Alzheimer’s disease (AD) is the most prevalent form of dementia, accounting for 60-80% of cases. Currently, there is no cure, and existing medications are palliative. It is a complex, multifactorial disease whose pathophysiological mechanisms are not fully understood. In recent years, the enzyme DYRK1A has emerged as a promising therapeutic target in AD, as it is involved in multiple biological functions. Specifically, studies have shown that alterations in DYRK1A, such as the phosphorylation of proteins like TAU and APP, correlate with AD progression. Therefore, DYRK1A is a highly promising target for designing new drugs to treat AD. Our objective in this work is to use predictive and generative AI tools to design non-toxic DYRK1A inhibitors. We present a successful strategy for integrating AI-based methods in the de-novo generation of molecules within a binary target drug discovery framework. The generative process is detailed, demonstrating its effectiveness through in-vitro studies focused on candidate molecules designed to inhibit DYRK1A kinase. This approach was conducted under a small-data regime, utilizing various AI techniques to develop an optimal model for generating viable candidates. The most promising candidate, with a novel structure, has been synthesized along with a family of derivatives. All of these compounds have been evaluated enzymatically against DYRK1A, showing nanomolarlevel activity, as well as antioxidant and anti-inflammatory potential. Thanks to the synergy between AI, computational techniques such as docking, and organic chemistry, we have developed a family of DYRK1A inhibitors with a highly promising pharmacological profile.

This work was financially supported by Projects:Projects:”IND2023/BMD 27452” financed by CAM,“2021TED2021-129970B-C21” financed by MCIN/AEI/10.13039/501100011033 and European Union “NextGenerationEU”/PRTR and Project ”SNEO 2022207" (NEOTEC) financed by CDTI (MCIN) and European Union “NextGenerationEU”/PRT.

1 p.-5 fig.-2 tab.

Peer reviewed