Skip to content

docs: add TOX.md and relax the review rule to solo self-merge - #244

Merged
rvnminers-A-and-N merged 1 commit into
mainfrom
docs/tox-heads-and-solo-review
Jul 28, 2026
Merged

rvnminers-A-and-N merged 1 commit into
mainfrom
docs/tox-heads-and-solo-review

Conversation

@rvnminers-A-and-N

Copy link
Copy Markdown
Collaborator

Aroma and mouthfeel each have a canonical modality doc; the head family that most needs careful framing had none. docs/TOX.md covers the twelve Tox21 heads β€” why they differ in kind from the flavor heads (wet-lab assay measurements, not documented percepts), why they are deliberately kept out of the substitute-match vector (you don't want reformulation swaps ranked by shared toxicity signal), why the Tox21 training molecules are not imported into the flavor universe (industrial/pharma compounds would pollute the map and search), the per-assay CV-AUROC table (0.719–0.900) with positives/n, the featurization gotcha (bare 2048-bit Morgan _fp, not the _feat block β€” mixing them raises a 2060-vs-2048 shape error), where it surfaces, and the honest limits. The framing throughout: assay activity is not toxicity β€” it flags for review, never clears.

Also relaxes the review rule. CONTRIBUTING.md said "one approval required, routed by CODEOWNERS", but that can't be satisfied while the lead is the sole contributor β€” GitHub blocks approving your own PR, and main's protection currently requires 0 approvals, so the doc described a rule nothing enforced. It now states self-merge on green CI, with an explicit note to restore the approval requirement in the doc and in branch protection together once a second contributor is active.

Docs-only, no behavior change. ruff check . clean and 25 unit tests pass.

Aroma and mouthfeel each have a canonical modality doc; the head family that most
needs careful framing had none. docs/TOX.md covers the twelve Tox21 heads: why they
differ from the flavor heads (wet-lab assay measurements vs documented percepts),
why they are deliberately excluded from the substitute-match vector, why the Tox21
training molecules are NOT imported into the flavor universe, the per-assay CV-AUROC
table (0.72-0.90), the featurization gotcha (bare 2048-bit Morgan, not the _feat
block), where it surfaces, and the honest limits β€” assay activity is not toxicity.

CONTRIBUTING: the documented "one approval, routed by CODEOWNERS" rule cannot be
satisfied while the lead is the sole contributor (GitHub blocks self-approval and
branch protection requires 0 approvals), so it now states self-merge on green CI,
with a note to restore the approval requirement β€” in the doc AND in branch
protection β€” as soon as a second contributor is active.

Signed-off-by: Austin L. <86896075+rvnminers-A-and-N@users.noreply.github.com>
@rvnminers-A-and-N
rvnminers-A-and-N merged commit 15e2116 into main Jul 28, 2026
4 checks passed
@rvnminers-A-and-N
rvnminers-A-and-N deleted the docs/tox-heads-and-solo-review branch July 28, 2026 09:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant