Improving compound–protein interaction prediction by building up highly credible negative samples

Abstract

Motivation: Computational prediction of compound–protein interactions (CPIs) is of great importance for drug design and development, as genome-scale experimental validation of CPIs is not only time-consuming but also prohibitively expensive. With the availability of an increasing number of validated interactions, the performance of computational prediction approaches is severely impended by the lack of reliable negative CPI samples. A systematic method of screening reliable negative sample becomes critical to improving the performance of in silico prediction methods.Results: This article aims at building up a set of highly credible negative samples of CPIs via an in silico screening method. As most existing computational models assume that similar compounds are likely to interact with similar target proteins and achieve remarkable performance, it is rational to identify potential negative samples based on the converse negative proposition that the proteins dissimilar to every known/predicted target of a compound are not much likely to be targeted by the compound and vice versa. We integrated various resources, including chemical structures, chemical expression profiles and side effects of compounds, amino acid sequences, protein–protein interaction network and functional annotations of proteins, into a systematic screening framework. We first tested the screened negative samples on six classical classifiers, and all these classifiers achieved remarkably higher performance on our negative samples than on randomly generated negative samples for both human and Caenorhabditis elegans. We then verified the negative samples on three existing prediction models, including bipartite local model, Gaussian kernel profile and Bayesian matrix factorization, and found that the performances of these models are also significantly improved on the screened negative samples. Moreover, we validated the screened negative samples on a drug bioactivity dataset. Finally, we derived two sets of new interactions by training an support vector machine classifier on the positive interactions annotated in DrugBank and our screened negative interactions. The screened negative samples and the predicted interactions provide the research community with a useful resource for identifying new drug targets and a helpful supplement to the current curated compound–protein databases.Availability: Supplementary files are available at: http://admis.fudan.edu.cn/negative-cpi/. Contact: sgzhou@fudan.edu.cnSupplementary Information: Supplementary data are available at Bioinformatics online.

Providing data science support for systems pharmacology and its implications to drug discovery

“…or explore a heterogeneous network of chemical and target space using graph mining, as discussed in the previous section. By using randomly generated or algorithmically derived negative cases [65], the drug-target interaction prediction can be formulated as a supervised classification problem. When the data points are sufficiently large, wide and deep Artificial Neural Network (ANN) outperformed random forests, gradient boosted decision tree ensembles, logistic regression, and several other state-of-the art methods in QSAR modeling [66,67] and compound virtual screening [68], and achieved the top performance in the compound toxicity prediction challenge [69].…”

Scalable and Accurate Drug–target Prediction Based on Heterogeneous Bio-linked Network Mining

“…They can produce physiological effects by binding to a "target" or multiple "targets" to enhance or inhibit functions carried out by these "targets" (Olayan, et al, 2017;Santos, et al, 2017). Exploring novel drug-target interactions (DTI) plays a crucial role in drug development, which facilitate the studying of drug action, disease pathology and drug side effects (Liu, et al, 2015;Santos, et al, 2017;Yuan, et al, 2016). Despite the availability of a variety of biological assays, experimental prediction remains laborious and expensive (Liu, et al, 2015;Wang, et al, 2013), which drives the biochemical experimentation (in vitro) to focus on some particular families of 'druggable' proteins, while the potentially larger number of small molecules are rarely systematically screened for paring (Vogt and Mestres, 2010;Yıldırım, et al, 2007).…”

“…Exploring novel drug-target interactions (DTI) plays a crucial role in drug development, which facilitate the studying of drug action, disease pathology and drug side effects (Liu, et al, 2015;Santos, et al, 2017;Yuan, et al, 2016). Despite the availability of a variety of biological assays, experimental prediction remains laborious and expensive (Liu, et al, 2015;Wang, et al, 2013), which drives the biochemical experimentation (in vitro) to focus on some particular families of 'druggable' proteins, while the potentially larger number of small molecules are rarely systematically screened for paring (Vogt and Mestres, 2010;Yıldırım, et al, 2007). Consequently, in order to lower the overall costs and uncover more potential screening targets, computational (in silico) methods have become popular and are commonly applied for poly-pharmacology and the drugs repurposing in drug development (Cheng, et al, 2012;Ding, et al, 2013).…”

“…Early computational methods attempted to predict associations by docking simulations based on the 3D structure of target proteins, or by comparing a query ligand with the ligands of target proteins. These methods were hindered by the lack of ample data, such as 3D structures of proteins or ligands with target proteins (Liu, et al, 2015;Olayan, et al, 2017;Yuan, et al, 2016). To provide the more robust solutions (Cheng, et al, 2007;Zhu, et al, 2005), machine learning algorithms were applied to predict drug-target associations.…”

“…These methods can generally be categorized into two types: 1) classification-based methods, and 2) inference-based methods (Ding, et al, 2013;Jacob and Vert, 2008;Klipp, et al, 2010). The features used as the input for the diverse classification models play an important role, and the features are generated from three types, 1) chemical structures/fingerprint of drugs and genomic sequence for proteins (Feng, et al, 2018;Liu, et al, 2015;Peng, et al, 2017;Wen, et al, 2017;Yuan, et al, 2016), 2) neighborhood or structure information for vertices (e.g., drugs or targets) in a heterogeneous network structure (Emig, et al, 2013), 3) similarity matrix. It is important to note that similarities are not feature vectors (Ding, et al, 2013).…”