Pitfalls of predicting complex traits from SNPs

By Naomi R. Wray, Jian Yang, Ben J. Hayes, Alkes L Price, Michael E. Goddard, Peter M. Visscher. 2013

720 citations recorded by the source corpus.

The success of genome-wide association studies (GWASs) has led to increasing interest in making predictions of complex trait phenotypes, including disease, from genotype data. Rigorous assessment of the value of predictors is crucial before implementation. Here we discuss some of the limitations and pitfalls of prediction analysis and show how naive implementations can lead to severe bias and misinterpretation of results.

Subfields: Genetics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Genetics

Citations in the corpus

Fully-Connected Neural Networks with Reduced Parameterization for Predicting Histological Types of Lung Cancer from Somatic Mutations (2020, background): “In that case, the association cannot be identified through GWAS because it tests each gene locus independently for association with a phenotype of interest [4].”
Efficient Bayesian mixed model analysis increases association power in large cohorts (2014, methodology): “Again, larger sample sizes will enable further gains 12,34 .”
Improved polygenic prediction by Bayesian multiple regression on summary statistics (2019, background): “(6), which is the kernel of the normal distribution with mean β̂j and variance σ(2)e=lj .”
Research review: Polygenic methods and their application to psychiatric traits. (2014, background): “Factors that can bias GPRS results have been discussed in detail elsewhere (Wray et al., 2013).”
Dissecting the genetics of complex traits using summary association statistics (2016, result): “…evaluating polygenic risk scores to avoid non-independence of training and validation samples (e.g. due to cryptic relatedness or shared population stratification), which could cause prediction accuracy to be overstated relative to what could be achieved in an independent validation sample 74,80 .”
Meta-analysis of genome-wide association studies for height and body mass index in ∼700,000 individuals of European ancestry (2018, background): “(9–11) and others (12) have helped clarifying the distinction between what can potentially be explained by all SNPs (aka SNP heritability, h(2)SNP) and what remains out of the reach of SNP array-based GWAS, for instance causal variants that are not tagged by genotyped or imputed SNPs.”
A tutorial on conducting genome‐wide association studies: Quality control and statistical analysis (2018, positive): “…for schizophrenia (the discovery sample) has been used to significantly predict the risk in target samples with various phenotypes, such as bipolar disorder, level of creativity, and even risk of immune disorders (Power et al., 2015; Purcell et al., 2009; Stringer et al., 2014; Wray et al., 2013).”
Efficient Test and Visualization of Multi-Set Intersections (2015, background): “…SNPs may be generally not the true causal variants for a phenotype, but are in linkage disequilibrium (LD) with one or more causal variants 26 , the genes closest to the predicted SNPs were annotated as the candidate causal genes (called the mapped genes) by the GWAS Catalog, resulting in…”
Multi-trait analysis of genome-wide association summary statistics using MTAG (2018, neutral): “Detailed results from DEPICT for each trait are found in Supplementary Tables 21–29.”