Downregulation of KRT proteins in the Neuroendocrine System of Amyotrophic Lateral Sclerosis with PT150 as a Multi-Target Neuroprotectant

Amyotrophic Lateral Sclerosis (ALS) is a fatal neurodegenerative disease that deteriorates motor neurons. Researchers have dis-covered proteins and genes such as C9orf72, Nfl, TDP43, SOD1, and FUS that exhibit unique traits in ALS patients. However, ALS still has a misdiagnosis rate of 55% and an expected lifespan of 3-5 years. This is partially due to the lack of specificity of existing biomarkers. Additionally, ALS mimic diseases, or diseases that mimic ALS very closely in pathology, often increase the diagnostic period as they have to be ruled out before a proper diagnosis can be made. This paper proposes the use of graph-based machine learning to model ALS heterogeneity along-side biological and neurospatial databases to discover and validate an ALS specific biomarker group along side molecular docking to discover a small-molecule therapeutic to modulate this biomarker. Two open source datasets were sourced. First, a gene expression profile with 7,038 gene expressions and second, a neuroproteomic network analysis with 2,105 proteins. To interpret the gene expression profile, a Graph Neural Network (GNN) was designed for classifiying each patient into either ALS, control, or ALS mimic that achieved an accuracy of 98% between ALS and controls and 85% between ALS and mimics. Consequently, Integrated Gradients identified the 20 most differentially expressed genes based on their affect on the GNN’s diagnosis of the ALS class. To interpret the proteomic profile, two-sample t-tests and Log2 Fold Change were applied with thresholds at 0.05 and 0.58 respectively. Consequently, this identified the 20 most differentially expressed proteins. After validating the biomarkers considering their neurospatial and pathological contexts, KRT6A, KRT6B, and KRT4 were identified as biomarkers in the pituitary gland with tau specificity of 86% showing potential to reduce the ALS misdiagnosis rate by up to 20% in clinical settings. They were shown to be correlated with cellular differentiation and human developmental pathways. These keratin proteins were down-regulated due to Glucocorticoid receptor a (GRa) inhibiting the activity of AP-1 transcription factor, which is supposed to regulate keratin proteins. Through molecular docking, it was shown that the PT150 ligand could inactivate Glucocorticoid Receptor Alpha and Cyclooxygenase 2 and regulate neuroinflammation and neuroendocrine dysregulation in ALS.