Machine learning-optimized long single-stranded DNA synthesis technology empowers high-precision diagnostic–therapeutic integration in living cells

While DNA nanotechnology holds transformative potential across biomedical and information storage applications, current technologies face critical limitations in synthesizing long single-stranded DNA (ssDNA) with high purity and homogeneity. To address these challenges, we developed Ouroborosyn-ssDNA, a nicking enzymatic assisted replication (NEAR) platform that synergizes enzymatic engineering with computational optimization. By integrating phi29 DNA polymerase and Nb.BbvCI nickase in formate-based buffers, we achieved extended ssDNA synthesis up to 15 000 nt while preserving sequence fidelity, resulting in a 4.73-fold yield enhancement compared to commercial buffers. Notably, machine learning-guided parameter optimization identified magnesium ion dynamics and thermal modulation as pivotal determinants of enzymatic efficiency. Furthermore, solid-phase synthesis using thiol-gold immobilized templates demonstrated 86.38% purification recovery via automated magnetic bead systems, enabling scalable production. To validate functional utility, we engineered six-helix bundle DNA origami-CRISPR complexes that achieved nucleolin-targeted genome editing in cervical cancer cells, coupling GFP-based diagnostics with therapeutic E7 oncogene disruption. These advancements directly overcome key limitations in enzymatic stochasticity and product heterogeneity through buffer engineering and computational optimization, establishing a scalable pathway for applications in precision nanomedicine, synthetic biology, and molecular data storage. This integrated strategy advances DNA nanotechnology from proof-of-concept studies toward standardized biomanufacturing of sequence-defined macromolecular architectures.