Large-scale trans-ethnic replication and discovery of genetic associations for rare diseases with self-reported medical data

Suyash Shringarpure, Wei Wang, Yunxuan Jiang, Alison Acevedo, Devika Dhamija, Briana Cameron, A. A. Jubb, Peng Yue, Lea Sarov‐Blat, Robert Gentleman, Adam AutonmedRxiv 2021 · DOI

Abstract

A key challenge in the study of rare disease genetics is assembling large case cohorts for well-powered studies. We demonstrate the use of self-reported diagnosis data to study rare diseases at scale. We performed genome-wide association studies (GWAS) for 33 rare diseases using self-reported diagnosis phenotypes and re-discovered 29 known associations to validate our approach. In addition, we performed the first GWAS for Duane retraction syndrome, vestibular schwannoma and spontaneous pneumothorax, and report novel genome-wide significant associations for these diseases. We replicated these novel associations in non-European populations within the 23andMe, Inc. cohort as well as in the UK Biobank cohort. We also show that mixed model analyses including all ethnicities and related samples increase the power for finding associations in rare diseases. Our results, based on analysis of 19,084 rare disease cases for 33 diseases from 7 populations, show that large-scale online collection of self-reported data is a viable method for discovery and replication of genetic associations for rare diseases. This approach, which is complementary to sequencing-based approaches, will enable the discovery of more novel genetic associations for increasingly rare diseases across multiple ancestries and shed more light on the genetic architecture of rare diseases.

Introduction

The study of rare diseases—those affecting fewer than 1 in 2,000 individuals—is traditionally hampered by the difficulty of assembling sufficiently large cohorts for genetic analysis. While clinical recruitment is slow and expensive, understanding the genetic architecture of these conditions is crucial for identifying risk modifiers and therapeutic targets. This study explores the potential of massive consumer-based cohorts and self-reported medical data to overcome these barriers.

Methods and Data

Researchers leveraged the 23andMe platform, utilizing a discovery cohort of over 1.6 million individuals (including 19,084 cases across 33 rare diseases). Participants reported diagnoses via web-based questionnaires mapped to Orphanet terms. Association testing was performed using the SAIGE algorithm, which supports trans-ethnic mixed-model analysis. This approach allowed for the inclusion of related individuals and diverse ancestries (African-American, East Asian, Latino, and South Asian), maximizing statistical power. Findings were replicated using the UK Biobank and independent 23andMe subsets.

The Viability of GWAS for Rare Traits

A key finding of this research is that GWAS remains a powerful tool even for traits with very low case counts. Simulations demonstrated a "Penetrance Paradox": GWAS often has higher power to detect associations for dominant diseases with lower penetrance. This occurs because causal variants must be more frequent in the population to maintain a specific disease prevalence, making them easier to "tag" with common SNPs.

Power simulations for rare disease GWAS across different prevalence and penetrance levels.

Validation and Discovery

The study successfully re-identified 29 known genetic associations, such as CFTR for Cystic Fibrosis and HTT for Huntington’s disease. This high rate of replication validates the quality of self-reported data for genomic research. Comparison of effect sizes between the 23andMe cohort and the GWAS Catalog shows strong agreement across multiple conditions.

Comparison of effect sizes between self-reported 23andMe data and the GWAS Catalog for four rare diseases.

Beyond validation, the study reported several novel genome-wide significant associations:

  • Duane Retraction Syndrome: The first GWAS for this disorder identified loci near OLIG1 and OLIG2.
  • Vestibular Schwannoma: A novel hit was found near the CDKN2A-CDKN2B locus, suggesting shared pathways with more common cancers.
  • Spontaneous Pneumothorax: Three novel loci were identified, including hits near POSTN and BCAR1.

Discussion and Conclusion

The results demonstrate that large-scale online collection of self-reported data is a viable and complementary method to sequencing-based approaches. By including diverse populations in a single mixed-model analysis, the study also highlights the importance of trans-ethnic data in increasing discovery power. While self-reporting has inherent biases and GWAS is limited by its reliance on common variant tagging, this model provides a scalable path forward for mapping the genetics of increasingly rare diseases globally.

Summary updated

Abstract

Abstract A key challenge in the study of rare disease genetics is assembling large case cohorts for well-powered studies. We demonstrate the use of self-reported diagnosis data to study rare diseases at scale. We performed genome-wide association studies (GWAS) for 33 rare diseases using self-reported diagnosis phenotypes and re-discovered 29 known associations to validate our approach. In addition, we performed the first GWAS for Duane retraction syndrome, vestibular schwannoma and spontaneous pneumothorax, and report novel genome-wide significant associations for these diseases. We replicated these novel associations in non-European populations within the 23andMe, Inc. cohort as well as in the UK Biobank cohort. We also show that mixed model analyses including all ethnicities and related samples increase the power for finding associations in rare diseases. Our results, based on analysis of 19,084 rare disease cases for 33 diseases from 7 populations, show that large-scale online collection of self-reported data is a viable method for discovery and replication of genetic associations for rare diseases. This approach, which is complementary to sequencing-based approaches, will enable the discovery of more novel genetic associations for increasingly rare diseases across multiple ancestries and shed more light on the genetic architecture of rare diseases.
BibTeX
@misc{shringarpure2021large,
  author = {Suyash Shringarpure and Wei Wang and Yunxuan Jiang and Alison Acevedo and Devika Dhamija and Briana Cameron and A. A. Jubb and Peng Yue and Lea Sarov‐Blat and Robert Gentleman and Adam Auton},
  title = {{Large-scale trans-ethnic replication and discovery of genetic associations for rare diseases with self-reported medical data}},
  year = {2021},
  doi = {10.1101/2021.06.09.21258643},
  howpublished = {medRxiv},
}
Figure