FOLD : Fully powered method for OverLapping Data.

FOLD/FOLD-split is a free, open-source meta-analysis framework, which achieves optimal performance when case-control studies include shared control samples.


Recent studies have developed several meta-analysis methods that can systematically account for overlapping subjects at the summary statistics level.

Existing methods : Bhattacharjee et al. 2012; Bulik-Sullivan et al. 2015; Han et al. 2016; Lin and Sullivan 2009; Zaykin and Kozbur 2010

We observed that when not all controls are shared between studies, the existing approaches show a severe power drop compared with splitting(a strategy that splits genotype data of overlapping samples into individual studies).

Image of the power.jpg

Striking phenomenon

  • Simulation code:

simulate-v1.0 (R script) <- Right click and download

  • Command

Inspired by this phenomenon, we propose a power-preserving meta-analysis framework for overlapping samples called FOLD (Fully-powered method for OverLapping Data).


FOLD obtains multiple summary statistics from each study, such that each statistic is calculated using samples that are homogeneous in terms of their information.

The frame of the gold standard approach and FOLD are

Image of the FOLD_scheme.jpg


  • Code and manual:

FOLD <- Right click and download


FOLD can be complicated if the number of configurations of sharing(T) is large. In such situations, splitting can be simpler.

FOLD-split helps with split individuals to maximize power.

Equal splitting and Case-based splitting are two widely known methods, we compared FOLD-split with Equal and Case-based splitting approaches

Power of the FOLD-split.jpg

Power of different methods:

Strategy A is a strategy that equally distributes shared controls into studies(Equal splitting)

Strategy B is a strategy that distributes shared controls proportionally to the number of cases in each study(Case-based splitting)


  • Code and manual:

FOLD-split <- Right click and download


FOLD: A method to optimize power in meta-analysis of genetic association studies with overlapping subjects


Eunji Kim (