Forging the Basis for Developing Protein–Ligand Interaction Scoring Functions

Abstract

In structure-based drug design, scoring functions are widely used for fast evaluation of protein-ligand interactions. They are often applied in combination with molecular docking and de novo design methods. Since the early 1990s, a whole spectrum of protein-ligand interaction scoring functions have been developed. Regardless of their technical difference, scoring functions all need data sets combining protein-ligand complex structures and binding affinity data for parametrization and validation. However, data sets of this kind used to be rather limited in terms of size and quality. On the other hand, standard metrics for evaluating scoring function used to be ambiguous. Scoring functions are often tested in molecular docking or even virtual screening trials, which do not directly reflect the genuine quality of scoring functions. Collectively, these underlying obstacles have impeded the invention of more advanced scoring functions. In this Account, we describe our long-lasting efforts to overcome these obstacles, which involve two related projects. On the first project, we have created the PDBbind database. It is the first database that systematically annotates the protein-ligand complexes in the Protein Data Bank (PDB) with experimental binding data. This database has been updated annually since its first public release in 2004. The latest release (version 2016) provides binding data for 16 179 biomolecular complexes in PDB. Data sets provided by PDBbind have been applied to many computational and statistical studies on protein-ligand interaction and various subjects. In particular, it has become a major data resource for scoring function development. On the second project, we have established the Comparative Assessment of Scoring Functions (CASF) benchmark for scoring function evaluation. Our key idea is to decouple the "scoring" process from the "sampling" process, so scoring functions can be tested in a relatively pure context to reflect their quality. In our latest work on this track, i.e. CASF-2013, the performance of a scoring function was quantified in four aspects, including "scoring power", "ranking power", "docking power", and "screening power". All four performance tests were conducted on a test set containing 195 high-quality protein-ligand complexes selected from PDBbind. A panel of 20 standard scoring functions were tested as demonstration. Importantly, CASF is designed to be an open-access benchmark, with which scoring functions developed by different researchers can be compared on the same grounds. Indeed, it has become a popular choice for scoring function validation in recent years. Despite the considerable progress that has been made so far, the performance of today's scoring functions still does not meet people's expectations in many aspects. There is a constant demand for more advanced scoring functions. Our efforts have helped to overcome some obstacles underlying scoring function development so that the researchers in this field can move forward faster. We will continue to improve the PDBbi...

Deep neural network affinity model for BACE inhibitors in D3R Grand Challenge 4

“…To build a BACE-specific affinity machine learning model, a dataset comprised of 222 published BACE ligand affinities (label value ytraining, dimension: 222×1) was extracted from PDBbind v.2017 which contains 14,761 protein-ligand complexes affinity data, see Table S2. [17] In this work, structure-based and/or ligand-based features were used as input features (X) for our machine learning models. To generate the structure-based features used for machine learning model training, ten AutoDock Vina-like scoring terms were generated for the 222 BACE-ligands using smina, ( Table 1).…”

“…We obtained a dataset (Xtraining) of 222 BACE-ligands deposited in PDB with their affinities (ytraining) extracted from PBDBind v2017 [17]. We investigated three aspects of applying machine learning in BACE-ligand affinity prediction: input feature selection, machine learning model selection, and machine learning model regularization.…”

“…Machine learning models have shown great potential in advancing current computer-aided drug designing methodology. [13] The advance of machine learning platforms and tools for chemistry, such as TensorFlow [14] and scikit-learn [15], and the public availability of high-quality proteindrug datasets, such as the PDB [16] and PDBbind [17], greatly enables the application of machine learning to drug design. Novel machine learning-based scoring methods, such as RF-Score [18] and KDEEP [19], or machine learning-optimized software, such as Vinardo [20], smina [21], RF-Score-v3 [22], have demonstrated superior performance in addition to their generalizability and accessibility.…”

“…RF (number of estimators: 200, maximum depth: 8) yieldedan R 2 of 0.61(19) for the cross-validation set. SVM regressor (kernel: Gaussian, regularization: C = 2.5) had a R 2 = 0.64(17) for the validation set. We compared DNN/MLP architecture with 10 × time.…”

Druggability Assessment in TRAPP using Machine Learning Approaches

“…A larger dataset, the Druggability augmented from PDBbind (DaPB) dataset, was constructed by filtering the 2017 release of the PDBbind refined set. 25 The PDBbind database collects the experimentally measured binding affinity data for the biomolecular complexes in the Protein Data Bank (PDB), and the PDBbind refined set is a subset containing 4154 protein-ligand complexes with better quality with regard to binding data, crystal structures and the nature of the complexes.…”