Systematic evolution of ligands by exponential enrichment (SELEX)-based aptamer discovery can be constrained by target availability, selection biases, and optimization requirements. We introduce Xelari, a modular pipeline for de novo DNA and RNA aptamer design from protein structures without target-specific SELEX data or retraining. Xelari combines algorithmic modules with surrogate models trained on physical or structural references. InNA reproduced nucleotide-amino-acid interaction energies with RMSE of 0.19 kcal/mol across 483,735 reference pairs. AliNA achieved a mean F1 of 0.757 on 339 PDB-derived RNA evaluation sequences. Edgar correlated with experimental folding energies but did not reproduce their absolute scale, supporting its use as a ranking heuristic. Under the reported calibration, ValentiNA showed MAE of 0.931 kcal/mol and Spearman ρ of 0.795 across 32 retrospective records associated with 27 structural templates; this analysis does not establish family-independent validation. Prospectively, three DNA aptamers designed against Treponema pallidum protein structures showed cell-associated binding in a treponeme assay (apparent K D , 0.91–2.57 μM). A fourth DNA aptamer served as a capture reagent in a serum-based hybrid aptamer-antibody assay for Tau441. These applications support feasibility without establishing intended-target engagement in the cellular assay or a general success rate, positioning structure-based design as a complement to experimental selection.
