Spaces:

SageBio
/

genomic-discovery-hackathon

Running

App Files Files Community

genomic-discovery-hackathon / index.html

gandreoletti

Update index.html

62be751 verified 9 days ago

Raw

History Blame Contribute Delete

46.8 kB

	<!doctype html>
	<html lang="en">
	<head>
	<meta charset="utf-8" />
	<meta name="viewport" content="width=device-width, initial-scale=1" />
	<title>Rare Disease, Real Kid: The MVA Hackathon 2026</title>
	<link rel="stylesheet" href="style.css" />
	</head>
	<body>
	<header class="hero">
	<div class="hero-inner">
	<div class="hero-partner">
	<span>Organized in partnership with</span>
	<img src="sage_logo.svg" alt="Sage Bionetworks" class="hero-logo" />
	</div>
	<div class="badge"> Rare Disease Challenge</div>
	<h1>Rare Disease, Real Kid: The MVA Hackathon 2026</h1>
	<p class="subtitle">This isn't a synthetic benchmark. It's the real genome and clinical story of a young child living with Mosaic Variegated Aneuploidy (MVA), an ultra-rare disease so poorly understood that there is still no treatment for it, only symptom management. The family is opening up the case to researchers everywhere, hoping to find the answer that's been out of reach until now. This challenge invites researchers, data scientists, citizen scientists, and clinicians everywhere</p>
	<div class="hero-actions">
	<a class="btn btn-primary" href="https://huggingface.co/datasets/SageBio/genomic-discovery-hackathon-data" target="_blank" rel="noopener">View Dataset</a>
	<a class="btn btn-outline" href="#tracks">Explore Tracks</a>
	</div>
	<div class="sponsor-row">
	<span class="sponsor-row-label">Made possible by</span>
	<div class="sponsor-row-logos">
	<img src="sage_logo.svg" alt="Sage Bionetworks" class="sponsor-logo" />
	<img src="aws-logo.png" alt="AWS — prize pool sponsor" class="sponsor-logo" />
	<img src="mva-society-logo.jpg" alt="MVA Society" class="sponsor-logo" />
	</div>
	</div>
	</div>
	</header>

	<!-- Tab nav -->
	<nav class="tab-nav" role="tablist">
	<button class="tab-btn active" role="tab" aria-selected="true" data-tab="overview">Overview</button>
	<button class="tab-btn" role="tab" aria-selected="false" data-tab="leaderboard">Leaderboard</button>
	<button class="tab-btn" role="tab" aria-selected="false" data-tab="submit">How to Submit</button>
	<button class="tab-btn" role="tab" aria-selected="false" data-tab="faq">FAQ</button>
	<button class="tab-btn" role="tab" aria-selected="false" data-tab="rules">Rules</button>
	<a class="tab-btn tab-btn-external" href="https://huggingface.co/datasets/SageBio/genomic-discovery-hackathon-data/discussions" target="_blank" rel="noopener">Community ↗</a>
	</nav>

	<main>
	<!-- ═══ OVERVIEW TAB ═══ -->
	<div id="tab-overview" class="tab-panel active">
	<div class="overview-layout">

	<div class="overview-content">

	<!-- What is this challenge -->
	<section id="overview-about" class="section">
	<div class="container">
	<h2>What is this challenge, exactly?</h2>
	<p>Every person's DNA contains millions of tiny variations, and almost all of them are harmless. But every so often, one of those variations disrupts something important, and the result is a disease. For many rare diseases, doctors and researchers already know a lot about the symptoms a condition causes, but they don't yet know exactly which genetic variation is responsible. Finding that answer is often the first step toward a real diagnosis, and eventually, a treatment.</p>
	<p>This is where you come in. We're releasing real, de-identified genetic and clinical data, shared with full, explicit permission from the family it belongs to, and asking the global research community to help make sense of it. You don't need to be a geneticist to contribute meaningfully: this challenge welcomes data scientists, machine learning engineers, clinicians, students, and citizen scientists alike. If you can work with data and think carefully about a problem, you can help here.</p>
	<p>The work is split into two tasks, described in plain terms below.</p>
	</div>
	</section>

	<!-- Why this challenge exists -->
	<section id="overview-story" class="section section-alt">
	<div class="container story-container">
	<div class="story-text">
	<h2>Why This Challenge Exists</h2>
	<p>This challenge exists because one family made a hard decision: to share their child's most private information, their genome, with strangers around the world, in the hope that someone, somewhere, might find an answer.</p>
	<p>Their kid lives with Mosaic Variegated Aneuploidy (MVA), a disease few clinicians will ever encounter in their careers. It's poorly understood, there is no disease-modifying treatment, and care today means managing symptoms and watching closely for cancer.</p>
	<p>When you work on this challenge, you're not solving a hypothetical or a synthetic benchmark. You're working alongside clinicians, patient advocates, and a family who are trusting the global research community with what matters to them most. That's the whole point of doing this in the open.</p>
	</div>
	<figure class="story-figure">
	<img src="child-drawing.png" alt="A child's drawing of himself smiling in a hospital bed" class="story-drawing" />
	<figcaption>Illustration courtesy of the <a href="https://mvasociety.org" target="_blank" rel="noopener">MVA Society</a></figcaption>
	</figure>
	</div>
	</section>

