A comparative study of family-specific protein–ligand complex affinity prediction based on random forest approach

Abstract

The assessment of binding affinity between ligands and the target proteins plays an essential role in drug discovery and design process. As an alternative to widely used scoring approaches, machine learning methods have also been proposed for fast prediction of the binding affinity with promising results, but most of them were developed as all-purpose models despite of the specific functions of different protein families, since proteins from different function families always have different structures and physicochemical features. In this study, we proposed a random forest method to predict the protein-ligand binding affinity based on a comprehensive feature set covering protein sequence, binding pocket, ligand structure and intermolecular interaction. Feature processing and compression was respectively implemented for different protein family datasets, which indicates that different features contribute to different models, so individual representation for each protein family is necessary. Three family-specific models were constructed for three important protein target families of HIV-1 protease, trypsin and carbonic anhydrase respectively. As a comparison, two generic models including diverse protein families were also built. The evaluation results show that models on family-specific datasets have the superior performance to those on the generic datasets and the Pearson and Spearman correlation coefficients (R p and Rs) on the test sets are 0.740, 0.874, 0.735 and 0.697, 0.853, 0.723 for HIV-1 protease, trypsin and carbonic anhydrase respectively. Comparisons with the other methods further demonstrate that individual representation and model construction for each protein family is a more reasonable way in predicting the affinity of one particular protein family.

A Review on Applications of Computational Methods in Drug Screening and Design

“…Compared to DTs, it is impossible that RF over-fits the data, and the RF has been used for bioactivity data classification [81], toxicity modeling [82], and drug target prediction [83], etc. Wang et al [84] used the RF approach to model the binding affinity of protein-ligand on 170 HIV-1 proteases complexes, 110 trypsin complexes, and 126 carbonic anhydrase complexes, which demonstrated that individual representation and model construction for each protein family is a more reasonable way in predicting the affinity of one particular protein family.…”

Machine‐learning scoring functions to improve structure‐based binding affinity prediction and virtual screening

“…As previously argued in the context of diverse test sets, Wang et al’s SF did not perform better than RF‐Score in every target class despite using a far more precise description of the complex. Actually, the new SF only outperformed RF‐Score on HIV protease and carbonic anhydrase, half of the evaluated target classes. The situation could be different when building family‐specific SFs, as complexes of a protein family tend to be more similar and thus more likely to be well described by a single precise characterization (e.g., adding features that are rare across all targets but common within the family such as the presence of a particular metal ion).…”

A machine learning approach towards the prediction of protein–ligand binding affinity based on fundamental molecular properties

“…Xue et al 24 used 94 drugs against human serum albumin and have used SVM models for prediction the binding affinity. Deng et al 27 has used 105 diverse protein-ligand complexes and Wang et al 29 used pdbbind version 2012. Wang et al…”