Psilocybin gene cluster

The psilocybin genes (psiD, psiK, psiM, psiH, psiR and the transporters psiT1 and psiT2) sit together on chromosome 10. The figure shows every place in this stretch where the genomes in the collection differ, and which genomes carry which version. i Each genome's assembly is compared with the P. Envy reference (MGC_Penvy_1). At each site, a filled cell means the genome carries the less common of the two versions among the genomes whose assembly covers the site, whether that is the reference's base or the change, so the picture does not depend on which culture happens to be the reference. Genomes with similar patterns are grouped together; the reference is outlined in orange. Samples you are not authorized to see are left out of the rows but still counted in the frequencies.

Method: minimap2 (asm20) alignment of each assembly to GCF_017499595.1, paftools variant calls, bcftools merge over NC_063008.1:1,655,000-1,678,000, coding consequences with bcftools csq. Assemblies covering less than half of the reference are not placed. Gene models are the RefSeq annotation with four corrected: RefSeq fuses psiT1 and psiK into one protein and stops psiH before its heme-binding cysteine, so psiT1, psiK, psiH and psiT2 follow curated models that reproduce the characterised proteins (UniProt P0DPB1, P0DPA8, P0DPA7, P0DPB2).

The seven enzymes of the cluster as 3D models, with every amino-acid change seen in the collection. Pick an enzyme, then flip between the Reference (the P. Envy reference genome's protein) and an Alt version: by default the most common version that differs from P. Envy, or any other version from the list. The track under the model shows each change along the protein; its height is the share of samples that carry it. P. Envy itself carries the less common residue at several of these sites.

Method: each sample's seven proteins are read from its de novo assembly, or from a consensus of its reads where the assembly misses part of a gene (miniprot with the characterised UniProt proteins, intron curation, exon-model transfer for psiT1 and psiM, length, stop-codon and motif checks); a protein with an ambiguous position is not counted. Every distinct protein is folded once with ESMFold, P. Envy's included, and superposed on P. Envy's model. Each sample contributes one protein per enzyme; we do not say how many copies a culture carries. A single amino-acid change rarely moves a predicted backbone, so the flip mostly shows the changed side chains and where they sit; larger shifts in pale (low prediction confidence, pLDDT < 50) regions are not real differences. "Buried" means less than 25% of the residue is exposed to solvent in the model; for the transporters, which sit in a membrane, "surface" can mean facing the membrane.