Genome-scale screening of drug-target associations relevant to Ki using a chemogenomics approach

Abstract

The identification of interactions between drugs and target proteins plays a key role in genomic drug discovery. In the present study, the quantitative binding affinities of drug-target pairs are differentiated as a measurement to define whether a drug interacts with a protein or not, and then a chemogenomics framework using an unbiased set of general integrated features and random forest (RF) is employed to construct a predictive model which can accurately classify drug-target pairs. The predictability of the model is further investigated and validated by several independent validation sets. The built model is used to predict drug-target associations, some of which were confirmed by comparing experimental data from public biological resources. A drug-target interaction network with high confidence drug-target pairs was also reconstructed. This network provides further insight for the action of drugs and targets. Finally, a web-based server called PreDPI-Ki was developed to predict drug-target interactions for drug discovery. In addition to providing a high-confidence list of drug-target associations for subsequent experimental investigation guidance, these results also contribute to the understanding of drug-target interactions. We can also see that quantitative information of drug-target associations could greatly promote the development of more accurate models. The PreDPI-Ki server is freely available via: http://sdd.whu.edu.cn/dpiki.

Figure 1. Outline of our methodology.

(A) Interaction features are calculated by combing the fingerprint descriptors from drugs and the CTD and amino acid composition descriptors from protein sequences. These feature vectors are used to find the optimal RF parameters which most accurately separate the positive and negative training sets. The independent validation sets are used for further validation for the RF model. (B) Once the RF model is constructed, we can predict new unknown drug-target associations or screen all cross-linking associations.

Figure 2. ROCs and precision-recall curves for Naïve Bayes (green) and random forest (red) with full and selected features.

(A) ROCs (B) precision-recall curves.

Figure 3. The plot of K_i versus prediction probability on 5-fold cross validation.

non-interaction: red and interaction: green. Linear relationship between K_i and prediction probability could be observed with correlation coefficient of 0.65.

Figure 4. ROCs and precision-recall curves with different K_i thresholds using RF.

(A) ROCs (B) precision-recall curves. The auPRCs drop with the decreasing of K_i thresholds. However, the varying trend of auROCs is consistent with that of auPRCs.

Figure 5. The predictive probability plot of screening all cross-linking drug-target pairs. The size of predictive probability gradually varies from green to red.

Figure 6. Drug-target interaction network using drug-target pairs with prediction probability above 0.99.

Drugs and targets are presented by red circle and blue triangle, respectively. Drug-target interactions are represented by the edges connecting related drugs and targets.