	<!-- Tracks -->
	<section id="tracks" class="section section-alt">
	<div class="container">
	<h2>Two Tracks</h2>
	<div class="card-grid">
	<div class="info-card track-card">
	<div class="info-icon">🧬</div>
	<h3>Track 1 — Variant Prediction</h3>
	<p><strong>In plain terms:</strong> you'll be given a patient's genomic data consisting of raw sequencing reads (FASTQ) and called variants (VCF), along with a description of their symptoms, and asked to identify which specific variant(s) is actually causing their condition. Secondary or incidental findings are welcome too — if something else in the data looks biologically interesting or clinically relevant, flag it. Think of it like being handed a long list of suspects and a description of a crime, and asked to identify who's actually responsible, using clues from biology instead of an alibi — but if you spot something else worth noting along the way, that counts too.</p>
	<p>Your ranked guesses are automatically checked against the answer researchers have already confirmed through clinical testing, and your score updates on a <strong>live public leaderboard</strong> — so you can see immediately how well your approach worked and try to improve it.</p>
	<p class="track-format">What you submit: a CSV file of ranked variant predictions with an estimated probability of causality, and a detailed description of your method(s). Full definitions and examples are on the "How to Submit" tab.</p>
	</div>
	<div class="info-card track-card">
	<div class="info-icon">💊</div>
	<h3>Track 2 — Drug Repositioning</h3>
	<p><strong>In plain terms:</strong> once you know (or have a strong guess at) what's biologically going wrong, the next question is: does a medication that already exists (approved for some other condition entirely) happen to work on this problem too? This is a real, established strategy in medicine: it's often far faster and cheaper to repurpose a drug that's already proven safe than to invent a brand-new one from scratch.</p>
	<p>For this track, you'll characterize the mechanism behind the variant(s) you're working from and write up your reasoning and proposed candidate medication(s). There's no automatic scoring here but instead, a panel of independent expert judges (researchers, clinicians, and patient advocates) will read every submission and evaluate it on its scientific merit and real-world potential to help.</p>
	<p class="track-format">What you submit: a GitHub repository of your code, a written detailed description of your method (PDF or Markdown) including your variant mechanism characterization and proposed candidate medication(s), and a 3-minute pitch video.</p>
	</div>
	</div>
	</div>
	</section>


	<!-- Partners & Sponsors -->
	<section id="partners" class="section">
	<div class="container">
	<h2>Our Partners & Sponsors</h2>
	<div class="card-grid">
	<div class="info-card partner-card">
	<img src="sage_logo.svg" alt="Sage Bionetworks" class="partner-logo" />
	<h3>Sage Bionetworks</h3>
	<p>A nonprofit biomedical research organization organizing this challenge and hosting the data infrastructure, in partnership with the family at its center.</p>
	</div>
	<div class="info-card partner-card">
	<img src="mva-society-logo.jpg" alt="MVA Society" class="partner-logo" />
	<h3>MVA Society</h3>
	<p>A patient advocacy organization supporting families affected by Mosaic Variegated Aneuploidy, helping ground this challenge in what actually matters to patients and families.</p>
	</div>
	<div class="info-card partner-card">
	<img src="aws-logo.png" alt="AWS" class="partner-logo" />
	<h3>AWS</h3>
	<p>Amazon Web Services sponsors the $25,000 prize pool for this challenge, supporting open, crowdsourced rare disease research.</p>
	</div>
	</div>
	</div>
	</section>

	<!-- Prizes -->
	<section id="prizes" class="section section-alt">
	<div class="container">
	<h2>Prizes</h2>
	<p><strong>$25,000 total prize pool</strong> — AWS-sponsored.</p>
	<div class="prize-grid">
	<div class="prize-card gold">
	<div class="prize-place">1st</div>
	<div class="prize-amount">$12,000</div>
	</div>
	<div class="prize-card silver">
	<div class="prize-place">2nd</div>
	<div class="prize-amount">$7,000</div>
	</div>
	<div class="prize-card bronze">
	<div class="prize-place">3rd</div>
	<div class="prize-amount">$4,000</div>
	</div>
	<div class="prize-card special">
	<div class="prize-place">Innovation / Community</div>
	<div class="prize-amount">$2,000</div>
	</div>
	</div>
	</div>
	</section>

	<!-- Judging -->
	<section id="judging" class="section">
	<div class="container">
	<h2>Judging Criteria</h2>

	<h3 class="sub-heading">Track 1 — Automated Scoring</h3>
	<p>Because this challenge centers on one patient with a clinically confirmed diagnosis, we score submissions the way large-scale rare disease benchmarks score their solved cases. Two scores are computed automatically the moment you submit:</p>
	<div class="card-grid">
	<div class="info-card">
	<h3>Rank points</h3>
	<p>How high the true variant(s) landed in your ranked list. Landing your #1 guess exactly right scores highest; landing it in your top 10 still earns partial credit. If you correctly identify only one of the two variants in a compound-heterozygous pair, you get half credit.</p>
	</div>
	<div class="info-card">
	<h3>F-max</h3>
	<p>The best balance of precision and recall your submission achieves at any confidence threshold, rewarding submissions that pinpoint the right answer without burying it in noise.</p>
	</div>
	</div>
	<p style="margin-top: 1rem;">Secondary or incidental findings won't hurt your automated score; they're simply set aside for the judging panel to review qualitatively, the same way they'd be handled in a real clinical genetics workflow.</p>

