Intelligent Framework for Adverse Drug Event Identification Using Large Language Models and Retrieval-Augmented Generation: Development and Evaluation Study.

Gap: disease-targeted therapeutic (DTT) small molecules capable of modifying or significantly slowing AD progression are still clinically unavailable. Unabated, AD remains as an intractable progressive neurodegenerative plague that is poised to debilitate >150 million people by 2050 globally []. Due to the multi-factorial nature of AD, multi-mechanistic druggable targets seem to offer the opportunity to discover novel therapeutic single molecules capable of disrupting key cascades involved in disease progression. In search of central nervous system (CNS) druggable targets against AD, we decided on multi-mechanistic σ1 chaperone receptors []. Broadly, when σ1 receptors are activated, they bind to a diverse set of substrate proteins (ion channels included) and modify downstream signaling to yield neuronal protection in neurodegenerative disorders. In this contribution, our rationale for pursuing σ1 ligands is largely supported by such experimental evidence as: σ1 exhibits low receptor/biomarker expression densities in cognitive brain centers (prefrontal cortex, hippocampus, etc.) in early AD, σ1 agonists restored memory in AD animals and provided neuroprotection, and that σ1 activation yields anti-amnesic effects as well as stimulates synaptic plasticity [–]. As characterized in literature, σ1 receptors are 223 amino acid intra-mitochondria endoplasmic reticulum (ER) peptides with multi-functional signaling/neuroprotective roles [anti-inflammatory, anti-apoptosis, mitigating against reactive oxygen species (ROS) levels, increasing autophagy, β-amyloid (Aβ) plaque reduction, Ca excitotoxicity modulation, etc.] [–]. To date, ligands like Blarcamesine or ANAVEX®2-73 (a σ1 agonist/M2 antagonist) have yielded promising AD clinical trials data [,]. Curiously, multiple central nervous system (CNS) active drugs (anti-depressants, cholinesterase inhibitors, opioids, anti-psychotics, etc.) possess σ1 activity and are being revisited for anti-neurodegenerative effects [,]. We therefore designed/synthesized LA based σ1 ligands to possess enhanced hydrophobic binding interactions and blood brain barrier (BBB) penetrant logP values ≥2. Our molecules feature a common dithiolane – amide motif substituted with a sampling of electronically/conformationally disparate alkyl/aryl functionalities (available in our laboratory inventory at the time, for exploratory medicinal chemistry). Secondarily, even though we were less focused on the anti-ROS aspects of the derivatives, LA’s dithiolane antioxidant properties are worth mentioning because this can be a synergistically beneficial mechanism. Essentially, under physiological conditions the dithiolane moiety equilibrates as oxidized/cyclized LA and reduced/open ring metabolite dihydrolipoic acid (DHLA, ), and both molecules are known direct/indirect antioxidants []. It is further indicated that the antioxidant/anti-inflammatory properties partly reside in LA’s chelation of Zn, Pb, and Cu plus the more robust DHLA’s complexation with Fe, Zn, Hg, Pb, and Cu [,]. Interestingly, the metal hypothesis of AD implicates Al, Fe, Zn, and Cu imbalances as contributory factors []. Physiological redox equilibration of LA and DHLA by lipoamide dehydrogenase, glutathione reductase, and thioredoxin reductase [,]. This manuscript shares aspects of design, synthesis, and σ1 pharmacological binding affinities of twenty-one hydrophobic 5-(1,2-dithiolan-3-yl)pentanamides. Design considerations included structural adherence to the reported σ1 ligand’s pharmacophore illustrated in []. This pharmacophore proposes that σ1 ligand’s chemical basis of action is related to their ability to form a minimum of two hydrophobic and one H-bonding interactions within receptor’s active site [].