A multi-omics and machine-learning framework reveals shared druggable targets for Alzheimer's disease and periodontitis and identifies amentoflavone and taraxerone as potential dual-action natural products.

Alzheimer's disease (AD) and periodontitis (PD) are highly prevalent chronic disorders in ageing populations and share multiple pathological features, particularly persistent inflammatory and immune dysregulation. Despite increasing evidence supporting an association between the two diseases, the shared druggable molecular mechanisms underlying their comorbidity remain incompletely understood, and effective therapeutic strategies capable of simultaneously targeting shared disease-related processes are limited. This study aimed to identify shared druggable molecular candidates associated with AD and PD and to evaluate the potential biological effects of the natural products amentoflavone (AF) and taraxerone (TA) in AD-PD-related inflammatory cell models. An integrative strategy combining Mendelian randomisation (MR), transcriptomic profiling, machine learning, and structure-based virtual screening was used to identify and prioritise shared molecular candidates associated with AD and PD. Molecular docking and molecular dynamics (MD) simulations were performed to evaluate the predicted binding modes and interaction stability of AF and TA with selected core proteins. In vitro experiments were conducted using LPS-induced mouse BV2 microglia and mouse primary gingival fibroblasts (MGFs) to assess the effects of AF and TA, alone and in combination, on core-gene expression, inflammatory mediator secretion, oxidative stress, apoptosis, and iNOS/CD206-defined microglial activation states. MR, transcriptomic, and machine-learning analyses prioritised six shared key genes, including FCGRT, LTBP1, CD59, SPRED1, SURF2, and ZDHHC2, with FCGRT, SPRED1, and SURF2 selected for subsequent structural and cellular investigations. Molecular docking and MD simulations suggested stable predicted interactions of AF and TA with these three proteins. AF formed persistent hydrogen-bond networks, particularly in the SPRED1 and SURF2 complexes, whereas TA interactions were predominantly hydrophobic. MM-PBSA analysis yielded favourable binding free energies ranging from - 62.4 to - 99.7 kJ/mol. AF at 10 µM and TA at 20 µM were selected as non-cytotoxic working concentrations for subsequent cellular experiments. Both compounds partially restored LPS-associated alterations in FCGRT, SPRED1, and SURF2 expression, reduced TNF-α, IL-1β, and IL-6 secretion, increased IL-10 production, attenuated oxidative stress by reducing ROS, MDA, and NO levels while increasing SOD activity, decreased apoptosis, and shifted the iNOS/CD206-defined microglial activation profile away from an LPS-induced pro-inflammatory state toward an anti-inflammatory or repair-associated response. AF + TA co-treatment produced greater combined effects than either single treatment in several assays; however, quantitative combination analyses are required to determine whether these effects represent pharmacological synergy or additive interactions. This study identified FCGRT, SPRED1, and SURF2 as prioritised shared molecular candidates associated with AD and PD and provided computational and cellular evidence supporting further investigation of AF and TA as potential natural-product-based therapeutic candidates. AF and TA showed stable predicted interactions with the selected proteins and modulated gene-expression, inflammatory, oxidative-stress, apoptotic, and microglial-activation-related phenotypes in LPS-induced cellular models. These findings provide additional insight into the shared molecular landscape of AD and PD and support further experimental evaluation of AF and TA, including direct target-engagement studies, quantitative combination analyses, and in vivo validation.