Quantifying the unobserved protein-coding variants in human populations provides a roadmap for large-scale sequencing projects

James Zou, Gregory Valiant, Paul Valiant, Konrad J. Karczewski, Siu On Chan, Kaitlin E. Samocha, Monkol Lek, Shamil Sunyaev, Mark J. Daly, Daniel G. MacArthur

TL;DR

This paper introduces "UnseenEst," a novel algorithm that accurately estimates the frequency distribution of all genetic variants in a population, including those not yet observed in current sequencing data. This provides a crucial roadmap for planning future large-scale human sequencing projects, like the U.S. Precision Medicine Initiative, by predicting how many new variants will be discovered and their expected frequencies.

What Problem Does This Paper Address?

As human genome sequencing projects grow exponentially in size, researchers need to understand the statistical power of these initiatives. Specifically, they need to predict how many new genetic variants will be discovered as cohort sizes increase and what their frequencies will be. Without this foresight, designing effective studies and evaluating the potential of large datasets for precision medicine becomes challenging. This paper tackles the problem of accurately quantifying the landscape of unobserved rare genetic variants to inform the design and interpretation of these ambitious sequencing efforts.

Methods

The core of this work is the development of a new algorithm called UnseenEst. This algorithm uses a linear programming approach to estimate the frequency distribution of all variants, including those not present in the current sequencing data, based on the observed site frequency spectrum (SFS) of a given cohort. UnseenEst is designed to accurately extrapolate from current data to cohort sizes that are an order of magnitude larger. The authors validate UnseenEst by training it on a subset of data and then predicting the number of variants in the full dataset, showing high accuracy. They also categorize variants by their functional consequence (synonymous, missense, loss-of-function) and biochemical properties to provide more granular predictions.

Datasets

The primary dataset used in this study is the Exome Aggregation Consortium (ExAC) dataset, which contains exome sequencing data from 60,706 healthy individuals. The authors also created a U.S. Census-matched cohort of 33,778 individuals from the ExAC dataset for their analyses.

Key Insights and Ideas

  • The vast majority of human genetic variation remains undiscovered: Even with large sequencing efforts, a significant fraction of possible variants, especially rare ones, are yet to be observed.
  • UnseenEst provides a powerful predictive tool: The algorithm can accurately forecast the discovery rate of new variants and their frequencies, enabling better study design.
  • Loss-of-function (LoF) variants are particularly rare and informative: These variants are likely deleterious and occur at very low frequencies, making their discovery and study crucial for understanding gene essentiality and disease.
  • Gene essentiality can be inferred from LoF variant frequencies: Genes with very low LoF allele frequencies are strongly intolerant to inactivation, suggesting they are essential for human survival.
  • Population composition matters: Optimizing the ancestral composition of sequencing cohorts can boost the overall discovery rate of rare variants.

Key Results

  • With a cohort of 500,000 individuals, UnseenEst predicts that we will capture 7.5% of all possible loss-of-function (LoF) variants and 12% of all possible missense variants.
  • The study estimates that approximately 2,900 genes have LoF allele frequencies lower than 10⁻⁵, indicating strong intolerance to inactivation.
  • UnseenEst accurately predicted the number of unique variants observed in the full ExAC cohort when trained on 10% of the data.
  • The algorithm demonstrated high accuracy in estimating allele frequencies for different variant classes (synonymous, missense, LoF, CpG).

Limitations and Future Work

  • Independence assumption: The UnseenEst model assumes variants are independent, which is a simplification. Population demography and linkage disequilibrium can introduce correlations, especially for common variants. However, the authors note that for very rare variants, this assumption is a better approximation.
  • Proxy for LoF: The definition of LoF variants used in the study (stop codon introduction or splice site disruption) serves as a proxy for true loss-of-function and doesn't explicitly assess protein function.
  • Future work: The authors suggest that UnseenEst can be applied to estimate the discovery rate of cohorts with different ancestral compositions, opening avenues for more tailored study designs. The framework is also generalizable beyond genetics for estimating properties of any finite list of probabilities.

Summary updated

Abstract

As new proposals aim to sequence ever larger collection of humans, it is critical to have a quantitative framework to evaluate the statistical power of these projects. We developed a new algorithm, UnseenEst, and applied it to the exomes of 60,706 individuals to estimate the frequency distribution of all protein-coding variants, including rare variants that have not been observed yet in the current cohorts. Our results quantified the number of new variants that we expect to identify as sequencing cohorts reach hundreds of thousands of individuals. With 500K individuals, we find that we expect to capture 7.5% of all possible loss-of-function variants and 12% of all possible missense variants. We also estimate that 2,900 genes have loss-of-function frequency of less than 0.00001 in healthy humans, consistent with very strong intolerance to gene inactivation.
BibTeX
@inproceedings{zou2016quantifying,
  author = {James Zou and Gregory Valiant and Paul Valiant and Konrad J. Karczewski and Siu On Chan and Kaitlin E. Samocha and Monkol Lek and Shamil Sunyaev and Mark J. Daly and Daniel G. MacArthur},
  title = {{Quantifying the unobserved protein-coding variants in human populations provides a roadmap for large-scale sequencing projects}},
  booktitle = {Nature Communications},
  year = {2016},
  doi = {10.1038/ncomms13293},
}
Figure