HyDB — an observation corpus and sequence registry for PET hydrolases

Read full text
Authors: Oussama Chahed, Tesnim Chahed
Cited by: 1

Abstract

Version 1.9 (build 0.1.120) — the replay verdict no longer depends on the machine that runs it. In version 1.8 the audit hook allowed Linux paths written in the code (/usr, /lib, /bin) and therefore refused the macOS standard library under /Library/Frameworks/Python.framework/…: a reader on macOS obtained 46 replayed blocks where the author obtained 98, two measurements of one archive. The readable roots of the Python installation now come from sysconfig, and a path belonging to the curation tree is refused before any filesystem call — exists, isfile, stat, listdir, glob and sys.path insertions are covered, not only open. On this archive: registry suite 16 of 16, PETHyDB suite 98 of 102, 0 invariant failures, exit code 0. The four remaining blocks are no longer described in one lump: proof/PORTABILITY.tsv gives each one the first path it was refused and a declared reason — a third-party supplement, a raw source file, an internal intermediate table this deposit does not publish, and a scan of the curation tree. The raw-evidence manifest generator also stops listing editor lock files, which had no business being cited as evidence. HyDB is a provenance-tracked corpus of experimental observations on poly(ethylene terephthalate) (PET) hydrolases with the sequence registry they are anchored to. python3 VERIFY.py - standard library only - 22 property checks, 0 failures. Contents (1.7, internal build 0.1.118): 40,929 observations across five views - PET 28,094, MHET 1,317, SYS 5,056, CORE 6,289, AVAL 173; 12,147 at the direct-PET counter; 6,040 sequence records carrying 5,946 distinct normalised-residue fingerprints; 313 sources; 46 rounds. What changed in 1.7. Four corrections raised by an external audit, none of which alters a measured value. A factual error about PlasticDB is removed. Earlier versions stated that PlasticDB reports 875 proteins and that the difference from its FASTA export could not be reproduced. PlasticDB reports 875 microbial species and 329 proteins; the deposited FASTA snapshot contains 329 entries, of which 328 pass the common filter and yield 310 distinct fingerprints. The protein count and the snapshot agree, and the earlier wording implied a discrepancy that does not exist. The extended-retrieval paragraph now matches the deposited trace. Its 24 additional rows are not all article-only retrievals: nine come from data deposits, of which eight concorded, and 15 from article-only retrievals, of which three concorded. The profiles differ in evidential completeness and are reported separately. The deposited sample is portable. validation/semantic_attribution_sample.tsv carried absolute paths from the author's machine on all 238 rows: exact for one reader, unusable for every other. They are replaced by the source file name and its SHA-256, which identify the file without depending on where it sits. File counts are stated as measured. The archive holds 198 files in 15 directories; the manifest covers 196 of them, all but itself and the verifier. Two build defects found while making these corrections, and fixed at the source. A rebuild had silently omitted proof/upstream_curation_tables.zip and proof/upstream_manifest.tsv - ten megabytes of upstream proof material - because their temporary directory had been cleared; the build now refuses to produce a package without them instead of describing whatever it finds. And a Python import cache written after the manifest step had entered the archive unlisted, so the announced file count stopped being the package's own; caches are now removed before sealing. Reading the data. An empty condition cell means not reported, never zero. counts_direct_pet is not a boolean - test it for the exact string YES; the negative case is written NO, FALSE and NO_BY_VIEW. Version numbering. Zenodo version 1.7, following 1.6 (10.5281/zenodo.22100286), which remains available and unaltered. The filename carries the internal curation build number, 0.1.118. Archive: HyDB_public_v0.1.118.zip, 26652616 bytes, sha256 e6bec81fe8f730568ace51ba92b1187ebcebc228f4f0aec63087846bb866a37b.

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
26 August 2026
Language
EN
Document type
Dataset