	<h3 class="sub-heading" style="margin-top: 2.5rem;">Track 2 — Panel Judging</h3>
	<p>Track 2 submissions are evaluated by an independent expert panel across four weighted dimensions over a ~2–3 month judging window. Here's what each one actually means:</p>
	<div class="criteria-stacked-bar" aria-hidden="true">
	<div class="csb-seg" style="--w:35%; --c:#1f9bea" data-tip="Scientific Rigor — 35%"></div>
	<div class="csb-seg" style="--w:25%; --c:#59c959" data-tip="Patient Impact — 25%"></div>
	<div class="csb-seg" style="--w:25%; --c:#7b5ea7" data-tip="Innovation — 25%"></div>
	<div class="csb-seg" style="--w:15%; --c:#e0509a" data-tip="Scalability — 15%"></div>
	</div>
	<div class="criteria-list">
	<div class="criterion">
	<div class="criterion-label">
	<span><span class="criterion-swatch" style="--c:#1f9bea"></span>Scientific Rigor</span><strong>35%</strong>
	</div>
	<p class="criterion-desc">Is the variant mechanism characterization sound, and is the proposed drug repositioning candidate well-supported by that mechanism?</p>
	</div>
	<div class="criterion">
	<div class="criterion-label">
	<span><span class="criterion-swatch" style="--c:#59c959"></span>Patient Impact</span><strong>25%</strong>
	</div>
	<p class="criterion-desc">If this were used in the real world, how much difference could it make to this patient's — or similar patients' — diagnosis, treatment, or quality of life?</p>
	</div>
	<div class="criterion">
	<div class="criterion-label">
	<span><span class="criterion-swatch" style="--c:#7b5ea7"></span>Innovation</span><strong>25%</strong>
	</div>
	<p class="criterion-desc">Did you bring a genuinely creative angle, method, or tool to the problem, rather than a routine or purely off-the-shelf approach?</p>
	</div>
	<div class="criterion">
	<div class="criterion-label">
	<span><span class="criterion-swatch" style="--c:#e0509a"></span>Scalability</span><strong>15%</strong>
	</div>
	<p class="criterion-desc">Could your approach realistically be applied to other patients, other diseases, or larger datasets — not just this one case?</p>
	</div>
	</div>
	</div>
	</section>

	<!-- How to participate -->
	<section id="how-to-participate" class="section section-alt">
	<div class="container">
	<h2>How to Participate</h2>
	<p>No prior hackathon experience needed! Here's the whole process from start to finish.</p>
	<ol class="steps">
	<li>
	<span class="step-num">1</span>
	<div>
	<strong>Request dataset access</strong>
	<p>Because this data comes from a real patient, access isn't automatic. You'll fill out a short form on the <a href="https://huggingface.co/datasets/SageBio/genomic-discovery-hackathon-data" target="_blank" rel="noopener">dataset page</a> confirming you understand and agree to the usage rules (things like not sharing the data further and deleting it once the challenge ends). A real person reviews every request, so approval isn't instant — plan to request access a few days before you want to start working.</p>
	</div>
	</li>
	<li>
	<span class="step-num">2</span>
	<div>
	<strong>Choose your track</strong>
	<p>Decide whether you want to try identifying the causal variant (Track 1), propose a repurposed medication (Track 2), or take on both — many teams do. You don't need to finish one to start the other.</p>
	</div>
	</li>
	<li>
	<span class="step-num">3</span>
	<div>
	<strong>Build & submit</strong>
	<p>Work through the data at your own pace, using whatever tools, software, or methods you prefer. When you're ready, submit your results: a GitHub repository of your code, a 3-minute pitch video, plus a CSV file (Track 1) or written report (Track 2). Full instructions with examples are on the "How to Submit" tab.</p>
	</div>
	</li>
	</ol>
	</div>
	</section>

	<!-- CTA -->
	<section class="section">
	<div class="container cta-block">
	<h2>Ready to compete?</h2>
	<p>Questions and discussion are welcome in the <a href="https://huggingface.co/datasets/SageBio/genomic-discovery-hackathon-data/discussions" target="_blank" rel="noopener">HF Discussions page</a>.</p>
	<div class="hero-actions">
	<a class="btn btn-primary" href="https://huggingface.co/datasets/SageBio/genomic-discovery-hackathon-data" target="_blank" rel="noopener">View Dataset →</a>
	<a class="btn btn-outline-dark" href="#" data-tab-target="submit">How to Submit →</a>
	</div>
	</div>
	</section>

	</div><!-- /overview-content -->

	<!-- Sticky sidebar TOC -->
	<nav class="overview-toc" aria-label="On this page">
	<span class="toc-label">On this page</span>
	<a class="toc-link" href="#overview-about">About</a>
	<a class="toc-link" href="#tracks">Two Tracks</a>
	<a class="toc-link" href="#partners">Partners & Sponsors</a>
	<a class="toc-link" href="#prizes">Prizes</a>
	<a class="toc-link" href="#judging">Judging Criteria</a>
	<a class="toc-link" href="#how-to-participate">How to Participate</a>
	<a class="toc-link toc-top" href="#">↑ Back to top</a>
	</nav>

