mRNA output ergonomics: naming, separators, ranking-report (#270) - #271
Merged
Conversation
The header still mixed pre-mRNA language ("frozen config groups",
``EpitopeConfig`` / ``VaccineConfig`` class names) with post-mRNA
content. Section is now described in user terms:
- modality-agnostic vs peptide-only is called out per section,
- ``manufacturability``'s actual role (tie-break + report
auto-skip) is documented instead of an outdated
``-> VaccineConfig`` mapping that no longer holds,
- the resolution order (CLI > per-modality > cross-modality >
built-in default) is named once at the top so the per-section
text doesn't have to re-state it.
First batch of fixes from the issue. Touches the mRNA writer + CLI plumbing; the broader items (mRNA ranking-decisions report, configurable antigen-name template) ride in follow-up commits on this branch. - Construct names: ``seq_NNN`` -> ``mrna_NNN``. Mixed peptide+mRNA outputs in the same downstream pipeline collide on a bare ``seq_001``; modality-stamped names mirror the peptide writer's ``peptide_NNN`` scheme. - FASTA ``antigens=`` field now uses ``;`` instead of ``,``. Comma is a token boundary in many FASTA-parsing tools' description-line handlers; ``;`` is the GenBank convention for compound description fields. - ``layers.csv`` -> ``mrna-sequence-parts.csv``. ``layers`` next to ``cds.fasta`` / ``full.fasta`` is too cryptic; the new name spells out what each row describes. - ``--mrna-max-constructs`` cap warning now reports the dropped count: ``dropping 379 / 384 ranked antigen(s) past the cap`` rather than the legacy single-name ``including X``. Operators can tell whether they lost 1 or 379. Tests pin the FASTA ``;``-separator format, the dropped-count warning shape, and the cross-modality construct-name pattern (``mrna|peptide_NNN``). Bump version to 2.18.0.
Continuation of #270 ergonomics. Real-LENS-run feedback drove these: - mrna assembly cap is **top-k selection**, not an error. Log at INFO instead of WARNING; drop "spill" / "drop" panic language. Names selected count, total ranked, and the next-best not-selected antigen for traceability. - ``--mrna-max-constructs`` default: 1 -> **2** (BioNTech FixVac canonical = 2x pentatope, 10 antigens at 5/construct, Sahin 2017). Bare default now produces a credible vaccine out of the box on a real LENS file. - ``_resolve_mhc_for_linker_optimizer``: thread MHC alleles inferred from LENS / pVACseq through to the per-junction optimizer (was failing on the LENS path because the external arg parser doesn't carry --mhc-alleles). Targeted hint when alleles are inferred but --mhc-predictor is missing instead of one generic "set both" warning. - ``_log_args_summary``: replace the flat ``Namespace(...)`` dump (100+ keys, no grouping) with grouped sections — Inputs / MHC / Vaccine design / Peptide / mRNA / Processing / Outputs / Reports / Config — and hide values that match parser defaults unless ``--verbose``. Auto-inferred state (LENS-derived alleles, …) surfaces under a separate ``[auto-inferred]`` block. - LENS antigen_source breakdown logged BEFORE filter messages so the operator sees the input composition before the per-filter drop counts. - LENS rows lacking ``rna_reads_covering_genomic_origin`` now log at INFO with the antigen_source breakdown — ERV / SPLICE / FUSION rows are *expected* not to carry that column. Only flags SNV / INDEL gaps as a WARNING (those should always have it). - Pepsickle "peptide not in pep_context" warning: drop the "File upstream if this rate is non-trivial" tail. - "report_df has N duplicate (peptide, allele) rows" demoted from INFO to DEBUG. Multi-source duplicates are by design, not actionable; logging them at INFO every run was noise. Filed as follow-ups (not in this PR): - #272 — Pepsickle Predictions vs EpitopePredictions split - #273 — Config flatten: drop vaccine_constructs wrapper, move manufacturability under peptide
Five housekeeping fixes from the post-review punch list, plus the
test debt:
- Move ``_MISSING_SENTINEL = object()`` above ``_log_args_summary``
so it's defined before the function that reads it.
- Replace ``__import__('logging').DEBUG`` with ``logging.DEBUG``
in epitope_io.py — the module imports ``logging`` at the top.
- Cache ``_ARG_GROUPS_BY_NAME = dict(_ARG_GROUPS)`` at module level
so the per-call ``dict(_ARG_GROUPS).get(...)`` reconstruction
goes away.
- Drop the misleading "OR is a non-empty string" claim from
``_log_args_summary``'s docstring — only the
differs-from-default check controls visibility.
- Narrow the bare ``except Exception`` in
``_resolve_mhc_for_linker_optimizer`` to
``(AttributeError, ValueError, KeyError)``. A genuine model-load
error (e.g. missing weights) inside mhctools must propagate, not
be swallowed into "linker optimizer disabled".
- ``LENS_PROVENANCE_MARKER`` constant in
``vaxrank.external_input`` shared between producer (the LENS
PatientInfo synth) and consumer (the entry-point alleles
filter), so the string can't drift between the two.
Test debt added (tests/test_entry_point_helpers.py):
- Branch tests for ``_resolve_mhc_for_linker_optimizer``: alleles
inferred but predictor missing, both missing, real model-load
failure propagates.
- Smoke + ``--verbose`` + auto-inferred-section tests for
``_log_args_summary`` (with parser_defaults snapshot captured
*before* overrides, mirroring the real arg-parser snapshot).
- caplog-style ordering test for the up-front LENS antigen-source
breakdown — must log before the per-filter "skipped X rows"
lines, and SNV / INDEL must precede other kinds with
``(missing)`` last.
Tests use a local capture handler attached directly to the target
logger, not pytest's ``caplog``: importing
``vaxrank.cli.entry_point`` transitively triggers isovar's
``logging.config.fileConfig`` which wipes pytest's caplog handlers
from the root logger. Pre-existing project quirk; documented in
the test module.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Addresses #270 (mRNA output ergonomics). First batch lands in commit 1; the broader items (mRNA ranking-decisions report, configurable antigen-name template, signal-peptide / start-codon polish) ride in follow-up commits on this branch before merge.
Commit 1 — naming, separators, dropped-count
seq_NNN→mrna_NNN. Modality-stamped namespace mirrors the peptide writer'speptide_NNNscheme so mixed peptide+mRNA outputs in the same downstream pipeline don't collide on bareseq_001.antigens=separator:,→;. Comma is a token boundary in many FASTA-parsing tools' description-line handlers;;is the GenBank convention.layers.csv→mrna-sequence-parts.csv. Next tocds.fasta/full.fasta, the legacy name was too cryptic — the new name spells out the per-element decomposition.--mrna-max-constructscap warning now reports dropped count:dropping 379 / 384 ranked antigen(s) past the capinstead of naming only one antigen asincluding X. Operators can tell whether they lost 1 or 379.Coming on this branch (TODO)
{gene}_{protein_change}style; long form retained in manifest).default.yamlheader rewrite (drop pre-mRNA "frozen config groups" / EpitopeConfig / VaccineConfig language; commitd2a51bd).Default.yaml header
Already in commit 0 on this branch. The PR also rewrites the header to match the post-2.17 layout (drops outdated class-name mappings, adds the resolution-order line at the top instead of repeating it per section).
Test plan
./test.sh— 714 passed, 0 failures;-separator, dropped-count warning shape, cross-modalitymrna|peptide_NNNpatternBumps version to 2.18.0.