Conversation
Outline for mhcflurry 3.0: switch to mhcseqs for allele sequences, curate updated training data, retrain models with broader allele coverage.
The original plan was authored before any of the 2.3.0 perf / recipe
/ infra work landed. Rewrite to reflect the actual foundation 3.0
will start from, and resolve the gaps surfaced in PR review.
Major changes:
* New "Foundation: 2.3.0 baseline" section enumerating what 3.0
inherits as starting infrastructure (layered SHM, recipe
tightening, --gpu-batched calibrate, calibrate filter, unified
pipeline, compare_runs harness). 3.0 does not re-derive these.
* Allele-sequences section now quantifies the size implications
of the three representation options (pseudosequence 34 aa /
contact subset ~50 aa / full groove ~180 aa) — input tensor
growth, per-fit memory growth (~5x for full groove), wall-time
growth. Recommended default: contact subset; ablate vs full
groove rather than picking blind.
* New "Benchmark plan" section with quantitative thresholds per
slice (existing alleles vs new human alleles vs non-classical),
referencing the compare_runs.py / compare_new_vs_public.py
tooling that landed in 2.3.0.
* New "Compatibility / migration" section covering bundle
incompatibility, prediction-delta documentation, deprecation
window for 2.x bundles, and mhcgnomes alias audit.
* Decisions:
- H2-D*q fix lands in 2.3.0 patch (cheap, isolated, irrelevant
after 3.0's mhcseqs replaces the source). Removed the
"Consider..." ambiguity.
- Class II is explicitly out of scope; reframed for a hypothetical
4.0 track.
- Sequencing diagram showing 3.0 branches from tagged 2.3.0;
3.0 cannot start full training until 2.3.0 is tagged.
* Cleaned up dependency table to reference 2.3.0 (the actual
foundation) rather than 2.2.0.
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.
MHCflurry 3.0
Tracking branch for the next major release. See
V3_PLAN.mdfor the full plan.Key changes
Allele coverage (mhcseqs vs current)
Dependencies