Modernizing the family-based association testing (FBAT) approach and its software implementation in the FBAT-program

Developed for linkage and candidate gene studies, family-based association tests (FBATs) have played a fundamental role in the disease mapping efforts of the human genome project. Originally conceived for studies of affected offspring and their parents, FBATs are joint tests for linkage and association that are robust against

confounding due to population substructure by virtue of their design. In terms of the trait/phenotype of interest, FBATs do not require distributional assumptions. In combination with its user-friendly software implementation, the robustness and model-free features of FBAT made the software a frequently used analysis tool that has

contributed to several thousand research projects world-wide. Since its first release in 2001, the package has been constantly adapted to the changing analysis needs of its user community. The approach has been extended to general pedigrees, complex phenotypes, and haplotype analyses. With the arrival of genome-wide association

studies (GWAS) and whole-genome sequencing studies (WGS), multiple-testing strategies and rare-variant analysis approaches have been developed and implemented, which has enabled its continuing use in current WGS studies. However, the full potential of FBAT for WGS studies is restricted by the available null-hypotheses

that can be tested and by its current software implementation that does not support modern, efficient file-formats for genotype data and that does not utilize parallel/distributed computing. In this application, we will expand the testing capabilities of FBAT by implementing additional null-hypotheses that are relevant for the analysis of WGS

data. In WGS analyses, multiple association signals are often detected within the same genomic region and it is unclear whether they stem from multiple disease susceptibility loci (DSLs) or whether a single DSL creates multiple association signals through linkage disequilibrium. We will address this research question by enabling

FBAT to condition the test statistics on other loci with association signals in the same region and thereby, assess the independent contribution of the tested loci to the association. Many genetic association signals have been discovered in studies of unrelated individuals with large sample sizes and relatively small effect sizes. These

findings may be false positives or caused by unmeasured confounding in the association analysis. While the contribution of family-based studies is often limited in large scale mapping efforts as their sample sizes are much smaller than the number of unrelated samples, family-based studies can play a pivotal role in validating findings

given their robustness against genetic confounding and their non-parametric modelling of the outcome. We will introduce equivalence testing into FBAT, which will allow the user to conclude the absence of a genetic effect of meaningful magnitude. Regarding the software implementation, FBAT will be able to accommodate commonly

used binary file formats for the genetic data in addition to the plain text files, which are currently the only supported files formats. To take full advantage of modern Linux clusters with multi-core CPUs, we will add parallel and distributed computing capabilities to FBAT.