Enhancing DNA oligomer self-assembly efficiency via a hairpin-constrained sequence optimization approach

DNA-mediated multivalent protein self-assembly requires precise design of single-stranded DNA target pairing regions to achieve efficient DNA oligomer self-assembly. However, intrastrand pairing leading to hairpin structure formation during this process often reduces self-assembly efficiency. In this study, we developed a hairpin-constrained DNA (HC-DNA) sequence optimization approach, which integrates a deep learning-based prediction model for accurate hairpin structure identification and employs an intrastrand mutation mechanism to effectively reduce intrastrand pairing length. Experimental results demonstrate that while maintaining the stability of the target pairing region, HC-DNA markedly reduces the melting temperature of single-stranded hairpins, achieving a substantially greater reduction than approaches without hairpin constraints. Furthermore, DNA tetramer self-assembly experiments demonstrate that HC-DNA-optimized sequences exhibit substantially improved self-assembly efficiency and homogeneity compared with sequences designed without hairpin constraints. Overall, HC-DNA provides an efficient computational design tool for the precise construction of DNA oligomers and multivalent molecular assemblies. Efficient DNA-mediated protein self-assembly is often hindered by hairpin structures formed through intrastrand pairing, reducing assembly efficiency. Here, the authors introduce a hairpin-constrained DNA optimization approach using deep learning to reduce hairpin stability, thereby enhancing self-assembly efficiency and homogeneity.