Biochemical Logic Computation Through Repurposing Natural bZIP Protein Interaction Networks

Abstract

Competitive protein dimerization networks offer an alternative to transcriptional genetic circuits by enabling fast molecular decision-making through direct protein-protein interactions. Implementing these networks currently requires challenging de novo protein design. In this work, we circumvent this limitation by repurposing natural, experimentally pre-characterized basic leucine zipper (bZIP) transcription factor networks. By keeping natural binding affinities and modulating only component monomer concentrations via a customized genetic algorithm, we comprehensively evaluate the computational versatility and robustness of these natural substrates. We identified 135 individual networks capable of implementing Boolean logic, with the most versatile natural clusters computing up to 15 of the 16 possible two-input logic gates, including the non-linearly separable XOR and XNOR. Computational versatility increased with network size and connectivity, and robustness analysis revealed that many optimized networks preserved reliable logical behavior despite substantial stochastic expression noise. These results demonstrate that natural bZIP networks possess substantial latent computational capacity and can perform reliable biochemical computation through concentration tuning alone, providing a realistic foundation for developing scalable protein-based biocomputing platforms for future synthetic biology applications.