Abstract
Transthyretin amyloidosis (ATTR) is a genetically diverse disorder caused by destabilising mutations in the transthyretin (TTR) protein, leading to pathological aggregation. While stabilisers like tafamidis and acoramidis are approved, their efficacy across TTR variants remains unclear. This study presents an in silico pipeline combining AlphaFold3 for structure prediction, ESM2 for sequence embeddings, DiffDock-L and AutoDock Vina for molecular docking, and DiffSBDD for ligand generation. Simulations show that binding affinities of approved ligands vary significantly among TTR variants, with some mutations (e.g., W61L, Y98F) reducing binding despite being distant from the binding site. Embedding-based clustering highlights potential benign mutations and supports scalable variant classification. Additionally, customised ligand optimisation can recover binding affinity in specific cases, though effects are mutation-dependent. These findings emphasise the need for variant-aware therapeutic strategies. This integrative approach offers a foundation for precision drug design in ATTR, enabling the development of personalised stabilisers tailored to individual mutational profiles.