Computational reconstruction of Tyrannosaurus rex collagen for biomanufacturing of paleo-inspired biomaterials with tuneable stiffness

Abstract

Ancient protein fragments preserved in fossilized tissues offer evolutionary sequence information inaccessible through conventional genomics. Here we demonstrate computational reconstruction of full-length Tyrannosaurus rex type I collagen (COL1A1) by integrating paleo-proteomics data with protein language models. Using TOC large protein model with pattern-constrained generation to maintain collagens Gly-X-Y triplet architecture, we generated five computationally designed sequences incorporating validated T. rex peptide fragments. These generated variants maintain triple helix structural features (23.0% Gly-X-Y content) and exhibit enhanced lysine content (up to 84 residues vs. 57 in wild-type chicken) to increase enzymatic crosslinking capacity. To enable spatial and temporal control of material properties such as stiffness, we engineered a doxycycline-inducible lysyl oxidase-4 (LOX4) expression system in TOC t-rex cell line that achieved 9-fold enzyme induction. This biomimetic approach, inspired by functionally graded fish scale architecture, decouples constitutive collagen production from inducible crosslinking, enabling biomanufacturing of materials with tuneable mechanical gradients. We validate stable transgene integration, confirm tight transcriptional control, and demonstrate system compatibility with scalable cell culture. This work establishes a platform combining paleo-proteomics, protein language modelling, and bioinspired engineering to produce next-generation biomaterials from computationally reconstructed extinct proteins.