which a ligand (often a small organic molecule) can bind to a protein. The binding free energy of a ligand to a protein can be predicted in different ways, and hence docking programs can be classified into one of the following three categories. 1- Force-field based 2- Empirical scoring functions 3- Knowledge-based potentials [1]. Different programs, using all three strategies, have been successfully used in many different drug discovery projects [2]. Autodock Vina [3] (referred to as Vina from here on) is the successor to Autodock 4, a highly successful docking program [4,5]. However, Vina is a different program and uses a different scoring function and global optimization algorithm. It is two orders of magnitude faster [3,6], and has shown similar or improved accuracy [3,6]. Vina is open source and is currently used by many groups worldwide for docking and virtual screening. The original paper describing Vina was published in 2010 and has now over 3000 citations according to Google Scholar. For the estimation of ligand-receptor affinity, Vina uses an empirical scoring function which is inspired by the X-score function [7]. As stated by the authors, the nature of the scoring function is “more of a machine learning than a physics-based function” [3]. The aim of the present work was to develop a simpler scoring function based on Vina with fewer parameters and with a more physics-based character, that is, a scoring function composed of terms which are readily identified as some of the traditional terms used in force fields [2]. The Vina scoring function is implemented not only in the Vina program but also in other closely related programs as iDock and Smina [8,9]. The Smina program is a fork of Vina recently published by Koes et al. [9] which maintains most of Vina ´s functionality, and adds a myriad of new features, many of them associated with energy minimization and the possibility to easily develop new scoring functions, which makes it a very interesting tool. Smina also provides a very easy, hassle-free way of defining the search space, also called bounding box, which in Vina was not straightforward. Koes et al. [9] also developed a novel scoring function (referred to as Dk_scoring from here on), selecting energetic terms using the forward selection algorithm, and then assigning weights to each term using linear regression between experimental and predicted binding energies of the CSAR 2012 dataset [10]. The Dk_scoring function shows improved correlation between the calculated and experimental binding affinity for the training set used, as compared to Vina. On the other hand, it is apparently less efficient than Vina when used for docking, and also at predicting poses which closely resemble the crystallized protein-ligand complexes used for training [9]. We have used Smina as a tool to develop Vinardo (Vina RaDii Optimized), a scoring function which shares component terms with the Vina scoring function: steric interactions, hydrophobic interactions, and non-directional H-bonds. Despite sharing component terms, Vinardo displays several differences with Vina; a modified steric interaction term, new atomic radii, and simplified interactions (using a lower number of parameters). Vinardo is implemented as an optional scoring function in Smina. To compare the docking abilities of Vinardo and Vina, we performed re-docking assays on four high quality datasets. To measure the scoring and ranking abilities of Vinardo, we repeated the scoring function analyses performed in CASF 2013 [11]. Finally, we tested virtual screening capabilities by docking a multitude of active and inactive compounds against different proteins available in the DUD database, and verifying Vinardo's capability to rank active compounds above inactive ones. As we will show in the following sections, Vinardo displays consistently superior docking, scoring, ranking and virtual screening capabilities for all datasets analyzed, compared to Vina and Dk_scoring. 2 MATERIALS AND METHODS 2.1 Scoring Functions 2.1.1 Vina An empirical scoring function calculates the affinity, or fitness, of protein-ligand binding by summing up the contributions of a number of individual terms [1]. Each of these terms generally represent an important energetic factor in protein-ligand binding. There are several parameters involved in each of these functions which can be modified to improve the predictions. Lastly, each term is weighted (multiplied by a constant) before being summed up into the final predicted binding affinity. In the following paragraphs we give a brief description of the Vina scoring function in order to compare it to Vinardo. For a more detailed description of the Vina scoring function, the reader is referred to the original paper by Trott and Olson [3]. The binding energy is predicted as the sum of distance-dependent atom pair interactions (Eq 1). [formula: E=∑epair(d)Eq 1] Here d is the surface distance calculated with Eq 2, where r is the interatomic distance and Ri and Rj are the radii of the atoms in the pair. [formula: d=r−Ri−RjEq 2] Every atom pair interacts through a steric interaction given by the first three terms of Eq 3. Also, depending on the atom type, there could be hydrophobic and non-directional H-bonding interactions, given by the last two terms of Eq 3. [formula: epair(d)={w1*Gauss1(d)+w2*Gauss2(d)+w3*Repulsion(d)+w4*Hydrophobic(d)+w5*HBond(d)Eq 3] Steric interaction in Vina is assessed using three terms (Eqs 4 to 6). The combination of an attractive Gaussian function with a repulsive parabolic function reproduces the general shape of a typical Lennard-Jones interaction, provided the Gaussian term is negative and the parabolic positive. A graphical representation of these functions is shown in S1 Fig (also see Figure 1 in [3]). If both atoms in the pair are hydrophobic the linear function in Eq 7 is included. Also, if the pair consists of an H-bond donor and an H-bond acceptor, Eq 8 is added. These last two equations are simple piecewise linear and their effect can be thought of as modifying