	</div><!-- /overview-layout -->
	</div><!-- /tab-overview -->

	<!-- ═══ LEADERBOARD TAB ═══ -->
	<div id="tab-leaderboard" class="tab-panel">
	<section class="section">
	<div class="container">
	<h2>Track 1 Leaderboard</h2>
	<p>Rankings update automatically each time a valid submission is evaluated. Only the best-scoring submission per team is shown.</p>
	<div class="leaderboard-meta">
	<span>Metric: <strong>Rank points & F-max</strong> (simialr to Stenton et al., 2024)</span>
	<span class="lb-note">Track 2 rankings are not shown here as those are reviewed privately by judges.</span>
	</div>
	<div class="lb-placeholder">
	<div class="lb-icon">🏁</div>
	<p>The leaderboard will go live once submissions open.<br>Check back after the challenge launch.</p>
	</div>
	</div>
	</section>
	</div><!-- /tab-leaderboard -->

	<!-- ═══ SUBMIT TAB ═══ -->
	<div id="tab-submit" class="tab-panel">
	<section class="section">
	<div class="container">
	<h2>How to Submit</h2>

	<h3 class="sub-heading">Track 1 — Variant Prediction</h3>
	<ol class="steps">
	<li>
	<span class="step-num">1</span>
	<div>
	<strong>Prepare your file</strong>
	<p>Create a CSV file with one row per proposed causal variant, or proposed compound-heterozygous variant pair. Rank your predictions by your own estimated probability of causal relationship (EPCR) — a number between 0 and 1. An EPCR of 1 means you're fully confident this is the causal variant. An EPCR close to 0 means you're including it as a long-shot or low-confidence guess rather than a serious candidate — there's no fixed "confident enough" cutoff, only your ranking relative to your own other guesses matters for scoring. Coordinates must use GRCh38. Up to 10 candidate rows are accepted — this is one patient, not a cohort, so we're asking for your best-ranked guesses, not an exhaustive list. Flag whether each row is your primary candidate or a secondary/incidental finding using the finding_type column. Secondary or incidental findings won't hurt your automated score — include them if you want them considered.</p>
	<a class="download-template" href="track1_submission_template.csv" download>⬇ Download submission template (.csv)</a>
	<table class="schema-table">
	<thead><tr><th>Field</th><th>Type</th><th>Description</th></tr></thead>
	<tbody>
	<tr><td><code>proband_id</code></td><td>string</td><td>Patient identifier (provided in the dataset)</td></tr>
	<tr><td><code>chrom_1</code></td><td>string</td><td>Chromosome of the first (or only) variant (e.g. <code>chr15</code>)</td></tr>
	<tr><td><code>pos_1</code></td><td>integer</td><td>Base-pair position of the first variant (GRCh38)</td></tr>
	<tr><td><code>ref_1</code></td><td>string</td><td>Reference allele of the first variant</td></tr>
	<tr><td><code>alt_1</code></td><td>string</td><td>Alternate allele of the first variant</td></tr>
	<tr><td><code>chrom_2</code></td><td>string</td><td>Chromosome of the second variant, if proposing a compound-heterozygous pair (leave blank otherwise)</td></tr>
	<tr><td><code>pos_2</code></td><td>integer</td><td>Base-pair position of the second variant (leave blank if not applicable)</td></tr>
	<tr><td><code>ref_2</code></td><td>string</td><td>Reference allele of the second variant (leave blank if not applicable)</td></tr>
	<tr><td><code>alt_2</code></td><td>string</td><td>Alternate allele of the second variant (leave blank if not applicable)</td></tr>
	<tr><td><code>epcr</code></td><td>float</td><td>Estimated probability of causal relationship, in the range (0, 1]</td></tr>
	<tr><td><code>finding_type</code></td><td>string</td><td>Either <code>primary</code> or <code>secondary</code> — flag whether you believe this variant explains the patient's known phenotype, or whether it's a secondary/incidental finding (something unrelated to the primary condition, but still worth surfacing).</td></tr> </tbody>
	</table>
	</div>
	</li>
	<li>
	<span class="step-num">2</span>
	<div>
	<strong>Write up your work</strong>
	<p>Push your documented, reproducible code to a public GitHub repository. Each model must include a detailed description of your method(s)</p>
	</div>
	</li>
	<li>
	<li>
	<span class="step-num">3</span>
	<div>
	<strong>Name your files</strong>
	<p>Use this naming convention so submissions are easy to track:</p>
	<table class="schema-table">
	<thead><tr><th>File</th><th>Naming pattern</th><th>Example</th></tr></thead>
	<tbody>
	<tr><td>Track 1 CSV</td><td><code><teamname>_track1_model<N>.csv</code></td><td><code>helixsquad_track1_model1.csv</code></td></tr>
	<tr><td>Methods description</td><td><code><teamname>_method<N>.txt</code></td><td><code>helixsquad_track1_method1.txt</code></td></tr>
	</tbody>
	</table>
	<p><code><N></code> is 1, 2,3,4,5, or 6 matching your Track 1 submission cap, so a team can submit up to 6 distinct models without filename collisions.</p>
	</div>
	</li>
	<li>
	<span class="step-num">4</span>
	<div>
	<strong>Sign in & upload</strong>
	<p>Sign in with your Hugging Face account on the leaderboard Space, then submit your CSV file along with your method using the submission form. Teams may submit up to 6 times; only the best-scoring submission counts toward final ranking.</p>
	</div>
	</li>
	<li>
	<span class="step-num">5</span>
	<div>
	<strong>See your score</strong>
	<p>Your CSV submission is evaluated automatically using rank points and F-max against the clinically confirmed answer on the leaderboard. Your method(s) will be reviewed separately as part of judging criteria like innovation and scalability.</p>
	</div>
	</li>
	</ol>

