Guidelines for investigating causality of sequence variants in human disease

Daniel G. MacArthur, Teri A. Manolio, David Dimmock, Heidi L. Rehm, Jay Shendure, Gonçalo R. Abecasis, David R. Adams, Russ B. Altman, Stylianos E. Antonarakis, Euan A. Ashley, Jeffrey C. Barrett, Leslie G. Biesecker, Don F. Conrad, Gregory M. Cooper, Nancy J. Cox, Mark J. Daly, Mark Gerstein, David B. Goldstein, Joel N. Hirschhorn, Suzanne M. Leal, L Pennacchio, J Stamatoyannopoulos, Shamil Sunyaev, David Valle, Benjamin F. Voight, Wendy Winckler, Chris GunterNature 2014 · DOI

Abstract The discovery of rare genetic variants is accelerating, and clear guidelines for distinguishing disease-causing sequence variants from the many potentially functional variants present in any human genome are urgently needed. Without rigorous standards we risk an acceleration of false-positive reports of causality, which would impede the translation of genomic research findings into the clinical diagnostic setting and hinder biological understanding of disease. Here we discuss the key challenges of assessing sequence variants in human disease, integrating both gene-level and variant-level support for causality. We propose guidelines for summarizing confidence in variant pathogenicity and highlight several areas that require further resource development.

Introduction

As high-throughput sequencing becomes a standard tool in clinical medicine, researchers and clinicians are increasingly confronted with an abundance of rare genetic variants. Every human genome contains hundreds of rare variants that appear "functional"—for instance, those that disrupt a protein-coding sequence. However, the vast majority of these variants are benign and unrelated to specific disease phenotypes.

The primary challenge lies in the "narrative potential" of the human genome: the ease with which a compelling, yet incorrect, story can be constructed to link a rare variant to a patient's symptoms. Previous studies have indicated that up to 27% of published "severe disease mutations" were likely common polymorphisms or lacked sufficient evidence of pathogenicity. Such false assignments of causality lead to erroneous medical advice, unnecessary clinical interventions, and a waste of research resources.

A Hierarchical Framework for Investigating Causality

To address these challenges, a rigorous, hierarchical evidence-based framework is proposed. This framework distinguishes between the implication of a gene in a disease and the implication of a specific variant within that gene.

1. Gene-Level Implication

Before a specific variant can be labeled as pathogenic, the gene itself must be robustly linked to the disease. This requires either strong statistical support—demonstrating a significantly higher burden of rare mutations in patients compared to healthy controls—or definitive experimental evidence. To maintain statistical rigor across the ~20,000 genes in the human genome, a Bonferroni-corrected P-value threshold of $1.7 \times 10^{-6}$ is recommended for sequencing studies.

2. Variant-Level Implication

Once a gene's involvement is established, individual variants are evaluated across three primary categories of evidence:

  • Genetic Evidence: This includes the variant's co-segregation with the disease within families and its absence or extreme rarity in large-scale population databases.
  • Informatic Evidence: This involves assessing whether the DNA site is evolutionarily conserved and using computational algorithms to predict the variant's impact on protein structure or function.
  • Experimental Evidence: This requires laboratory assays or animal models to demonstrate that the variant actually disrupts the protein's biological function.

The Role of Reference Datasets and Null Models

The framework relies heavily on large-scale reference datasets to establish the "background noise" of human variation. Resources such as the Exome Variant Server (EVS), the 1000 Genomes Project, and centralized repositories like ClinVar and LOVD are essential for filtering out common variants that are unlikely to cause rare diseases.

A critical component of this evaluation is the use of quantitative null models. The paper highlights the case of the TTN (Titin) gene, the largest in the human genome. Because of its size, TTN is a large target for random mutations. For example, observing four de novo mutations in TTN within an autism cohort might seem significant, but statistical modeling shows such a result can occur by chance ($P = 0.14$). This underscores the danger of relying on "intuitive" genetics without accounting for gene size and mutation rates.

Key Insights and Clinical Implications

  • The "Gene First" Rule: Pathogenicity cannot be confidently assigned to a variant if the underlying gene has not been statistically linked to the phenotype.
  • Conservation vs. Causality: Evolutionary conservation is a useful indicator but does not constitute proof of disease causality.
  • The Power of Negative Evidence: Identifying a putative "disease-causing" variant in healthy, elderly individuals in population databases is often the most definitive evidence against its pathogenicity.
  • Standardized Nomenclature: It is vital to distinguish between "damaging" variants (those that disrupt a protein) and "pathogenic" variants (those that cause disease).
  • Transparency: Researchers must clearly communicate the level of certainty, using terms like "suggestive" when evidence is not definitive, to prevent clinical misinterpretation.

Limitations and Future Directions

While these guidelines provide a necessary foundation, several challenges remain. For extremely rare "N-of-1" cases, traditional statistical significance may be unattainable, requiring a careful synthesis of weaker evidence. Furthermore, the issues of variable penetrance—where a mutation does not cause disease in all carriers—and the assessment of non-coding regulatory elements ("dark matter") require more extensive, well-phenotyped cohorts and new experimental frameworks. Finally, the scientific community must shift incentive structures to reward the replication and rigorous validation of genetic findings over the mere discovery of novel associations.

Summary updated

Abstract

The discovery of rare genetic variants is accelerating, and clear guidelines for distinguishing disease-causing sequence variants from the many potentially functional variants present in any human genome are urgently needed. Without rigorous standards we risk an acceleration of false-positive reports of causality, which would impede the translation of genomic research findings into the clinical diagnostic setting and hinder biological understanding of disease. Here we discuss the key challenges of assessing sequence variants in human disease, integrating both gene-level and variant-level support for causality. We propose guidelines for summarizing confidence in variant pathogenicity and highlight several areas that require further resource development.
BibTeX
@inproceedings{macarthur2014guidelines,
  author = {Daniel G. MacArthur and Teri A. Manolio and David Dimmock and Heidi L. Rehm and Jay Shendure and Gonçalo R. Abecasis and David R. Adams and Russ B. Altman and Stylianos E. Antonarakis and Euan A. Ashley and Jeffrey C. Barrett and Leslie G. Biesecker and Don F. Conrad and Gregory M. Cooper and Nancy J. Cox and Mark J. Daly and Mark Gerstein and David B. Goldstein and Joel N. Hirschhorn and Suzanne M. Leal and L Pennacchio and J Stamatoyannopoulos and Shamil Sunyaev and David Valle and Benjamin F. Voight and Wendy Winckler and Chris Gunter},
  title = {{Guidelines for investigating causality of sequence variants in human disease}},
  booktitle = {Nature},
  year = {2014},
  doi = {10.1038/nature13127},
}
Figure