Computational Preprocessing Framework for Small Molecule and α-synuclein Data towards Parkinson’s Drug Repurposing

Parkinson’s disease (PD) is a progressive
neurodegenerative disorder characterized by the loss of
dopaminergic neurons and abnormal aggregation of the α
synuclein protein, resulting in motor and cognitive deficits. The
complexity of PD pathology and the restricted blood-brain
barrier (BBB) permeability of most compounds present major
challenges for therapeutic discovery. This study presents an
implemented computational preprocessing framework for
curating small-molecule and α-synuclein datasets, supporting
reliable drug-target interaction (DTI) prediction and drug
repurposing. During drug preprocessing, a multi-step curation
pipeline was executed to standardize and refine PubChem
compounds through canonicalization, charge normalization,
tautomer correction, and PAINS filtering, followed by Lipinski
and Veber criteria with CNS multiparameter optimization to
identify molecules with optimal drug-like neuroactive profiles.
The protein preprocessing stage employed a confidence-guided
refinement approach on the AlphaFold-Predicted α-synuclein
model (AF-P37840-F1-v6), combining per-residue pLDTT
scoring, masking of low-confidence regions, and intrinsic
disorder was assessed with IUPred2A, and structural repair
under physiological conditions to generate a confidence
weighted structure optimized for computational modelling.
Together, these harmonized preprocessing stages establish a
reproducible, domain-specific foundation for subsequent
ensemble docking, feature extraction, and graph-based
learning. This framework enhances data quality, consistency,
and interpretability in computational neurotherapeutics,
supporting the future discovery of small-molecule modulators
that effectively target α-synuclein and potentially slow
Parkinson’s disease progression.
Keywords—data preprocessing; machine learning; molecular
data standardization; intrinsically disordered protein; Parkinson’s
disease