	<h3 class="sub-heading" style="margin-top: 2.5rem">Track 2 — Drug Repositioning</h3>
	<ol class="steps">
	<li>
	<span class="step-num">1</span>
	<div>
	<strong>Write your report</strong>
	<p>Prepare a written report (PDF or Markdown) proposing repositioned drug candidates supported by your analysis. Your report should include a characterization of the variant's mechanism — e.g., loss-of-function or gain-of-function, the pathway disrupted, and the downstream biological consequence , as the basis for your repositioning rationale.</p>
	</div>
	</li>
	<li>
	<span class="step-num">2</span>
	<div>
	<strong>Package your work</strong>
	<p>Push your documented, reproducible code to a public GitHub repository, and record a 3-minute pitch video walking through your reasoning and proposed candidates.</p>
	</div>
	</li>
	<li>
	<li>
	<span class="step-num">3</span>
	<div>
	<strong>Name your files</strong>
	<p>Use this naming convention so submissions are easy to track:</p>
	<table class="schema-table">
	<thead><tr><th>File</th><th>Naming pattern</th><th>Example</th></tr></thead>
	<tbody>
	<tr><td>Track 2 report</td><td><code><teamname>_track2_report.pdf</code> (or <code>.md</code>)</td><td><code>helixsquad_track2_report.pdf</code></td></tr>
	<tr><td>Pitch video</td><td><code><teamname>_video.mp4 (or .mov)</code></td><td><code>helixsquad_pitch_video.mp4</code></td></tr>
	</tbody>
	</table>
	<p> Track 2 has no model number since only one final submission is accepted per team.</p>
	</div>
	</li>
	<li>
	<span class="step-num">4</span>
	<div>
	<strong>Submit before the deadline</strong>
	<p>Upload your report along with your GitHub repo link and video via the submission form. Late submissions cannot be accepted, as materials go directly to human judges for scheduling.</p>
	</div>
	</li>
	<li>
	<span class="step-num">5</span>
	<div>
	<strong>Await judging</strong>
	<p>The independent panel reviews all submissions over a ~2–3 month window. Results are announced after the judging period closes.</p>
	</div>
	</li>
	<h3 class="sub-heading" style="margin-top: 2.5rem">Know what to include in your methods write-up</h3>
	<div class="methods-note">
	<p>Not sure how much detail to include, or what kind of questions your write-up should answer? Examples of what we're looking for in the methods section such as model description, data sources, how compound-heterozygous or secondary findings were handled, and so on can be found here:</p>
	<a class="download-template" href="mva_methods_description_form.xlsx" download>⬇ Methods description examples (.xlsx)</a>
	</div>
	</section>
	</div><!-- /tab-submit -->

	<!-- ═══ FAQ TAB ═══ -->
	<div id="tab-faq" class="tab-panel">
	<section class="section">
	<div class="container">
	<h2>FAQ</h2>
	<div class="faq-list">

	<details class="faq-item">
	<summary>Who can participate?</summary>
	<p>Anyone with a Hugging Face account can participate. Dataset access is gated and approved manually — request access on the <a href="https://huggingface.co/datasets/SageBio/genomic-discovery-hackathon-data" target="_blank" rel="noopener">dataset page</a>.</p>
	</details>

	<details class="faq-item">
	<summary>Can I participate in both tracks?</summary>
	<p>Yes. Track 1 and Track 2 are scored independently. You can submit to one or both.</p>
	</details>

	<details class="faq-item">
	<summary>How is Track 1 scored?</summary>
	<p>Submissions are scored the way large-scale rare disease benchmarks (like Stenton et al., 2024) score solved cases applied to this single, clinically confirmed answer. Two metrics are computed automatically: <strong>rank points</strong> (based on how high the true variant(s) land in your ranked list, with partial credit if you recover only one of two compound-heterozygous variants) and <strong>F-max</strong> (the best precision/recall balance across your submitted confidence thresholds).</p>
	</details>

	<details class="faq-item">
	<summary>Are secondary or incidental findings scored?</summary>
	<p> They won't hurt your automated Track 1 score — include them as additional rows if you'd like. They're set aside for qualitative review by the judging panel rather than folded into the automated rank points/F-max calculation, since there's no fixed "correct" answer for secondary findings the way there is for the primary causal variant.</p>
	<p> Use the optional notes column to briefly explain why you're flagging it (e.g., "well-established pathogenic variant, unrelated to primary phenotype, recommend clinical follow-up")</p>
	</details>

	<details class="faq-item">
	<summary>How many submissions can I make?</summary>
	<p>Track 1 allows up to 6 submissions per team; only your best-scoring submission is shown on the leaderboard. Track 2 accepts one final report per team. No resubmissions, so make it count.</p>
	</details>

	<details class="faq-item">
	<summary>What compute resources are available?</summary>
	<p>The challenge Space runs on CPU-basic hardware. You are welcome to use your own compute for training — only the final submission file needs to be uploaded here.</p>
	</details>

