SAveRUNNER: an R-based tool for drug repurposing

Abstract

Background: Currently, no proven effective drugs for the novel coronavirus disease COVID-19 exist and despite widespread vaccination campaigns, we are far short from herd immunity. The number of people who are still vulnerable to the virus is too high to hamper new outbreaks, leading a compelling need to find new therapeutic options devoted to combat SARS-CoV-2 infection. Drug repurposing represents an effective drug discovery strategy from existing drugs that could shorten the time and reduce the cost compared to de novo drug discovery.

Results: We developed a network-based tool for drug repurposing provided as a freely available R-code, called SAveRUNNER (Searching off-lAbel dRUg aNd NEtwoRk), with the aim to offer a promising framework to efficiently detect putative novel indications for currently marketed drugs against diseases of interest. SAveRUNNER predicts drug-disease associations by quantifying the interplay between the drug targets and the disease-associated proteins in the human interactome through the computation of a novel network-based similarity measure, which prioritizes associations between drugs and diseases located in the same network neighborhoods.

Conclusions: The algorithm was successfully applied to predict off-label drugs to be repositioned against the new human coronavirus (2019-nCoV/SARS-CoV-2), and it achieved a high accuracy in the identification of well-known drug indications, thus revealing itself as a powerful tool to rapidly detect potential novel medical indications for various drugs that are worth of further investigation. SAveRUNNER source code is freely available at https://github.com/giuliafiscon/SAveRUNNER.git , along with a comprehensive user guide.

Keywords: Drug repurposing; Network medicine; Network theory.

Fig. 1

SAveRUNNER conceptual organization. SAveRUNNER takes as input a list of drug-target interactions and disease-gene associations, and releases as output predicted drug-disease associations by performing seven steps (dashed box of this flowchart). In particular, Steps 1–3 bring to the construction of a proximity-based bipartite drug-disease network, where nodes are both drugs and diseases, edges are the statistically significant drug-disease associations (p value $\le 0.05$, or z-score $\le -1.65$), weighted according to the proximity values; Steps 4–7 bring to the construction of a similarity-based bipartite drug-disease network, where the weights represent the adjusted similarity measure computed to prioritize the predicted drug-disease associations by rewarding the associations between drugs and diseases belonging to the same network neighborhood. Finally, the drug-disease associations predicted by SAveRUNNER were evaluated by performing a ROC curve probability analysis (solid line box of this flowchart). The ROC curve is computed for SAveRUNNER algorithm by plotting the true positive rate (TPR) placed on Y-axis against the false positive rate (FPR) placed on X-axis at various threshold settings. Diagonal grey line represents the line of no-discrimination between positive class (known drug-disease associations) and negative class (unknown drug-disease associations)

Fig. 2

Used data resources. Summary of the all input data collection with the corresponding links to retrieve them

Fig. 3

ROC curves for predicting drug–disease associations. The ROC curve is computed for SAveRUNNER algorithm (light blue curve) and BiRW algorithm (orange curve) by plotting the true positive rate (TPR), i.e., sensitivity placed on Y-axis against the false positive rate (FPR), i.e., 1-specificity placed on X-axis at various threshold settings. Diagonal grey line represents the line of no-discrimination between positive class (known drug-disease associations) and negative class (unknown drug-disease associations)