Repository navigation
Conversation
1. partial3: a mid-gene premature stop (projection truncates the CDS and demotes path exons to three_prime_UTR) is no longer reported as a complete 3 end; partial3=false now requires the projected CDS to reach the terminal exon (empty three_prime_UTR). 2. honor the GFF3 phase column of prediction CDS rows: each chain carries the phase of its 5-prime-most CDS row and uses it to seed inherited frames, so 5-prime-partial inputs are no longer frame-shifted. 3. try_add_exon no longer inserts exons that carry an in-frame stop in every reading frame (empty determine_good_phases result was let through the guard). 4. terminal/single exons trim their trailing 3 bp only when those 3 bp actually encode a stop under the configured genetic code (CDS rows of many pipelines exclude the stop codon). 5. the genome-fetch-failure fallback in resolve() no longer emits unchecked CDS marked complete (hit_stop=false there now). 6. fully repeat-masked introns are registered with score 0 (EVM parity): masking zeroes an intron score but must not sever trellis connectivity (TE-intron genes in repeat-rich genomes stay whole). Each fix carries a targeted unit test. Full suite passes (194 lib tests); golden consensus output unchanged on the EVM testing data set.
Adds three opt-in behaviours developed while validating on a 2.5 Gb, 80%-repeat plant genome (all off by default; default output unchanged, goldens untouched): - --require-atg-start: prediction exons typed initial/single whose start codon is not ATG are demoted so they cannot begin a gene (EVM canonical START gate, limited to the start codon). At resolution a CDS not starting with ATG is trimmed in-frame to the first ATG (<=120 codons); when the path holds no in-frame ATG at all (truncated evidence start), the 5-prime-most exon is extended upstream (cap 5 kb) to the first ATG bounded by the nearest upstream in-frame stop, with the 5 UTR trimmed to just past that stop. - consensus-side 3 extension under --extend-terminal-stop: an ORF that runs off the end of its path extends the 3-prime-most exon downstream to the first in-frame stop (EVM extend_to_terminal). - --intron-score-cap <bp|auto>: caps the unmasked length contributing to one introns score (weight x min(len, cap)); auto selects smooth logarithmic scaling (linear up to 1 kb, then weight x 1 kb x (1 + ln(len/1 kb))) for species-agnostic anti-fusion control. On the same inputs and screening, screened + gffread -J validated proteins score BUSCO C 85.6% -> 94.7% (embryophyta_odb10), vs EVM 93.9%, with all retained proteins structurally clean.
Open
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.
Depends on / includes #5 (first commit here is that PR; easiest reviewed commit-by-commit).
While validating combinr on a 2.5 Gb, 80%-repeat plant genome we needed three
opt-in behaviours to reach EVM-level output. All are off by default —
with them off, output is unchanged (golden tests untouched).
--require-atg-start— prediction exons typed initial/single whosestart codon is not ATG are demoted so they cannot begin a gene (EVM's
canonical START gate, limited to the start codon only). At resolution, a
CDS not starting with ATG is trimmed in-frame to the first ATG (≤ 120
codons); when the path holds no in-frame ATG at all (truncated evidence
start), the 5'-most exon is extended upstream (cap 5 kb) to the first ATG
bounded by the nearest upstream in-frame stop, with the 5'UTR trimmed to
just past that stop. Rationale: homology/transferred predictions (weight
10 in our weights) frequently carry 5' truncations and otherwise outrank
ATG-starting ab-initio starts (weight 2). Before: 2,218/22,484 screened
proteins lacked an ATG start and were all discarded by
gffread -J(BUSCO C dropped 91.5% → 85.6%). After: zero such proteins, and ~1,400
evidence-rich genes with clipped starts were recovered as complete.
--extend-terminal-stop) — anORF that runs off the end of its path extends the 3'-most exon downstream
to the first in-frame stop (cap
MAX_TERMINAL_EXTEND), EVM'sextend_to_terminalhalf that combinr had not ported. 43/50 of ourscreened-out BUSCO loci were exactly this case.
--intron-score-cap <bp|auto>— caps the unmasked length contributingto one intron's score (
weight × min(len, cap));autoselects smoothlogarithmic scaling (
weight × lenup to 1 kb, thenweight × 1 kb × (1 + ln(len/1 kb))), which needs no species-specifictuning. Uncapped, a 15-38 kb intron from a misassembled transcript
outweighs any intergenic signal and the trellis absorbs the neighbouring
real gene (observed repeatedly; e.g. a 15.4 kb intron bridging into the
BUSCO gene at Chr01:188,523,992 from an upstream transcript artefact).
A 1000/2000/3000/5000/auto sweep gave 92.9/93.9/93.7/93.7/93.9 (BUSCO C);
automatched the best hard cap without tuning. Vertebrates withgenuinely long introns (10-100 kb) should prefer
autoover a low hardcap, since a hard cap can let strong intergenic signals split real
long-intron genes.
Net result — same inputs, same screening,
gffread -Jvalidated proteinsets, BUSCO embryophyta_odb10 (n=1614):
-JAll retained proteins are structurally clean (ATG start, terminal stop, no
in-frame stops). Full
cargo testpasses, including new tests for eachbehaviour.