	<details class="faq-item">
	<summary>What is the data license?</summary>
	<p>The underlying dataset is <strong>gated</strong> and subject to the terms of the Hackathon Rules and Data Transfer Agreement — access requires approval, redistribution is prohibited, and all data must be deleted upon conclusion of the Hackathon. Participant <em>submissions and results</em> (predictions, code, reports) are released under <strong>CC BY 4.0</strong> and may be reused with attribution.</p>
	</details>

	<details class="faq-item">
	<summary>When will winners be announced?</summary>
	<p>Results will be announced after the ~2–3 month judging window closes. The exact date will be posted in the <a href="https://huggingface.co/datasets/SageBio/genomic-discovery-hackathon-data/discussions" target="_blank" rel="noopener">HF Discussions page</a>.</p>
	</details>

	<details class="faq-item">
	<summary>I have a question not answered here.</summary>
	<p>Post in the <a href="https://huggingface.co/datasets/SageBio/genomic-discovery-hackathon-data/discussions" target="_blank" rel="noopener">HF Discussions page</a>. Organizers monitor it regularly.</p>
	</details>

	</div>
	</div>
	</section>
	</div><!-- /tab-faq -->

	<!-- ═══ RULES TAB ═══ -->
	<div id="tab-rules" class="tab-panel">
	<section class="section">
	<div class="container">
	<h2>Challenge Rules</h2>
	<p>To participate in this Hackathon, you must review and accept the following terms. Please read carefully and proceed only if you agree to comply with the conditions below.</p>

	<h3 class="sub-heading">Terms of Service & Eligibility</h3>
	<ul class="rules-list">
	<li>You must abide by the Hugging Face Terms of Service and by these Hackathon Rules.</li>
	<li>You must be 18 years of age or older to participate.</li>
	<li>Each member of your Hackathon team must individually register as a participant and agree to these Hackathon Rules.</li>
	</ul>

	<h3 class="sub-heading">Data Privacy & Recontact Restrictions</h3>
	<ul class="rules-list">
	<li>You agree that you will not attempt to recontact the data subject, data subject family members, or any points of contact at the MVA Society.</li>
	<li>Your use of the data will comply with all applicable laws, rules, regulations, and professional standards.</li>
	<li>If any unauthorized disclosure of the data occurs or is suspected, you agree to contact Sage Bionetworks' Privacy and Compliance Office via <a href="https://sagebionetworks.jira.com/servicedesk/customer/portal/20" target="_blank" rel="noopener">Sage's Help Center</a>.</li>
	<li>You will not release or otherwise grant data access to anyone, and you will establish appropriate safeguards to prevent unauthorized data use.</li>
	</ul>

	<h3 class="sub-heading">Hackathon Flow</h3>

	<p><strong>Registration.</strong> Participants must register with Hugging Face and request access to the data, including a valid email address and attestation to accepting these Hackathon Rules.</p>

	<p><strong>Hackathon tasks.</strong> The Hackathon consists of two tasks:</p>
	<ol class="rules-list">
	<li><strong>Task 1 — Variant identification:</strong> identify the disease-causing variant(s) using genetic variant information and phenotypic presentations. Participants will not be given the NHS-validated disease-causal variants; you will develop computational models and submit a list of variants you believe are causal.</li>
	<li><strong>Task 2 — Drug repositioning:</strong> leveraging the genetic information and variant prediction from Task 1, identify existing market-approved medications whose mechanism of action may target the disease-causing pathway(s) or affect symptoms and improve quality of life.</li>
	</ol>

	<p><strong>Team formation.</strong></p>
	<ul class="rules-list">
	<li>Teams are not required.</li>
	<li>Participants may form teams based on skills and interests; teams often combine programmers, designers, and domain experts.</li>
	</ul>

	<p><strong>Data & resource provision.</strong></p>
	<ul class="rules-list">
	<li><strong>File types:</strong> de-identified genomic data in VCF format (SNPs, indels, structural variants); raw sequencing data optionally in BAM/CRAM. Phenotypic data as standardized HPO terms (e.g., HP:0001250 Seizure, HP:0002904 Hyperbilirubinemia).</li>
	<li><strong>Data size:</strong> ~120 GB. A single-subject dataset (one patient) — small in scale, but genomically sensitive; size details are subject to change pending final file export.</li>
	<li><strong>Access:</strong> participants request access via Hugging Face's gated-dataset mechanism. Once approved, each participant downloads directly via their own Hugging Face account/token — no data may be emailed, shared via drive, or distributed through any channel outside the Hugging Face access-controlled dataset repo. You must not repost any data on Hugging Face.</li>
	<li><strong>Compute:</strong> participants are responsible for their own compute environment and any associated costs; Sage Bionetworks does not provide compute resources for the Hackathon.</li>
	</ul>

	<h3 class="sub-heading">Data Deletion</h3>
	<ul class="rules-list">
	<li>Data retention after the Hackathon is not permitted for any purpose.</li>
	<li>Participants must delete all copies of the Hackathon dataset from every environment where it was accessed or stored, including:
	<ul class="rules-list">
	<li>Local machines and personal hard drives (delete files and empty trash/recycle bin);</li>
	<li>Cloud compute instances and virtual machines (terminate instances and delete associated storage volumes);</li>
	<li>Notebooks and notebook environments (e.g., Jupyter, Google Colab — delete any cached or downloaded data files);</li>
	<li>Private code repositories (remove data files from repo history, not only from the working directory);</li>
	<li>Any intermediate files, derived datasets, or outputs that contain or could reconstruct the original data.</li>
	</ul>
	</li>
	<li>All data must be deleted within 30 days of Hackathon close. Do not wait until publications or code releases are finalized before deleting the dataset — only the challenge summary preprint and publicly shareable outputs (code, model weights, predictions) may be retained after this deadline.</li>
	<li>All participants must email <strong>[team]@synapse.org</strong> to confirm data has been deleted from personal compute environments. If you do not confirm, Sage Bionetworks may contact you directly after 30 days.</li>
	</ul>

