Comprehensive Network Map of ADME-Tox Databases

In the last decade, many statistical-based approaches have been developed to improve poor pharmacokinetics (PK) and to reduce toxicity of lead compounds, which are one of the main causes of high failure rate in drug development. Predictive QSAR models are not always very efficient due to the low number of available biological data and the differences in the experimental protocols. Fortunately, the number of available databases continues to grow every year. However, it remains a challenge to determine the source and the quality of the original data. The main goal is to identify the relevant databases required to generate the most robust predictive models. In this study, an interactive network of databases was proposed to easily find online data sources related to ADME-Tox parameters data. In this map, relevant information regarding scope of application, data availability and data redundancy can be obtained for each data source. To illustrate the usage of data mining from the network, a dataset on plasma protein binding is selected based on various sources such as DrugBank, PubChem and ChEMBL databases. A total of 2,606 unique molecules with experimental values of PPB were extracted and can constitute a consistent dataset for QSAR modeling.