Laboratory of Computational Systems Biology

 

Here we have searchable datasets with biological systems to be used to develop targeted-scoring functions (Seifert, 2009). We used the SAnDReS program (Xavier et al., 2016) to generate the datasets and the AutoDockTools (Morris et al., 2009) to create PDBQT files. We obtained the crystallographic structures from the Protein Data Bank (PDB) (Berman et al., 2000) and binding affinity data from BindingDB (Chen et al., 2001), Binding MOAD (Benson et al., 2008), and PDBbind (Wang et al., 2004). The major application of the data available here is the development of targeted-scoring functions making use of machine-learning techniques such as the ones available in the programs SAnDReS (Xavier et al., 2016) and Taba (da Silva et al., 2020). We expect to use this structural and binding information to explore the scoring function space (Heck et al., 2017; Bitencourt-Ferreira & de Azevedo, 2019) and design scoring functions targeted to each of the datasets available here. We used these datasets to evaluate the predictive performance of the Taba program (da Silva et al., 2020). Taba is an acronym for Tool to Analyze the Binding Affinity. This program uses a physical mass-spring model to generate machine-learning models to predict binding affinity based on the atomic coordinates of protein-ligand complexes. We applied it to a dataset of the cyclin-dependent kinase (EC 2.7.11.22) with inhibition constant (Ki) data.

 

References

 

Benson ML, Smith RD, Khazanov NA, Dimcheff B, Beaver J, Dresslar P, Nerothin J, Carlson HA. Binding MOAD, a high-quality protein-ligand database. Nucleic Acids Res. 2008; 36(Database issue): D674–D678.

 

Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE. The Protein Data Bank. Nucleic Acids Res., 2000; 28(1): 235–242.

 

Bitencourt-Ferreira G, de Azevedo WF Jr. Exploring the Scoring Function Space. Methods Mol Biol. 2019a; 2053: 275–281.

 

Chen X, Liu M, Gilson MK. BindingDB: a web-accessible molecular recognition database. Comb Chem High Throughput Screen. 2001; 4(8): 719–725.

 

da Silva AD, Bitencourt-Ferreira G, de Azevedo WF Jr. Taba: A Tool to Analyze the Binding Affinity. J Comput Chem. 2020; 41(1): 69–73.

 

Heck GS, Pintro VO, Pereira RR, de Ávila MB, Levin NMB, de Azevedo WF. Supervised Machine Learning Methods Applied to Predict Ligand-Binding Affinity. Curr Med Chem. 2017; 24(23): 2459–2470.

 

Morris GM, Huey R, Lindstrom W, Sanner MF, Belew RK, Goodsell DS, Olson AJ. AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility. 2009; J Comput Chem 30: 2785–2791.

 

Seifert MH. Targeted scoring functions for virtual screening. Drug Discov Today. 2009;14(11-12):562–569.

 

Wang R, Fang X, Lu Y, Wang S. The PDBbind database: collection of binding affinities for protein-ligand complexes with known three-dimensional structures. J Med Chem. 2004; 47(12): 2977–2980.

 

Xavier MM, Heck GS, de Avila MB, Levin NM, Pintro VO, Carvalho NL, Azevedo WF Jr. SAnDReS a Computational Tool for Statistical Analysis of Docking Results and Development of Scoring Functions. Comb. Chem. High Throughput Screen. 2016; 19(10): 801–12.

 

 

Keywords: Protein; PDBQT; receptor-ligand; ligand; interactions; protein-ligand interactions; drug design; drug discovery; protein crystallography; molecular docking; simulations; docking simulations; binding affinity.

 

Please cite the following reference if SAnDReS program was useful: Xavier MM, Heck GS, de Avila MB, Levin NM, Pintro VO, Carvalho NL, Azevedo WF Jr. SAnDReS a Computational Tool for Statistical Analysis of Docking Results and Development of Scoring Functions. Comb Chem High Throughput Screen. 2016; 19(10): 801–812.

SAnDReS draws inspiration from several protein-ligand projects that we have been working on in the last decades. These projects began in the 1990s with pioneering studies focused on intermolecular interactions between cyclin-dependent kinase and inhibitors (de Azevedo et al., 1996; 1997). SAnDReS is a free and open-source (GNU General Public License) computational environment for the development of machine-learning models for prediction of ligand-binding affinity. SAnDReS is also a tool for statistical analysis of docking simulations and evaluation of the predictive performance of computational models developed to calculate binding affinity. We have implemented machine learning techniques to generate regression models based on experimental binding affinity and scoring functions such as PLANTS and MolDock scores. The scikit-learn library has a wide spectrum of supervised machine learning techniques for regression, such as Ordinary Least Squares and Ridge Regression. SAnDReS was developed using Python programming language, and SciPy, NumPy, scikit-learn, and Matplotlib libraries. Data obtained from any protein-ligand docking program can be analyzed by SAnDReS, the only requisite is to have protein structures in Protein Data Bank (PDB) format, ligands in Structure Data File (SDF) format, docking and scoring function data in comma-separated values (CSV) format. This program has been applied to several datasets comprised of crystallographic structures for which there is information for the ligand-binding affinity, in order to generate scoring functions tailored to the biological system of interest (Xavier et al., 2016).

 

Molecular docking simulation of roscovitine against the structure of CDK2.

 

References

 

de Azevedo WF Jr, Mueller-Dieckmann HJ, Schulze-Gahmen U, Worland PJ, Sausville E, Kim SH. Structural basis for specificity and potency of a flavonoid inhibitor of human CDK2, a cell cycle kinase. Proc Natl Acad Sci U S A. 1996; 93(7): 2735–2740.

