Computational design of blue melanin by motif‐scaffolding a pentapeptide‐repeat protein

Abstract

Blue coloration in natural and synthetic melanins arises from light reflection rather than absorption, that is, structural rather than pigmentary color. Producing melanin with blue pigmentary color is challenging as it requires the formation of long, homogeneous melanin polymer from tyrosine (Tyr) amino acid. Here, we describe a bottom-up design strategy for producing blue melanin by polymerizing Tyr-containing pentapeptides into a ~60-unit polymer that absorbs at 617 nm. To narrow the sequence space from 160,000 possible amino acid combinations, we used RFdiffusion, a protein deep learning model, to motif-scaffold a pentapeptide-repeat protein (PRP) to identify sequences ideal for blue melanin formation. Top peptide designs were experimentally validated, generating de novo melanin colors (blue, purple, green) alongside known colors (red, yellow, brown). Under acidic conditions, blue melanin remained thermally stable at autoclave temperature of 121°C and photostable for weeks under 600 lux illumination. We also demonstrated the application of blue melanin as an electrophoretic ink. De novo melanin design from simple peptides with sequence-to-color tunability could potentially transform how colorants are sourced and produced. Our approach with computational design should also inspire the development of new deep-learning tools to directly predict colors from amino acid sequences.