	<h3 class="sub-heading">Presentations & Judging</h3>
	<ul class="rules-list">
	<li><strong>Format:</strong> each team's submission includes a written report, a GitHub repository (code and workflows, if applicable), and a 3-minute recorded pitch video covering results, methodology, data insights, and proposed solutions.</li>
	<li><strong>Judging panel:</strong> a panel of expert judges — including researchers, clinicians, and patient advocates — will review all submissions.</li>
	<li><strong>Timeline:</strong> judging takes place over approximately 2–3 months following submission close. Winners are announced at the conclusion of the judging period.</li>
	<li><strong>Format note:</strong> presentations are pre-recorded rather than live; there is no live Q&A round in the current design.</li>
	</ul>

	<h3 class="sub-heading">Evaluation & Scoring</h3>
	<p>Track 1 and Track 2 are scored independently, using different methods:</p>
	<p><strong>Track 1 — Variant Prediction (automated):</strong> scored automatically against the NHS-validated causal variant(s) using <strong>rank points</strong> and <strong>F-max</strong> (see the Judging Criteria section of the Overview tab for details). No panel review is involved in this score.</p>
	<p><strong>Track 2 — Drug Repositioning (panel-judged):</strong> reviewed by the expert judging panel across four weighted criteria:</p>
	<ul class="rules-list">
	<li>Scientific Rigor — <strong>35%</strong></li>
	<li>Patient Impact — <strong>25%</strong></li>
	<li>Innovation — <strong>25%</strong></li>
	<li>Scalability — <strong>15%</strong></li>
	</ul>
	<p>A valid submission is completion of <strong>either</strong> task — Track 1 and Track 2 each accept independent submissions, and teams are welcome to enter one or both. Each submission includes a documented report or CSV (as applicable), a GitHub repository (code/workflows), and a 3-minute recorded pitch video.</p>

	<h3 class="sub-heading">Prizes</h3>
	<ul class="rules-list">
	<li><strong>Total prize pool: $25,000</strong>, sponsored by AWS</li>
	<li>🥇 1st Place — $12,000</li>
	<li>🥈 2nd Place — $7,000</li>
	<li>🥉 3rd Place — $4,000</li>
	<li>🌟 Innovation / Community Award — $2,000, awarded at judges' discretion to a submission demonstrating exceptional creativity, community impact, or patient-centered design, independent of overall placement</li>
	</ul>

	<h3 class="sub-heading">Research Progress & Post-Hackathon Data Handling</h3>
	<ul class="rules-list">
	<li><strong>Publications:</strong> participants may publish and present on Hackathon results, subject to the embargo policy below.</li>
	<li><strong>Data:</strong> all personally identifiable information and raw genomic data must be deleted from participants' systems within 30 days of Hackathon close (see Data Deletion section above). Data hosted on Synapse will be sequestered and inaccessible to external users at that point; the Hugging Face–hosted copy will be removed per the Data Transfer Agreement's deletion policy. Any use of the data beyond the Hackathon's scope requires separate, explicit permission from the donors of the data — this restriction does not lapse over time.</li>
	<li><strong>Results and code:</strong> anonymized results — identified variants, statistical analyses, proposed solutions — remain available to the research community via Synapse and/or Hugging Face after the Hackathon ends. Submissions may be rerun by the organizers (Sage Bionetworks, Hugging Face, MVA Society) and are released under a CC-BY license, with participant names shared as part of open-access attribution.</li>
	<li><strong>Leaderboard/landing page visibility:</strong> the Hugging Face competition landing page and leaderboard may remain live after close for visibility and recognition, or may be removed by mutual agreement between Sage Bionetworks and Hugging Face.</li>
	<li><strong>Further development:</strong> participants are encouraged to continue developing their approaches using publicly available or synthetic datasets — not the original patient data, which must be deleted.</li>
	</ul>

	<h3 class="sub-heading">Embargo Policy</h3>
	<ul class="rules-list">
	<li>The embargo period begins at Hackathon close and ends upon public posting of the challenge summary report or preprint by the organizing team.</li>
	<li>During the embargo period, participants may not submit manuscripts for peer-reviewed publication that use the challenge dataset.</li>
	<li>Participants are free to publicly share their code, models, and derived outputs (e.g., features, predictions) at any time without restriction.</li>
	<li>Participants may present preliminary findings in conference abstracts, posters, or talks with prior written approval from the organizers.</li>
	<li>Once the challenge summary report or preprint is publicly posted, participants are free to submit independent publications without restriction, provided they follow the Citation and Acknowledgement requirements below.</li>
	</ul>

