A highly limited amino acid library from asteroid Bennu yields wide-ranging protein folds.

AI protein design software is used to explore the world of ancient protein structures, and very small, primordial amino acid libraries are found to produce a wide variety of key folds. Data from asteroid analyses and protobiotic chemistry experiments are utilised to constrain several primitive amino acid libraries of varying sizes. Protein design software is then used to construct sequences of these amino acids that could emulate a broad range of key protein folds including metabolic (all enzymes of the reverse tricarboxylic acid cycle), redox, electron-transfer and ribosomal, among others. AlphaFold2 is employed to predict 3D structures from these sequences, which are compared to their native counterparts using the Template Modelling Score. A library of only 6 amino acids-those of highest measured abundance on asteroid Bennu-can reproduce all folds in the test set. Two libraries of just 7 amino acids, constrained by the Miller-Urey experiment and Murchison meteorite data, and one library with 8 amino acids constrained by another Miller-Urey experiment, are also able to produce all folds considered. It is also demonstrated that a 6 amino acid alphabet-the 5 most abundant on Bennu supplemented by cysteine that could have been supplied by atmospheric haze chemistry-can yield a ferredoxin-like protein with a plausible Fe-S binding geometry. Such broad protein folding with very limited amino acid libraries has significant implications for the origins of life, synthetic biology and medical applications.