mva-hackathon-2026

Allele-resolved rescue screening for BUB1B-associated MVA1

This repository contains the public, reproducible submission package for the 2026 Rare Disease, Real Kid: MVA Hackathon. It has two linked outputs:

1. A Track 1 prediction of a compound-heterozygous BUB1B pair.
2. A Track 2 proposal that tests upstream rescue of each allele before testing downstream protection from aneuploidy-induced stress.

The result is a research hypothesis, not a cure or a treatment recommendation. No medicine should be given on the basis of this repository.

Headline result

The alleles are 10,911 bp apart. Short reads do not establish phase and parental data were not supplied, so in trans is inferred, not proven. The second allele remains a VUS in isolation until segregation or functional evidence is available.

The ready-to-upload Track 1 file is results/MarxistLeninist_bub1b_compound_het.csv.

Track 2 concept

The proposal is an experimentally gated, three-axis screen:

• Read through the nonsense allele: azithromycin is the exposure-bridged first screen, with gentamicin as a stronger but substantially more toxic mechanistic comparator. The exact BUB1B context is UGA-A; it must be tested directly because NMD, the +4 base, and the identity of the inserted amino acid may sharply limit functional rescue.
• Stabilize the missense allele: arimoclomol is FDA-approved for NPC and amplifies a stress-induced heat-shock response. This is an indirect bridge, not evidence that it rescues p.Asn1002Lys.
• Test an orthogonal abundance pathway: prescription nicotinic acid is an approved NAD precursor, while NMN/SIRT2 increased BUBR1 abundance in mice. Niacin is not NMN, so this remains a biomarker-gated comparator.
• Buffer downstream proteotoxic stress: sirolimus (rapamycin) is the only approved candidate in the shortlist with drug-level rescue in a 2026 fly model of SAC-loss MVA-like microcephaly. It does not repair chromosome segregation and carries major immunosuppression and cancer-risk caveats.

Candidates advance only if they restore BUBR1 abundance or checkpoint function, reduce new chromosome-segregation errors, work near approved human exposure, and do not preferentially preserve premalignant aneuploid cells.

Reproduce public checks

Python 3.11+ is sufficient; no patient data or third-party package is required.

make check

python scripts/readthrough_context.py \
  --wild-type-codon TTA --codon-position 2 --alternate G --plus-four A

python scripts/structure_context.py AF-O60566-F1-model_v6.pdb --residue 1002

The last command is optional and requires the public AlphaFold model linked in the Track 2 report. Structural output is explicitly hypothesis-generating.

Repository map

• reports/MarxistLeninist_track1_report.md - variant analysis and limitations
• reports/MarxistLeninist_track2_report.md - drug rationale and validation plan
• reports/pitch_script.md - approximately three-minute narration
• SUBMISSION_CHECKLIST.md - exact form fields and action-time safety gates
• results/ - submission CSV and public evidence/provenance records
• scripts/ - submission validator, privacy gate, and reproducible context checks
• tests/ - synthetic/unit tests only
• DATA_GOVERNANCE.md - release boundary and privacy safeguards

Data boundary

The gated genome, phenotype document, read evidence, sample-wide annotations, and intermediate files are deliberately absent. The repository publishes only the minimum derived facts needed for the competition. Run make privacy before every public commit.

Acknowledgement

This work was made possible through the Hackathon, organized by Sage Bionetworks in partnership with the MVA Society, Hugging Face, and BEACON (The Benchmarking, Evaluation, and Assessment Consortium for Science), with prize sponsorship from AWS and Anthropic. We are deeply grateful to the child and their family who generously contributed their data and their story to advance research into this rare disease. We acknowledge their trust in making this Hackathon possible.

Released under CC BY 4.0. The underlying gated dataset has separate, more restrictive terms and is not redistributed here.