	<h3 class="sub-heading">Submission Reuse & Recontact Consent</h3>
	<ul class="rules-list">
	<li><strong>Submission reuse:</strong> by registering, participants acknowledge that submissions may be rerun by the Hackathon organizers (Sage Bionetworks, Hugging Face, and the MVA Society). Submissions are subject to a CC-BY license, and your name will be shared as part of the open-access attribution.</li>
	<li><strong>Recontact:</strong> by providing your email address and participating, you consent to being recontacted by the Hackathon organizers for purposes related to the Hackathon, including but not limited to follow-up communications, feedback requests, and potential future research opportunities.</li>
	</ul>

	<h3 class="sub-heading">Post-Hackathon Information Sharing</h3>
	<ul class="rules-list">
	<li>All submitted results will be accessible to the broader research community to foster collaboration and innovation.</li>
	<li>Leading submissions may be prominently featured on Hugging Face, Sage Bionetworks, or Synapse-based websites.</li>
	<li>Follow-up activities such as blog posts, video recaps, and social media highlights may showcase key submissions and behind-the-scenes moments.</li>
	<li>This open-access approach ensures the knowledge generated benefits the broader research community while inspiring future advances in rare disease research.</li>
	</ul>

	<h3 class="sub-heading">Citation & Acknowledgement</h3>
	<p>Any publication, preprint, conference abstract, or public communication arising from participation in this challenge must include the following acknowledgement:</p>
	<blockquote style="border-left: 3px solid #0d7fa5; padding: 0.75rem 1.25rem; margin: 1rem 0; background: #f0f8fe; font-style: italic; color: #3a4f60;">
	"This work was made possible through the Hackathon, organized by Sage Bionetworks in partnership with the Vall d'Hebron Institute of Oncology (VHIO), the MVA Society, and Hugging Face, with cloud computing support provided by AWS. We are deeply grateful to the patient and family who generously contributed their data and their story to advance research into this ultra-rare disease. We acknowledge their courage and trust in making this challenge possible."
	</blockquote>
	<ul class="rules-list">
	<li><strong>Data subject privacy:</strong> any publication referencing the underlying patient data must not include any information that could re-identify the data subject or their family, beyond what is already publicly available through the family's own blog posts and public communications.</li>
	<li><strong>Dataset citation:</strong> the Hackathon dataset must be cited using the reference provided on the challenge Synapse page at the time of publication.</li>
	</ul>
	</div>
	</section>
	</div><!-- /tab-rules -->

	</main>

	<footer>
	<div class="container">
	<p>The MVA Hackathon 2026 · Hosted on <a href="https://huggingface.co/spaces" target="_blank" rel="noopener">Hugging Face Spaces</a></p>
	<p class="footer-credit">Organized in partnership with <a href="https://sagebionetworks.org" target="_blank" rel="noopener">Sage Bionetworks</a>, the MVA Society, and AWS</p>
	</div>
	</footer>
	<script>
	document.querySelectorAll('.tab-btn').forEach(btn => {
	btn.addEventListener('click', () => {
	document.querySelectorAll('.tab-btn').forEach(b => { b.classList.remove('active'); b.setAttribute('aria-selected', 'false'); });
	document.querySelectorAll('.tab-panel').forEach(p => p.classList.remove('active'));
	btn.classList.add('active');
	btn.setAttribute('aria-selected', 'true');
	document.getElementById('tab-' + btn.dataset.tab).classList.add('active');
	});
	});

	// Smooth scroll for TOC links
	document.querySelectorAll('.overview-toc .toc-link').forEach(link => {
	link.addEventListener('click', e => {
	const href = link.getAttribute('href');
	if (href === '#') {
	e.preventDefault();
	window.scrollTo({ top: 0, behavior: 'smooth' });
	} else {
	const target = document.querySelector(href);
	if (target) {
	e.preventDefault();
	target.scrollIntoView({ behavior: 'smooth', block: 'start' });
	}
	}
	});
	});

	// "How to Submit" CTA button
	document.querySelectorAll('[data-tab-target]').forEach(btn => {
	btn.addEventListener('click', e => {
	e.preventDefault();
	const target = btn.dataset.tabTarget;
	document.querySelectorAll('.tab-btn').forEach(b => { b.classList.remove('active'); b.setAttribute('aria-selected', 'false'); });
	document.querySelectorAll('.tab-panel').forEach(p => p.classList.remove('active'));
	const tabBtn = document.querySelector(`.tab-btn[data-tab="${target}"]`);
	if (tabBtn) { tabBtn.classList.add('active'); tabBtn.setAttribute('aria-selected', 'true'); }
	const tabPanel = document.getElementById('tab-' + target);
	if (tabPanel) { tabPanel.classList.add('active'); window.scrollTo({ top: 0, behavior: 'smooth' }); }
	});
	});
	const tocLinks = document.querySelectorAll('.overview-toc .toc-link:not(.toc-top)');
	const tocTop = document.querySelector('.toc-top');
	const tocSections = Array.from(tocLinks).map(l => document.querySelector(l.getAttribute('href')));
	const onScroll = () => {
	let current = tocSections[0];
	tocSections.forEach(sec => { if (sec && window.scrollY >= sec.offsetTop - 80) current = sec; });
	tocLinks.forEach(l => l.classList.toggle('active', l.getAttribute('href') === '#' + current?.id));
	if (tocTop) tocTop.classList.toggle('visible', window.scrollY > 300);
	};
	window.addEventListener('scroll', onScroll, { passive: true });
	onScroll();
	</script>
	</body>
	</html>