Exploring the computational methods for protein-ligand binding site prediction

Abstract

Proteins participate in various essential processes in vivo via interactions with other molecules. Identifying the residues participating in these interactions not only provides biological insights for protein function studies but also has great significance for drug discoveries. Therefore, predicting protein-ligand binding sites has long been under intense research in the fields of bioinformatics and computer aided drug discovery. In this review, we first introduce the research background of predicting protein-ligand binding sites and then classify the methods into four categories, namely, 3D structure-based, template similarity-based, traditional machine learning-based and deep learning-based methods. We describe representative algorithms in each category and elaborate on machine learning and deep learning-based prediction methods in more detail. Finally, we discuss the trends and challenges of the current research such as molecular dynamics simulation based cryptic binding sites prediction, and highlight prospective directions for the near future.

Keywords: Deep learning; Ligand binding site; Machine learning; Protein; Protein–ligand binding.

Fig. 1

3D schematic of a protein structure and its binding ligands generated from the PDB website. The protein shown above is the crystal structure of human deoxyhaemoglobin at 1.74 Å resolution, published on PDB (Access Code: 4HHB). The amplified ligand is [HEM (PROTOPORPHYRIN IX CONTAINING FE)] 142: C with its bonds (Hydrogen, Halogen, et al).

Fig. 2

A simple schematic of SVM A hyperplane divides the points into two categories.

Fig. 3

A simple model of a convolutional neural network Hidden Layers are used to generate the classification result (multiple convolutional layers and pooling layers can be set in a CNN).

Fig. 4

A simple demonstration of deep belief network DBNs are constructed by combining multiple RBMs. Training of DBNs is performed layer by layer. The hidden layer is first inferred from the data vector, and this hidden layer is used as the input data vector of the next layer.