Recently developed methods have made it possible to leverage summary association statistics to perform a wide range of analyses, many of which previously required individual-level data. As the availability of summary association statistics continues to grow (Table 1), summary statistics will continue to be broadly used in analyses involving single-variant association tests, gene-based association tests, fine-mapping, polygenic prediction and inferring polygenic architectures, and cross-trait analysis. The use of summary data will entail a loss of accuracy in some applications, such as imputation, where methods that analyze individual-level data can use haplotypes to model nonlinear structure, and polygenic prediction, where methods that analyze individual-level data can reduce noise by fitting all markers simultaneously; however, when summary statistics are available in larger sample size than individual-level data, the advantage of larger sample size will outweigh those limitations. In addition, there are some settings where summary statistic based methods are the method of choice even when individual-level data is available, such as identifying functional annotations that are enriched for heritability, where methods that analyze individual-level data cannot currently handle a large number of overlapping annotations.
