The Power of Partnership: Democratizing Genetic Prevalence to Empower Patient Advocacy

By partnering with 18 patient organizations to utilize the publicly available GeniE tool for analyzing gnomAD data, this study demonstrates how democratizing genetic prevalence estimates for 22 rare conditions empowers patient advocacy and highlights the dynamic nature of these metrics as genomic databases evolve.

The Gist

Imagine you are trying to figure out how many people in a massive city have a specific, rare recipe for a cake that, if eaten by two people who both have it, causes a very serious illness.

For a long time, doctors tried to count these people by looking at hospital records. But this is like trying to count all the people in the city by only looking at the people who showed up at the bakery to buy a cake. Many people have the recipe but never got sick, never went to the doctor, or didn't even know they had it. So, the old counts were often wrong, incomplete, or just plain missing.

This paper describes a new, smarter way to count: Genetic Prevalence. Instead of waiting for people to get sick, the researchers looked at the "recipe books" (DNA) of thousands of healthy people to see how many carry the ingredients for the bad cake.

Here is the breakdown of their journey, explained simply:

1. The Problem: The "Static" Map

Imagine you have a map of the city drawn in 2018. It's okay, but the city has grown, new neighborhoods have been built, and some streets have changed names. If you try to find a house using that old map, you might get lost.

In genetics, the "map" is a database called gnomAD. For years, scientists used an older version of this map (v2.1) to guess how common rare diseases were. But in 2024, a much bigger, more detailed version (v4.1) was released. The researchers wanted to see: Does our new map change how many people we think are at risk?

2. The Team: The "Rare As One" Network

The researchers didn't work in an ivory tower. They partnered with 18 different patient advocacy groups (like the "Rare As One" network). Think of these groups as the local community leaders who actually know their neighborhoods best.

• The Scientists brought the high-tech maps and math tools.
• The Patient Groups brought the real-world context. They said things like, "Hey, we know a lot of people with this condition in Turkey, but your map doesn't show them," or "We found a specific mutation in our community that your map missed."

This partnership was crucial. It was like a GPS app that updates in real-time because the drivers (patients) are telling it where the roadblocks are.

3. The Tool: "GeniE" (The Calculator)

Before this, calculating these numbers required a PhD in computer science and a supercomputer. The team built a free, easy-to-use website called GeniE (Genetic Prevalence Estimator).

• Analogy: Think of GeniE as a "Nutrition Label Generator" for genes. You tell it which "ingredients" (genetic variants) cause the disease, and it instantly tells you how many people in the population likely have those ingredients.
• It makes the math transparent. Anyone can see exactly how the number was calculated, rather than just trusting a black box.

4. The Results: The Numbers Changed!

When they ran the numbers using the new, bigger map (v4.1), things shifted significantly.

• The "Shrinking" Effect: For many diseases, the estimated number of people at risk actually went down. Why? Because the new map was so much bigger and more detailed that it realized some "bad ingredients" were actually just harmless typos in the recipe book. The old map was overestimating the danger.
• The "Growing" Effect: For some specific groups (like people of African or East Asian descent), the risk went up. The old map didn't have enough people from these backgrounds, so it missed specific variants that are common in those communities. The new map caught them.
• The Big Takeaway: The "prevalence" (how common a disease is) isn't a fixed number written in stone. It's more like a weather forecast—it changes as we get better data.

5. Why This Matters

Why should you care if the number is 1 in 100,000 or 1 in 200,000?

• For Patients: It helps them understand their risk and find others like them.
• For Drug Companies: If a company wants to make a medicine for a rare disease, they need to know how many people might need it. If the number is too low, they won't bother making the drug. If the number is higher than they thought, it might be worth the investment.
• For Doctors: It helps them decide who should get tested.

The Bottom Line

This paper is a celebration of partnership. It shows that when scientists, patient advocates, and new technology work together, we get a clearer picture of the world.

They built a tool (GeniE) that democratizes this knowledge, meaning it's no longer locked away in expensive labs. It's now open for anyone to use, ensuring that as our understanding of human genetics evolves, our estimates of disease risk evolve with it, keeping everyone—from patients to policymakers—better informed.

In short: We used to guess how many people had a rare genetic condition by looking at a blurry, old photo. Now, with a high-definition camera (new data) and a team of local guides (patient groups), we have a crystal-clear picture that helps us build a better future for everyone.

Technical Summary

1. Problem Statement

Accurate disease prevalence estimation is critical for public health resource allocation, therapeutic development, and carrier screening. However, traditional methods suffer from significant limitations:

• Case Counting/ICD-10: Often biased due to lack of specific codes, under-diagnosis, or reliance on physician recognition.
• Newborn Screening (NBS): Limited to a small subset of conditions and varies geographically.
• Static Genetic Estimates: Existing genetic prevalence estimates (based on Hardy-Weinberg Equilibrium) are often static, relying on outdated variant classifications and allele frequencies (AF) from specific database versions. They rarely incorporate real-world insights from patient communities.
• Accessibility Gap: Calculating genetic prevalence typically requires advanced bioinformatics skills and access to specific database annotations, limiting utility for patient advocacy groups and non-computational researchers.

2. Methodology

The study employed a collaborative, multi-step approach to estimate genetic prevalence for 22 autosomal recessive (AR) conditions.

A. Partnership Model

• Collaborated with 18 Rare As One (RAO) patient organizations.
• Patient groups provided input on gene selection, condition-specific variant lists, and evidence for pathogenicity.
• Results were returned directly to these groups for validation and contextual interpretation.

