Limits...
Large-scale Direct Targeting for Drug Repositioning and Discovery.

Zheng C, Guo Z, Huang C, Wu Z, Li Y, Chen X, Fu Y, Ru J, Ali Shar P, Wang Y, Wang Y - Sci Rep (2015)

Bottom Line: However, the biological means on a large scale remains challenging and expensive even nowadays.The available computational models mainly focus on predicting indirect interactions or direct interactions on a small scale.This shows that the WES method provides a potential in silico model for drug repositioning and discovery.

View Article: PubMed Central - PubMed

Affiliation: Bioinformatics Center, College of Life Sciences, Northwest A&F University, Yangling, Shaanxi, 712100, China.

ABSTRACT
A system-level identification of drug-target direct interactions is vital to drug repositioning and discovery. However, the biological means on a large scale remains challenging and expensive even nowadays. The available computational models mainly focus on predicting indirect interactions or direct interactions on a small scale. To address these problems, in this work, a novel algorithm termed weighted ensemble similarity (WES) has been developed to identify drug direct targets based on a large-scale of 98,327 drug-target relationships. WES includes: (1) identifying the key ligand structural features that are highly-related to the pharmacological properties in a framework of ensemble; (2) determining a drug's affiliation of a target by evaluation of the overall similarity (ensemble) rather than a single ligand judgment; and (3) integrating the standardized ensemble similarities (Z score) by Bayesian network and multi-variate kernel approach to make predictions. All these lead WES to predict drug direct targets with external and experimental test accuracies of 70% and 71%, respectively. This shows that the WES method provides a potential in silico model for drug repositioning and discovery.

No MeSH data available.


The performance of the WES model based on CDK, Dragon, and CDK-Dragon features.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4496667&req=5

f1: The performance of the WES model based on CDK, Dragon, and CDK-Dragon features.

Mentions: Feature analysis. To investigate the effects of different structural features of the ligands on the model performance, we have used the Chemical Development Kit (CDK), Dragon and the CDK-Dragon hybrid features for model construction, respectively (see Methods for details). Table 1 illustrates the results in terms of precision and recall rates. Clearly, the hybrid model outperforms both the CDK and Dragon ones in recovering the negative links. Notably, the hybrid model for the leave-one-out cross-validation (LOOCV) performs well in predicting the binding (sensitivity 85%, SEN) and the non-binding (specificity 71%, SPE) patterns, with the accuracy of 78%, the precision (PRE 74%) and the area under the receiver operating curves (AUC) of 0.85, respectively. It is noted that all the scores (Z score for CDK and Dragon model and likelihood for CDK-Dragon hybrid model), used to make prediction, in this work were selected when the models achieve the highest F1 score in cross-validation otherwise specified (see Methods for details). The ROC curves (Fig. 1) show that all the three models are capable of catching sufficient information related to detect interactions at high true-positive rates against low false-positive rates at any threshold. With the increase of the AUC in the complete dataset, the hybrid model improves the ability to identify those known drug-target links, demonstrating that more chemical and pharmacological information introduced to build models can achieve better predictive activity.


Large-scale Direct Targeting for Drug Repositioning and Discovery.

Zheng C, Guo Z, Huang C, Wu Z, Li Y, Chen X, Fu Y, Ru J, Ali Shar P, Wang Y, Wang Y - Sci Rep (2015)

The performance of the WES model based on CDK, Dragon, and CDK-Dragon features.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4496667&req=5

f1: The performance of the WES model based on CDK, Dragon, and CDK-Dragon features.
Mentions: Feature analysis. To investigate the effects of different structural features of the ligands on the model performance, we have used the Chemical Development Kit (CDK), Dragon and the CDK-Dragon hybrid features for model construction, respectively (see Methods for details). Table 1 illustrates the results in terms of precision and recall rates. Clearly, the hybrid model outperforms both the CDK and Dragon ones in recovering the negative links. Notably, the hybrid model for the leave-one-out cross-validation (LOOCV) performs well in predicting the binding (sensitivity 85%, SEN) and the non-binding (specificity 71%, SPE) patterns, with the accuracy of 78%, the precision (PRE 74%) and the area under the receiver operating curves (AUC) of 0.85, respectively. It is noted that all the scores (Z score for CDK and Dragon model and likelihood for CDK-Dragon hybrid model), used to make prediction, in this work were selected when the models achieve the highest F1 score in cross-validation otherwise specified (see Methods for details). The ROC curves (Fig. 1) show that all the three models are capable of catching sufficient information related to detect interactions at high true-positive rates against low false-positive rates at any threshold. With the increase of the AUC in the complete dataset, the hybrid model improves the ability to identify those known drug-target links, demonstrating that more chemical and pharmacological information introduced to build models can achieve better predictive activity.

Bottom Line: However, the biological means on a large scale remains challenging and expensive even nowadays.The available computational models mainly focus on predicting indirect interactions or direct interactions on a small scale.This shows that the WES method provides a potential in silico model for drug repositioning and discovery.

View Article: PubMed Central - PubMed

Affiliation: Bioinformatics Center, College of Life Sciences, Northwest A&F University, Yangling, Shaanxi, 712100, China.

ABSTRACT
A system-level identification of drug-target direct interactions is vital to drug repositioning and discovery. However, the biological means on a large scale remains challenging and expensive even nowadays. The available computational models mainly focus on predicting indirect interactions or direct interactions on a small scale. To address these problems, in this work, a novel algorithm termed weighted ensemble similarity (WES) has been developed to identify drug direct targets based on a large-scale of 98,327 drug-target relationships. WES includes: (1) identifying the key ligand structural features that are highly-related to the pharmacological properties in a framework of ensemble; (2) determining a drug's affiliation of a target by evaluation of the overall similarity (ensemble) rather than a single ligand judgment; and (3) integrating the standardized ensemble similarities (Z score) by Bayesian network and multi-variate kernel approach to make predictions. All these lead WES to predict drug direct targets with external and experimental test accuracies of 70% and 71%, respectively. This shows that the WES method provides a potential in silico model for drug repositioning and discovery.

No MeSH data available.