JUN 12, 2025

Computational Design of de novo Transcription Factors for Targeted Genetic Repression

Abstract:

DNA-binding proteins (DBPs) play key roles in genetic regulation and manipulation in both natural and synthetic contexts. Aided by advances in machine learning and protein engineering, the design of de novo DBPs is now possible. While early attempts yielded small, single-chain proteins capable of sequencespecific DNA binding, these monomers could not alter gene expression. In contrast, many native transcription factors (TFs) are homodimers, enabling more protein-DNA contacts than de novo DBPs. We hypothesized that the dimeric nature of these TFs improves DNA-binding affinity and ensures the TF remains bound to the DNA even in the presence of native transcriptional machinery. To test this, de novo homodimeric TFs were computationally designed through a structure-based approach with the goal of achieving measurable gene repression in bacteria. RFdiffusion was used to design homodimeric TF backbones, which were subsequently fitted with an amino acid sequence using ProteinMPNN. The predicted sequences were folded with AlphaFold2, and the top 96 TFs were selected based on their predicted local distance difference test (pLDDT), predicted alignment error (PAE), and Cα atom root mean square distance (RMSD) when aligned to the RFdiffusion design. Selected TFs were expressed in E. coli, and their repression was measured through a fluorescence-based assay. Six TFs achieved over 4-fold repression, with the highest performing TF achieving nearly 20-fold repression. All TFs achieved considerable orthogonality and fold repression comparable to that of CRISPR interference systems, demonstrating that the design of effective de novo homodimeric TFs is indeed possible. In the future, the repression of de novo TFs might be enhanced by designing DNA-bending TFs or larger TF oligomers composed of more than two subunits. Successful designs could have applications in synthetic gene circuits, biosensors for various cellular processes, and robust therapeutics for genetic diseases.