Nature's economy as blueprint: regional overlap & block-based algorithm for protein design

Nature has optimized protein design through modular recombination of conserved and variable domains, allowing rapid adaptation with minimal energetic cost. Inspired by this principle, we present ROB-Fold, a deterministic rule-based algorithm that emulates natural evolution to identify conserved structural scaffolds across protein families.

Using experimentally validated and AI-predicted models from AlphaFold and the Protein Data Bank, ROB-Fold establishes motif centered analyses anchored on biologically meaningful patterns, into study cases, that can be extended for any protein family, such as CXXC in protein disulfide isomerases (PDIs) and EEVD in HSP70 chaperones. Around each anchor, bidirectional expansions are generated to detect recurrent sequence and structural elements. The algorithm translates residue frequencies into Chou-Fasman structural equivalence classes, projects consensus patterns onto secondary structure, and quantifies geometric persistence via Root Mean Square Deviation (RMSD).

Application to the PDI family revealed a conserved β-α-β-α-β core centered on the CXXC catalytic motif, delineating the thioredoxin-like scaffold responsible for redox activity. Similarly, analysis of the HSP70 family identified a recurrent "GPXVEEVD" pattern defining the nucleotide-binding domain, with conserved α-helical arrangements surrounding the EEVD motif. Across both cases, conserved regions exhibited RMSD values below 15 Å, confirming structural persistence and validating the cross-family generalization of the rule-based logic.

These findings demonstrate that explicit structural rules can recover evolutionary invariants typically detected by stochastic models while maintaining interpretability and computational efficiency. ROB-Fold thus provides a transparent and biologically grounded framework for identifying natural scaffolds, enabling a guide for the rational design of novel proteins.