PETase ANnotation and Triage System

Therapeutic PETase Design: PANTS mines metagenomic sequence space for PET-degrading enzymes to enable protein engineering and therapeutic discovery.

PET is the plastic in drinks bottles and polyester clothing, and fragments of it now turn up in human blood and tissue. A few bacterial enzymes can cut PET apart, but almost every one was found or engineered for an industrial recycling reactor: hot, alkaline, nothing like a human body. A therapeutic enzyme has to work at 37 °C in blood, and the published lineage has been moving steadily the other way.

So PANTS looks somewhere else. Metagenomics reads the DNA of whole microbial communities straight from the environment, without culturing anything, reaching the vast majority of bacteria nobody has ever grown in a lab. PANTS searches four communities with decades of plastic exposure and every reason to have evolved a use for it: the human gut, compost, the ocean surface where debris collects, and landfill. Some 15 million protein sequences are scanned for the polyesterase fold, then filtered on whether the three catalytic residues actually meet in space rather than merely appearing in the sequence. Survivors get a predicted structure, an active-site measurement and a rank.

Lineage and findings

Seven views of the same 25 enzymes. Every value is read from the table below, so nothing here is a separate transcription that could drift from it.

Named enzymes and their lineages 25 of 342 measured

This is the curated subset: the enzymes with a name, a known lineage and a structure. It is not the whole reference set. 342 enzymes carry a published measurement, almost all from PAZy, and those are what the activity head is trained and evaluated on — see all of them.

Grouped by lineage with each wild type heading the variants built on it. Read down the T_opt column within a lineage and the pattern is the one that motivates this project: engineering runs away from 37 °C, the therapeutic target, which is marked on every row at or below it. Each row opens a page carrying its structure, active-site geometry, mutation set, measured activity and every external identifier.

^† cleft width measured over fewer than four aromatic residues: it is a percentile of the distances between them, so with three instead of five it is not comparable with the rest of the column. TurboPETase is the case in point, and it is not an artefact: two of its eight substitutions delete an aromatic side chain outright. Where a source records several optima, the one measured on PET is shown in preference to one measured on a model ester, and failing that the one with the most independent publications behind it; hover any value for the basis and the alternatives. That choice matters: LCC's optimum is 50 °C on pNP-butanoate but above 70 °C on PET, and the first would place it near the therapeutic end of this table for a substrate the project is not about. T_opt is in °C and cleft width in Å; the coloured edge marks the family. %ID is against the wild type at the root of that enzyme's own lineage, not against IsPETase: LCC-ICCG is 98.6% identical to LCC and only about 50% to IsPETase, and only the first says how far engineering moved it. Wild types are 100% by definition. Optimum temperatures come from UniProt where an experimental value is curated and from the literature review otherwise; each enzyme page shows which, with its evidence code. Where a value is published as a range the table shows the midpoint and the page shows the range. Cleft width is measured on the structure named in the Structure column, so it mixes crystal structures with predictions; the page says which for each.

By source environment

Every marine plastisphere candidate sits below 40% identity to anything characterised: nothing in that cohort is a rediscovery. Compost, where the field has already looked, returns near-identical copies of known enzymes.

Ranked by retrieval bitscore

The retrieval baseline, which deliberately favours what is already known: the interesting candidates are the low-identity ones this ranking pushes down.