B. Data Sources & Variant Selection

• Database: Used two releases of the Genome Aggregation Database (gnomAD): v2.1 and v4.1 (released April 2024).
• Inclusion Criteria: Variants passed gnomAD QC filters (PASS) with an allele count (AC) $\ge$ 1.
• Sources for Pathogenicity:
  • ClinVar (Pathogenic/Likely Pathogenic/Conflicting).
  • HGMD (Disease Mutations).
  • gnomAD (High-confidence predicted Loss of Function [pLoF] and high-reliability missense variants with REVEL $\ge$ 0.932).
  • Patient-provided candidate variants.
• Triaging Strategy (v2.1 to v4.1): Due to the influx of new variants in v4.1 (n=4,600), a triage approach was used:
  • Variants with AC $\ge$ 15 received full curation.
  • Previously reviewed Variants of Uncertain Significance (VUS) underwent updated literature searches.
  • New high-confidence pLoF variants (AC $\ge$ 2) underwent pLoF curation to rule out nonsense-mediated decay (NMD) escape.

C. Calculation Framework

• Hardy-Weinberg Equilibrium (HWE): Used to calculate carrier frequency ($2q_a$) and genetic prevalence ($q_a^2$), where $q_a$ is the aggregate allele frequency of all disease-causing variants in a gene.
• Estimation Types:
  • Conservative: Includes only Pathogenic (P) and Likely Pathogenic (LP) variants.
  • Relaxed: Includes P, LP, and VUS-high variants.
• Exclusion Criteria: Variants were excluded if classified as VUS with limited case data, benign, likely mapping errors, or pLoF in homopolymer regions without supporting evidence.

D. Tool Development: GeniE

• Developed the Genetic Prevalence Estimator (GeniE), a web-based, open-source tool (broad.io/genie).
• Features: Allows users to input variant lists, perform calculations using gnomAD AFs, and download results. Includes a Dashboard for preliminary estimates and comparison with Orphanet data.

3. Key Results

A. Prevalence Estimates (gnomAD v4.1)

• Conservative Carrier Frequencies: Ranged from 1/164 (NEB) to 1/11,888 (HADH).
• Genetic Prevalence: Ranged from 1/194,753 to 1/94,984,678.
• Dynamic Changes: The median change in genetic prevalence between v2.1 and v4.1 was 0.806 (a 20% decrease on average), with a wide range of 0.50 to 14.96-fold changes.
  • Drivers of Change: Primarily updated allele frequencies (81% of shared variants had decreased AF in v4.1 due to larger sample sizes) and the addition of new variants.
  • Ancestry Specificity: While global estimates often decreased, specific ancestry groups (e.g., African, East Asian, Admixed American) showed increased carrier frequencies due to the discovery of population-specific founder variants (e.g., ABCC8 in African/East Asian groups; HPS1 Puerto Rican founder).

B. Variant Curation & Source Analysis

• Source Contribution: ~68% of included variants were unique to a single source.
• Database Reliance: ClinVar accounted for 85.6% of the cumulative carrier frequency, followed by gnomAD pLoFs (11.8%) and HGMD (2.5%).
• Classification Agreement:
  • Agreement with ClinVar P/LP classifications: 76.5%.
  • Agreement with HGMD Disease Mutations: 48.7% (dropping to 18.7% for HGMD-only variants).
  • Implication: Manual curation is essential; reliance on HGMD alone introduces significant noise.

C. Patient Feedback

• Impact: 16/17 organizations rated the estimates as "highly impactful."
• Utility: Data is being used for grant applications, biopharma engagement, strategic planning, and justifying genetic testing access in underrepresented communities.
• Discrepancies: Some estimates were lower than expected due to underrepresentation of specific ancestries in gnomAD (e.g., Middle Eastern populations for HADH) or missing founder variants.

4. Key Contributions

1. Democratization of Data: The creation of GeniE removes computational barriers, allowing patient advocates and non-bioinformaticians to generate and interpret genetic prevalence data.
2. Dynamic Modeling: Demonstrated that genetic prevalence is not a static number but an evolving metric that changes significantly with database updates (v2.1 vs. v4.1) and variant reclassification.
3. Participatory Science: Established a robust framework for integrating patient advocacy groups directly into the scientific curation and interpretation process, ensuring estimates align with real-world clinical and demographic realities.
4. Best Practices: Defined five key factors influencing interpretation: Representation, Curation, Sequencing Sensitivity, Gene-Disease Novelty, and Disease Spectrum (including penetrance and life expectancy).

5. Significance

• For Therapeutic Development: Provides industry and researchers with more accurate, up-to-date target population sizes, crucial for clinical trial design and market sizing.
• For Public Health: Highlights the limitations of current population databases (gnomAD) regarding global diversity, urging the inclusion of underrepresented ancestries to prevent underestimation of disease burden.
• For Patient Advocacy: Empowers organizations with quantitative data to advocate for funding, policy changes, and drug development, transforming them from passive recipients of data to active partners in genomic science.
• Future Outlook: The authors note current limitations (e.g., applicability only to fully penetrant AR conditions) and plan to expand GeniE to handle reduced penetrance, locus heterogeneity, and autosomal dominant/X-linked conditions.

Conclusion: The study successfully bridges the gap between complex genomic data and actionable public health insights by combining rigorous bioinformatics with patient-centered collaboration, proving that genetic prevalence is a dynamic, community-driven metric essential for the future of rare disease research.