Concatenative nonmonotonicity and optimal links in HP protein folding models
Bjørn Kjos-Hanssen
TL;DR
This work shows that in the $HP$ protein folding framework, the optimal fold energy is nonmonotone under concatenation across several standard lattices, contradicting the idea that adding sequence is always beneficial. It uses $Z$-optimality and isoperimetric arguments to construct explicit counterexamples in 2D and 3D rectangular, triangular, and hexagonal lattices, and demonstrates analogous nonmonotonicity via a Berger-Leighton–style construction. Beyond nonmonotonicity, the paper establishes that closed-chain folds can realize nontrivial knots and links as uniquely optimal folds for certain inputs, including a trefoil knot example, and demonstrates, under the levels-of-hydrophobicity model, that certain links are indispensable for optimality. Together, these results reveal rich topological structure in HP-folding models and connect computational protein folding questions to knot/link theory, with implications for how concatenation can nontrivially affect fold quality and for topological constraints in lattice-based protein models.
Abstract
The hydrophobic-polar (HP) model represents proteins as binary strings embedded in lattices, with fold quality measured by an energy score. We prove that the optimal fold energy is not monotonic under concatenation for several standard lattices, including the 2D and 3D rectangular, hexagonal, and triangular lattices. In other words, concatenating two polymers can produce a fold with strictly worse optimal energy than one of the polymers alone. For closed chains, we show that under the levels-of-hydrophobicity model of Agarwala et al. (1997), proper links can arise as uniquely optimal folds, revealing an unexpected connection between HP models and knot/link theory.
