Evaluate PEPlife2-associated half-life models and serum-specific validation data before integration

Open
  • Is something wrong?
#286 1 comment 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
30/100
Issue type
Feature
Clarity
Needs clarification
Activity status
Active
Tech stack
python
Domain
bioinformatics, data, machine-learning

Research direction

Start by reviewing #279 and #283 alongside the PEPlife2 thesis and author-maintained resource. Establish which model artifacts, code, licenses, splits, and serum records are available, then audit endpoints, units, chemical forms, provenance, repeated measurements, and overlap with the named models. Done means a reproducibility and generalization assessment with a documented decision on whether an optional local adapter is justified, including offline tests if integration proceeds.

Written by the indexing model from the issue text.

Description

Request and priority

Track the exploratory models associated with Alam's PEPlife2 work as a lower-priority evaluation under #279. The data resource may be valuable even if its published research models are not suitable for integration.

Primary thesis record (2025): https://repository.iiitd.edu.in/xmlui/handle/123456789/1837
Author-maintained resource: https://webs.iiitd.edu.in/raghava/peplife2/
The resource now cites a 2026 database publication: https://doi.org/10.3390/immuno6020026

Evidence and limitations

The thesis reports a K-nearest-neighbor half-life model for modified peptides and an XGBoost model for natural peptides. Reported validation R-squared is 0.83 for modified peptides but only 0.29 for natural peptides; these are not evidence of adequate accuracy for unmodified vaccine peptides.

The resource spans biological fluids, experimental settings, peptides and proteins. Database entries are not necessarily unique peptides or independent serum measurements. A published database is not proof that trained inference models are downloadable.

Do not conflate these thesis models with original PlifePred or the newer PlifePred2 package in #283. Their lineage, target definitions and data overlap require verification.

Acceptance

  • Establish which model artifacts, code, licenses and reproducible splits are actually available; identify any relationship to #283 before duplicating an adapter.
  • Audit records for matrix/species, intact-peptide half-life versus other pharmacokinetic readouts, units, censoring, chemical forms and repeated measurements.
  • Determine whether a usable human-serum subset exists, preserving source-paper and assay provenance. Keep proteins, plasma, intestine and in-vivo PK records distinct rather than pooling labels.
  • Check overlap with training data used by PeptiVerse, Cavaco, PlifePred and Tan models before designating records as independent validation.
  • Reproduce model results if artifacts permit and test family-disjoint, assay-aware generalization. Report natural-peptide limitations explicitly.
  • Integrate a predictor only if licensing, reproducibility and relevant-domain performance justify it. Otherwise conclude this branch as unsuitable for prediction while documenting any useful validation-data contribution.
  • Any eventual adapter must be optional/local, preserve endpoint and input identity, and include offline tests without uploading patient data.

This is an evaluation/data-validation ticket, not an endorsed serum predictor, an automatic retraining plan or sequence-optimization feature.

Dominant language
Python
Stars
109
Forks
27
Avg merge
4h 53m
Merged PRs (30d)
59

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from openvax/mhctools

All issues in openvax/mhctools

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.