We have recently reported computational models for prediction of cell-based anticancer activity using machine learning methods.

EK-DRD provides three retrieval methods for quickly searching and displaying the drug repositioning data, namely, text mining, chemical structure search, and protein sequence search. For chemical structure search, five algorithms, namely, substructure search, Markush search, two-dimensional (2D) and three-dimensional (3D) similarity calculations, and hybrid structure-similarity calculations, are used in EK-DRD. 2D similarity calculations are based on the FP2 fingerprint and performed using OpenBabel (O'Boyle, et al., 2011). 3D similarity adopts the weighted Gaussian algorithm (WEGA) for molecular-shape-similarity calculations (Yan, et al., 2013), which provides shape-, feature-and coefficient-based shape-feature combo-scoring functions for user selection.Our group also encoded in EK-DRD a new hybrid-similarity metric for calculating compound similarity that combines 2D fingerprint and 3D shape, called HybridSim, which was developed and validated to outperform the popular 2D FP2-, MACCS-, and 3D WEGA-based similarity methods (Shang, et al., 2017). All similarity methods use Tanimoto coefficient as a similarity function to quantify the similarity between two molecules. BLAST algorithm is used for protein sequence similarity search (Altschul, et al., 1990). We also developed an online network-display tool to virtually display the relationship among drugs, repositioning putative protein targets, and related diseases, in the form of an interactive network.

Hybrid-molecule-based drug design is the combination of two or more bioactive molecules into a single chemical entity. This strategy may be used to achieve better affinity and efficacy or improved properties compared with the parent molecules, to interact with two or multiple targets, to reduce undesirable side effects, to decrease drug–drug interactions, or to reduce the emergence of drug resistance. The approach offers the prospect of better drugs for the treatment of many human diseases. Research activity in this area is increasing and has attracted many practitioners worldwide. To accelerate the design and discovery of new hybrid-molecule-based drugs, it is essential to properly collect and annotate experimental data obtained from known hybrid molecules. To address this need, we have developed HybridMolDB (), a manually curated database dedicated to hybrid molecules for chemical biology and drug discovery. It contains structures, manually annotated design protocols, pharmacological data, some physicochemical properties, ligand efficiency, drug-likeness, and ADMET characteristics, and the biological targets of known hybrid molecules. HybridMolDB supports a range of query types, including searches by text, protein sequence, chemical structure similarity, and property ranges. The database serves as an open source facilitating the development and/or optimization of related in silico tools for the design and discovery of hybrid-molecule-based drugs and chemical probes.

Drug repositioning, or the identification of new indications for approved therapeutic drugs, has gained substantial traction with both academics and pharmaceutical companies because it reduces the cost and duration of the drug development pipeline and it reduces the likelihood of unforeseen adverse events. So far, there has not been a systematic effort to identify such opportunities, in part because of the lack of a comprehensive resource for an enormous amount of unsystematic drug repositioning information to support scientists who could benefit from this endeavor. To address this challenge, we developed a new database, Experimental Knowledge-Based Drug Repositioning Database (EK-DRD) by using text and data mining, as well as manual curation. EK-DRD contains experimentally validated drug repositioning annotation for 1861 FDA-approved and 102 withdrawn small molecule drugs. Annotation was done at four levels, using 70,212 target assay records, 3999 cell assay records, 585 organism assay records, and 8910 clinical trial records. Additionally, approximately 1799 repositioning protein or target sequences coupled with 856 related diseases and 1332 pathways are linked to the drug entries. Our webbased software displays a network for integrative relationships between drugs, their repositioning targets, and related diseases. The database is fully searchable and supports extensive text, sequence, chemical structure, and relational query searches. Availability: EK-DRD is freely available at