AI-driven de-novo design and development of non-toxic DYRK1A inhibitors for Alzheimer’s disease

In humans, dual-specificity Tyrosine-Phosphorylation-Regulated Kinase 1A is an enzyme encoded by the DYRK1A gene involved in various diseases, including DYRK1A syndrome, cancer, diabetes, and neurodegenerative pathologies such as Alzheimer’s disease (AD). AD is the most prevalent form of dementia, accounting for 60–80% of cases and remains an unmet medical challenge with no cure and just palliative treatments. Recent studies have identified DYRK1A as a promising therapeutic target in AD, given its involvement in multiple biological functions and its alterations correlated with AD progression. In this work, we leverage multiple Artificial Intelligence (AI) tools, including pre- dictive models and generative algorithms, to design non-toxic DYRK1A inhibitors. We construct a dual-target drug discovery framework integrating AI-driven meth- ods with classical techniques to identify novel compounds. An ensemble Quanti- tative Structure-Activity Relationship (QSAR) model is employed for predicting compound affinities, while Directed Message Passing Neural Networks (DMPNN) are used to assess toxicity. In a generative phase, a Hierarchical Graph Genera- tion model (HGG) facilitated the design of potential DYRK1A inhibitors. Promis- ing candidate molecules were refined through classical docking studies, leading to their synthesis and experimental validation. As a result, pyrazolyl-1H -pyrrolo[2,3- b]pyridine was identified as a potent DYRK1A inhibitor, leading to the synthesis of a new derivative series. Enzymatic assays demonstrated nanomolar-level inhibitory activity, while anti-inflammatory and antioxidant properties were confirmed through ORAC assays and LPS-induced pro-inflammatory response evaluations in BV2 mi- croglial cells. Pharmacological testing revealed that the mentioned compound and its derivatives exhibit significant DYRK1A inhibition alongside robust antioxidant and anti-inflammatory effects.