AI-augmented prediction of high-risk PINK1 variants associated with Parkinson's disease: integrating multilayered bioinformatics, MD simulation, and deep learning.

Parkinson's disease is a prevalent neurodegenerative disease, in which genetic mutations in many genes play an important role in its pathogenesis. Among these, a mutation in the PINK1 gene, a mitochondrial-targeted serine/threonine putative kinase 1 that protects cells from stress-induced mitochondrial dysfunction, is implicated in autosomal recessive Parkinsonism. However, the exact etiology is not well understood. Therefore, this study aimed to identify the most damaging non-synonymous single-nucleotide polymorphisms (nsSNPs) distributed in the kinase domain of the PINK1 gene and their structural and functional alterations using a range of bioinformatics and deep learning tools. Next, to find the possible impact of these mutations on PINK1 interactions and binding affinities, a protein-protein interaction and molecular docking analysis were conducted. Finally, molecular dynamics (MD) simulations were performed to observe the stability and dynamic behaviour of the pathogenic SNPs on the PINK1 protein over time. Our integrated bioinformatics and deep learning approaches predicted 5 SNPs (C166R, E240K, D362N, D362Y, and C388R) as high-risk candidates for disrupting PINK1 structure and function. In conclusion, we propose that the pathogenicity of these variants may provide an important clue to understanding the mechanism by which pathogenic nsSNPs contribute to PD, thereby enhancing future diagnostic value for the disease and serving as potential targets for new drugs.