“The most fruitful basis for the discovery of a new drug is to start with an old drug”: James Black (Nobel Laureate pharmacologist)
Drug repurposing (also known as drug repositioning, drug redirecting and drug re-tasking) represents the investigation of existing drugs for novel disease therapeutic. The existing drugs could be from any of the stages of the drug such as, already approved, archived drugs, drugs from currently undergoing clinical trials and discontinued. Drug repurposing is more effective and attractive in disease therapeutic when there is limited understanding of disease causality more specifically in epidemic and pandemic situation.
At present, due to the worldwide COVID-19 pandemic, drug repositioning has taken on a new urgency. Several approaches from different research domains have contributed independently or in combination towards drug repositioning. Among these are the computational approaches along with biological experiments gaining popularity. The available and accessible standardised heterogeneous information such as expression (gene/protein) level data, network information (genomic, proteomic, pathway and different semantic and expression), drug-target interaction, structural information, reports of patients of clinical trials, drug adverse event have paved the way of drug repurposing.
Different high-throughput experiments generate large-scale heterogeneous data. However, the data are changing day by day with the updated and new experimental results. Recent advanced computational approaches are capable to integrate all these large-scale data and acquire knowledge to generate novel insights into drug repositioning mechanism.
Computational and Mathematical approaches have attracted considerable attention in drug repurposing. Multi-level omics data is applied for computational drug repurposing using signature or structure-based approaches. The transcriptomic signature-based drug repositioning is a classification problem of drug identification that depends on the target transcriptomic signature. In structure-based approaches, a potential drug target is identified by assessing the complementary shape structure of the target protein (either conformed or computationally modelled 3D structure) and binding strength between them.
Computational drug repurposing is broadly classified into two groups; machine learning-based, network model-based.
Drug-drug similarity-based logistic regression is integrated with disease-disease similarity to predict the similar drug as repurposing target. A similarity- based approach is used in logistic regression learning to identify the therapeutic chemical class of a drug for repositioning. Multi-level information such as drug chemical structure, molecular target, gene expression data are combined to extract the drug similarity matrix to design an SVM kernel function for repurposing drug identification. Recently, deep architecture is utilised for drug repurposing strategy with the novel structural embedding of a drug molecule and the target structure.
Network models are widely used in drug repurposing as well as drug discovery strategy. Generally, the networks are constructed from multiple sources of data that represent various interacting information (protein-protein, gene-gene, drug-gene, drug-drug, and drug-disease, signalling pathway) or relationship between individual entities such as drug, protein, disease, gene, SNP. These heterogeneous networks are primarily utilised to infer the potential drug (existing)-disease pair from the candidate set of interactions. The semantic and text mining- based approach provides valuable insights into the network model for drug repurposing.
In a nutshell, computationally repurposing strategies significantly reduce the time and cost of de novo drug discovery. In any epidemic and pandemic situation drug repurposing become, the only alternative as there is limited understanding of any novel disease. Computationally approaches employed with multi-omics data analysis improve drug repositioning efficiency.