Drug Repurposing for Parkinson’s Disease with Deep Learning: A Virtual Screening Approach

This paper investigates Drug Repurposing for Parkinson’s Disease with Deep Learning. The methodology used is Drug Discovery via Virtual Screening. Various existing drugs were screened and binding scores obtained to find potential efficacy that may inhibit Parkinson’s Disease. An end-to-end Drug Repurposing pipeline was built for the study. We observe that Metformin, a widely administered Type-2 diabetes drug, Remdesivir an anti-viral drug and Simvastatin, a cholesterol drug, performed best with high binding scores on selected targets when compared to the FDA approved Parkinson’s drug Levodopa. In addition, Curcumin, a natural turmeric ingredient, also had high binding scores. The drugs identified had high efficacy for Parkinson’s SCNA and PINK1 targets. The research involved three different Deep Learning models on two benchmark datasets, namely KIBA and DAVIS. The methodology using Deep Learning was carried out after developing custom Python scripts that run on a Jupyter Notebook. The PD targets were further compared with a String diagram to identify the strength of interactions between them. The results also show strong potential when repurposed drugs are used along with existing FDA approved drugs for Parkinson’s Disease. Further research can be performed with in vivo clinical studies.