 

de Azevedo WF, Leclerc S, Meijer L, Havlicek L, Strnad M, Kim SH. Inhibition of cyclin-dependent kinases by purine analogues: crystal structure of human cdk2 complexed with roscovitine. Eur J Biochem. 1997; 243(1-2): 518–526.

 

Xavier MM, Heck GS, de Avila MB, Levin NM, Pintro VO, Carvalho NL, Azevedo WF Jr. SAnDReS a Computational Tool for Statistical Analysis of Docking Results and Development of Scoring Functions. Comb Chem High Throughput Screen. 2016; 19(10): 801–812.

 

 

Keywords: Protein-ligand interactions; drug design; drug discovery; molecular docking; simulations; docking simulations; binding affinity.

 

In our research, we see protein-ligand interaction as a result of the relation between the protein space (Smith, 1970) and the chemical space (Bohacek et al., 1996; Dobson, 2004; Kirkpatrick & Ellis, 2004; Lipinski & Hopkins, 2004; Shoichet, 2004; Stockwell, 2004), and we propose to represent these sets as a unique complex system, where the application of computational methodologies may contribute to generate models to predict protein-ligand binding affinities. Such approaches have the potential to create novel semi-empirical force fields to predict binding affinity with superior predictive power when compared with standard methodologies. SFSXplorer is an acronym for Scoring Function Space eXplorer. This computational tool explores the scoring function space with a hybrid algorithm, where we vary energy terms and adjust their relative weights using machine learning algorithms. We propose to use the abstraction of a mathematical space composed of infinite computational models to predict ligand-binding affinity. We named this space as the scoring function space (Heck et al., 2017; Bitencourt-Ferreira & de Azevedo Jr., 2019). By the use of supervised machine learning techniques is possible to explore this scoring function space and build a computational model targeted to a specific biological system.

 

References

 

Bohacek RS, McMartin C, Guida WC. The art and practice of structure-based drug design: a molecular modeling perspective. Med Res Rev. 1996; 16(1):3–50.

 

Bitencourt-Ferreira G, de Azevedo Jr. WF. Development of a machine-learning model to predict Gibbs free energy of binding for protein-ligand complexes. Biophys Chem. 2018; 240: 63–69.

 

Bitencourt-Ferreira G, de Azevedo WF Jr. Exploring the Scoring Function Space. Methods Mol Biol. 2019b; 2053: 275–281.

 

da Silva AD, Bitencourt-Ferreira G, de Azevedo WF Jr. Taba: A Tool to Analyze the Binding Affinity. J Comput Chem. 2020; 41(1): 69–73.

 

Dobson CM. Chemical space and biology. Nature. 2004; 432(7019):824–828.

 

Heck GS, Pintro VO, Pereira RR, de Ávila MB, Levin NMB, de Azevedo WF. Supervised Machine Learning Methods Applied to Predict Ligand-Binding Affinity. Curr Med Chem. 2017; 24(23): 2459–2470.

 

Kirkpatrick P, Ellis C. Chemical Space. Nature 2004; 432:823. PDF

 

Lipinski C, Hopkins A. Navigating chemical space for biology and medicine. Nature. 2004;432(7019):855–861.

 

Shoichet BK. Virtual screening of chemical libraries. Nature. 2004; 432(7019):862–865.

 

Smith JM. Natural selection and the concept of a protein space. Nature. 1970; 225(5232): 563–564.

 

Stockwell BR. Exploring biology with small organic molecules. Nature. 2004; 432(7019):846–854.

 

Xavier MM, Heck GS, de Avila MB, Levin NM, Pintro VO, Carvalho NL, Azevedo WF Jr. SAnDReS a Computational Tool for Statistical Analysis of Docking Results and Development of Scoring Functions. Comb Chem High Throughput Screen. 2016; 19(10): 801–812.

 

 

Keywords: Computational systems biology; systems biology; systems approach; machine learning; protein; ligand; interactions; protein-ligand interactions; drug design; drug discovery; binding affinity.

 

Please cite the following reference if the Taba program was useful:

The basic idea behind the Taba is that the determinant structural features responsible for ligand-binding affinity are already somehow imprinted in the three-dimensional structures of protein-ligand complexes. When we consider an ensemble of crystallographic structures, for which ligand-binding information data is available, we have the raw data that can be used by the program Taba to generate a target-based polynomial scoring function. To build this target-based polynomial scoring function (flowchart below), Taba reads all structures available for a biological system of interest and calculates the average distances for each type of pair of atoms. For instance, consider intermolecular Carbon-Carbon distances, where one Carbon belongs to the protein and the second one is in the ligand. Taba calculates the average intermolecular distance for Carbon-Carbon pair. Taba considers this length as the equilibrium distance for a Carbon-Carbon pair, taking an analogy with a mass-spring system. For a given structure, displacement from this equilibrium distance generates an increase in the energy of the system. Again, we consider this naïve analogy with the mass-spring system.

 

Flowchart showing the main steps used to generate targeted-scoring functions with Taba.

 

Reference

 

da Silva AD, Bitencourt-Ferreira G, de Azevedo WF Jr. Taba: A Tool to Analyze the Binding Affinity. J Comput Chem. 2020; 41(1): 69–73.

 

Keywords: Taba; tool to analyze the binding affinity; Protein-ligand interactions; drug design; drug discovery; molecular docking; simulations; docking simulations; binding affinity.