The amino acid alphabet was adaptively shaped by biophysical constraints of foldability

Modern proteins rely on an almost universal repertoire of 20 canonical amino acids which can be divided into 'early' and 'late' groups, based on when they were added to the amino acid alphabet.  However, many nonproteinogenic amino acids were also prebiotically available but not incorporated.  In a study investigating what drove this selectivity, researchers made peptides out of sets of amino acids and compared their solubility and how readily they arranged themselves into protein-like structures.  Starting from a core of seven early amino acids, Hlouchová and coworkers added sets of other amino acids, made peptide libraries from each, and compared them with peptides made from 19 of the canonical amino acids. Two of the libraries included noncanonical but prebiotically available amino acids and one included amino acids with linear side chains such as norvaline and norleucine. 

The canonical library formed more secondary structures like α helices and β sheets than either libraries made using only early amino acids, or libraries that contained one or more noncanonical amino acids.  The results suggest that foldability was a critical factor in the selection of the canonical alphabet, with unbranched aliphatic amino acids being purged from the proteinogenic alphabet - despite their high prebiotic abundance - because they generate peptides that are oversolubilised and have low packing efficiency.  The early canonical alphabet was most likely selected for supporting peptide folding, and was better adapted for creating proteins that bind cofactors, recognise RNA, and perform enzymatic catalysis.