An AI-assisted genomic census of 193,342 assemblies finds a GFP-barrel family in unreported lineages, carrying pseudogenes in both fungi and cnidarians

Abstract Green fluorescent protein and its homologues share an eleven-stranded beta-barrel enclosing a central helix, on which three consecutive residues cyclise to form the chromophore. A family of these barrels carrying a buried network of two arginines, two glutamates and a tyrosine, with a hydrophobic residue in place of the chromophore tyrosine, was recently assembled from protein records by Schneider and Zimmer. Here 193,342 genome assemblies were searched in all six reading frames, returning 396 loci, of which 140 read the complete buried network and collapse to 77 distinct carrying organisms. The charge network family reaches two sponge genera, a coronate jellyfish, three dinoflagellate transcriptome records and an archaeal genome, none of them represented in any proteome. In fungi and in cnidarians a substantial minority of loci carry an in-frame stop or a frameshift: 36 of 144 fungal loci, and 18 of 66 cnidarian loci across eleven genera. A Dollo reconstruction over a 28-genome Aspergillus clade requires two to six independent losses at the species level. The copies that survive are under purifying selection, at a median ratio of non-synonymous to synonymous substitution of 0.146 per species pair over 39 such pairs, and in one anthozoan lineage the gene is instead tandemly amplified to eleven copies. Two novel bacterial copies sit in mobile DNA, one inside a phiE-type prophage in Burkholderia pseudomallei and one beside an IS4 family transposase within a toxin-antitoxin array in a Myxococcota genome. A family that protein records show as intact is, in genomic sequence, intact in some lineages and breaking down in others.