Objective: To use an in silico screen to identify compounds that have potential to reduce α-synuclein (aSyn) oligomers and are amenable to drug repurposing for Parkinson’s disease (PD).

Background: Development of disease-modifying therapies for PD and translation into clinical use is expensive and slow. Repurposing of compounds, previously proven to be safe in humans and approved by regulatory agencies could reduce costs and accelerate drug development. However, methods to prioritize candidate drugs for repurposing are needed. IBM Watson for Drug Discovery (WDD) is a cognitive computing platform able to extract domain-specific text features (e.g., drugs, diseases) from the literature and identify connections between entities of interest. We used WDD to generate a predictive model to rank potential candidates for drug repurposing for PD.

Methods: We developed: 1) a training set of 15 chemical compounds known to reduce aSyn oligomers in vitro and/or in vivo based on published studies, and 2) a candidate set composed of all 620 individual active compounds in the Ontario Drug Benefit program database. WDD analyzed hundreds of thousands of Medline abstracts to learn text patterns and develop a semantic fingerprint for each compound and then, using machine learning, generated a predictive model to rank compounds from the candidate set based on their semantic similarity to the training set. 

Results: Leave-one-out cross-validation demonstrated that each compound in the training set was highly ranked by the model, suggesting that highly ranked compounds from the candidate set would have properties common to the training set. Following ranking of candidate compounds, directed PubMed searches and exploration using WDD applications for the top 52 compounds revealed: 9 compounds with existing evidence for inhibition of aSyn aggregation (4 of which have not yet been studied in human clinical trials or epidemiological studies of PD), and 12 compounds not previously associated with aSyn but with biologically plausible links to aSyn aggregation.  

Conclusions: Our approach using WDD to mine scientific literature to rank compounds with potential to reduce aSyn oligomers is novel and promising. Future work will perform necessary validation of prioritized compounds using both in vitro and in vivo models of aSyn aggregation and toxicity, as well as epidemiologic studies assessing incidence and outcomes in PD.

References: The authors would like to acknowledge: the Ontario Brain Institute and the Government of Ontario for providing access and training to the IBM Watson Drug Discovery platform.

« Back to 2017 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/in-silico-predictive-analytics-accelerating-identification-of-potential-disease-modifying-compounds-for-parkinsons-disease/