Xu et al. used virtual screening techniques to analyze 1200 DrugBank small molecules and identify agonists for Retinoid X receptor alpha (RXRα). RXRα plays an important role in several types of cancers and neurodegenerative diseases. The virtual screening successfully identified several agonists and antagonists of RXRα.

Pascal Heitel et al. used virtual screening techniques to analyze marketed drugs in the DrugBank database to identify two drugs that have the potential to modulate the activity of liver X receptor (LXR). The authors used DrugBank to build a dataset of drug structures and compounds that interact with the target of interest to implement an experimental machine learning technique called a self-organizing map.

Zali et al. used several techniques to virtually screening techniques the dataset of approved drugs from DrugBank and identify potential strong inhibitors of TLR4. The screening technique made use of the chemical structures and pathway information from DrugBank. Several strong inhibitors were identified that have potential for further research.

Using drug profiles from DrugBank, along with genomic data from BioVU, Challa et al., developed an algorithmic approach to help with selecting drugs for repurposing prioritization. Drug repurposing is the practice of using approved drugs to treat something separate from it’s originally approved purpose. It is often accompanied by an unmanageable number of drugs to prioritize (sometimes upwards of 11,000 agents), but with this new approach, the authors were able to select a manageable subset of around 250 drugs to research.

Using virtual screening methods, Singh et al. were able to investigate the DrugBank database for possible drugs that could target the allosteric site of Abl kinase, which is affected when there is a “chromosomal rearrangement resulting in the expression of Bcr-Abl fusion protein with deregulated Abl tyrosine kinase activity”. This rearrangement results in chronic myleoid leukemia (CML). Current approved drugs for CML target the ATP-binding site of Abl kinase, but the body often becomes resistant. They believe by finding drugs that target the allosteric site, which is remote from the ATB-binding site, is useful and through their screening, they were able to find a promising drug combination that could more effectively treat CML.

Using in-house and DrugBank databases, Ngo et al. used virtual screening using prediction models and molecular docking to identify and discover hit compounds to reverse multidrug resistance (MDR). The team utilized the P-gp structure (which plays an important role in MDR) and its binding structure as the predictor for the docking study determining interactions between ligands and receptors. In their study they attempted to “restore cancer cell sensitivity to cytotoxic drugs.

Peng et al. used the drug-target relationships from DrugBank to explore novel machine learning techniques for predicting drug-target relationships. DrugBank drug structures and drug-target relationships were used for training and validation of machine learning models. The team then explored the possibility of using the techniques for drug-repurposing and drug discovery by predicting unknown drug-target interactions and comparing these results to known relationships in DrugBank.

Utilizing drug interactions information on the DrugBank database, Zie et al. were able to build a deep-learning-based model to predict potential interactions. They used interaction prediction “as a binary classification task” and were able to show that the framework predicted more credible interactions than traditional experiments that are often difficult and timely due to data scarcity.