petadex/petai-activity-labels

PETai Activity Labels

13,092 measured activity and stability labels on 1,219 plastic-degrading enzyme sequences, curated from 696 published references plus one in-house halo screen. Built for iGEM Toronto 2026 (project PetaBite) as the supervised label set for the PETadex corpus of 307,155,746 catalytic ORFs.

PETadex supplies candidates. This supplies the targets. No activity label exists anywhere in the PETadex release, so the two are published separately and joined on sequence MD5.

Read this before training on the split

The shipped split config is unusable for evaluation. The test fold holds 7 negative labels on 5 of its 170 sequences. The in-house halo screen contributes 168 of the 200 negatives in the release and was allocated 178 to train, 12 to val and 0 to test. No classification metric, calibration curve or enrichment number computed on this test fold is interpretable.

A second defect compounds it: label provenance is inverted across the folds. Train is 19.6% LLM-extracted and test is 62.9%, and the 4,898 LLM labels contain zero negatives. A model trained here learns from human-curated measurements and is scored on unvalidated machine extractions.

The split is shipped anyway because the homology-clean clustering underneath it is sound and worth reusing. It is a correct 30%-identity partition with a passing leakage check. Only the stratification is wrong. A corrected re-split is planned as v2. Until then, treat split as a cluster assignment rather than an evaluation protocol, and re-stratify it yourself.

Configs

Config	Rows	Description
labels	13,092	One measurement per row, keyed by seq_md5
sequences	1,219	One row per sequence, with cluster and split assignment
split	1,137	The 30%-identity fold assignment for the clean sequences
clusters	1,219	Cluster membership at 90%, 50% and 30% identity

raw/sequences.fasta holds all 1,219 sequences, named by MD5. raw/summary.json is the release summary emitted by the build.

labels schema

Column	Type	Notes
seq_md5	string	Join key, MD5 of the uppercase sequence
source	string	activity 6,655, llm 4,898, kinetics 1,342, halo 197
task	string	quantitative 12,895, binary 197
variable	string	What was measured, 99.95% present
unit	string	See the unit warning below
substrate	string	92.16% present. PET 4,287, pNPB 1,078, BHET 822
value	float	Not one target. See below
is_negative	bool	200 of 13,092 (1.53%)
ph	float	53.25% present
temperature_c	float	29.74% present
time_h	float	9.99% present
ref	string	696 distinct references
curator	string	89.75% present

value is not a single regression target

The column mixes incompatible units and spans -372.91 to 1.21e9 as a result. Do not regress on it directly. The three collisions that matter:

• Relative activity is split across % (1,370), percent (917) and % relative activity (369). Same quantity, three spellings.
• Melting data, 904 labels in all, mixes fraction of folded protein (751), which are CD spectroscopy curve points rather than a melting temperature, with °C (148), K (3) and mJ mg⁻¹ (2). Only the 151 temperature-unit rows are actual Tm values.
• Kinetics are three separate targets, not one: kcat (559), Km (551) and kcat/Km (473).

Clean target choices, both well populated: binary halo activity for classification, and PET monomer release for regression.

Conditions are trainable but unavailable at inference

pH is on 53.25% of labels and temperature on 29.74%, with both on 3,398 (25.95%). This is far better coverage than the PETadex corpus, which has melting_point and thermostability null on all 307M rows.

The asymmetry is the catch. A PETadex candidate mined from Logan has no measured pH or temperature, so a model conditioned on them at training time has nothing to condition on at inference. Condition dropout during training is the standard fix.

Overlap with the PETadex corpus is small

Joining on exact sequence MD5 against all 131 orfs shards of petadex/petadex-corpus-full, 401 of 1,219 labelled sequences (32.90%) appear in the corpus at all. Of those, 327 come from NCBI nr, 109 from PAZy and 21 from Logan.

Two consequences. Corpus-derived columns cannot be model features, since they exist for only 216 of the 796 train sequences. And the 21 Logan hits are the only direct evidence of how a model trained here behaves on the population it is meant to screen.

This is exact-MD5 overlap and the label set is full of engineered point variants, so it understates true homology coverage. An MMseqs2 search against the 22,235 30%-identity centroids is the number that actually bears on transfer, and is not yet run.

Cluster-size asymmetry

Bootstrap confidence intervals must resample clusters, not sequences.

Fold	Sequences	30% clusters	Mean size	Max	Top 3 share
train	796	70	11.37	124	43.84%
val	171	63	2.71	16	23.39%
test	170	63	2.70	14	22.94%

Effective training diversity is nearer 70 families than 796 sequences.

Fragments

82 of the 1,219 sequences are fragments or contained duplicates and carry no split assignment. They hold 3,188 labels (24.35% of the release), including 1,301 PET measurements and 773 of the 904 melting labels (85.5%).

73 of those 82 name a contained_in parent that is already in a fold, so 3,150 of the 3,188 labels can be recovered by remapping onto the parent. Only 9 sequences with 38 labels are genuinely unrecoverable. The remap is not applied in this version.

Provenance and limitations

• 8,194 labels are human-curated, 4,898 are LLM-extracted from published PDFs. The LLM subset is unvalidated against a human-labelled gold set at release and contains zero negatives, which is itself evidence of extraction bias rather than of biology. Filter on source != "llm" for an evaluation set.
• The halo screen is in-house iGEM Toronto wet lab data under the condition BHET25 @ 120h, active at z>3.0. It is a BHET assay, not a PET polymer assay, so its negatives mean "no BHET hydrolysis detected" rather than "not a PETase".
• Values are transcribed from assays run under widely differing protocols. Cross-reference comparisons are not controlled.
• value has no uncertainty estimate.

Licensing

Not yet settled. The underlying measurements are facts extracted from 696 published references, but the aggregation, the in-house halo screen and the curation are iGEM Toronto 2026 work and no license has been chosen. Contact the team before redistribution or commercial use.

Citation

@misc{petai_activity_labels_2026,
  title  = {PETai Activity Labels},
  author = {{iGEM Toronto 2026}},
  year   = {2026},
  url    = {https://huggingface.co/datasets/petadex/petai-activity-labels}
}

Built 2026-10-14. Published 2026-10-15. See igem-toronto issue 116.