Skip to content

LNG_BOUNDARY_HIERARCHY: hierarchical root-complex decomposition (off by default) - #97

Merged
adamjohnwright merged 4 commits into
mainfrom
feat/boundary-hierarchy
Sep 26, 2026
Merged

adamjohnwright merged 4 commits into
mainfrom
feat/boundary-hierarchy

Conversation

@adamjohnwright

Copy link
Copy Markdown
Contributor

Decompose a root complex one hasComponent level at a time: join components that already exist as nodes, build missing nested complexes, fall back to terminal leaves. Fixes the IFN alpha/beta severing (ISGF3 [cytosol] never reached root ISGF3:KPNA1:KPNB1). Off by default; A/B in deltasignal specs/030.

🤖 Generated with Claude Code

adamjohnwright and others added 2 commits September 26, 2026 08:09
…el at a time -- join components the network already has, build missing nested complexes (deltasignal specs/030)

Flat decomposition went straight to base leaves, so a species the pathway
produces never reached the root complex that contains it (IFN alpha/beta:
ISGF3 [cytosol] -> root ISGF3:KPNA1:KPNB1, no binding reaction in Reactome;
200 held-out cases severed). Off by default; downstream-reuse rule unchanged.
Three tests. The one failing test (Class I MHC edges proportional) fails
identically on main -- it checks stale generated output.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…228, 238/10, p=9e-58; +28 outside IFN alpha/beta; experimental +2)

Flat-mode tests that stub every id as a Complex are pinned to LNG_BOUNDARY_HIERARCHY=0;
the leaf-reuse tests run under the new default with a one-level component stub.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
@adamjohnwright
adamjohnwright marked this pull request as ready for review September 26, 2026 13:06
…(two joins had welded a 162-node DSB loop); deterministic root order; nested complexes built per root; prefer a ROOT copy when joining (PDGFB: the pinned root copy never reached PDGF A/B heterodimer). All hierarchy-only; flat mode byte-identical. Three new tests, two mutation-checked.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
… uuid string (a uuid4 is new every run, so the join landed on a different copy run to run); root-ness for the root-copy preference is the pathway's own non-depletion targets, not a live in-degree (which counted this function's joins, so choice followed stid sort order, and counted depletion). Three tests, all red on the previous commit.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
@adamjohnwright
adamjohnwright merged commit 590301c into main Sep 26, 2026
4 checks passed
@adamjohnwright
adamjohnwright deleted the feat/boundary-hierarchy branch September 26, 2026 16:20
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