Background After the recent successes of genome-wide association studies (GWAS), one key challenge is to identify genetic variants that might have a significant joint effect on complex diseases but have failed to be identified individually due to weak to moderate marginal effect. One popular and effective approach is gene set based analysis, which investigates the joint effect of multiple functionally related genes (eg, pathways). However, a typical gene set analysis method is biased towards long genes, a problem that is especially severe in psychiatric diseases.
Methods A novel approach was proposed, namely generalised additive model (GAM) for GWAS (gamGWAS), for gene set enrichment analysis of GWAS data, specifically adjusting the gene length bias or the number of single-nucleotide polymorphisms per gene. GAM is applied to estimate the probability of a gene to be selected as significant given its gene length, followed by weighted resampling and computation of empirical p values for the rank of pathways. We demonstrated gamGWAS in two schizophrenia GWAS datasets from the International Schizophrenia Consortium and the Genetic Association Information Network.
Results The gamGWAS results not only confirmed previous findings, but also highlighted several immune related pathways. Comparison with other methods indicated that gamGWAS could effectively reduce the correlation between pathway p values and its median gene length.
Conclusion gamGWAS can effectively relieve the long gene bias and generate reliable results for GWAS data analysis. It does not require genotype data or permutation of sample labels in the original GWAS data; thus, it is computationally efficient.
- gene set enrichment analysis
- generalised additive model
- cancer: lung
Statistics from Altmetric.com
Funding This work was partially supported by National Institutes of Health grant MH083094, 2009 NARSAD Maltz Investigator Award (to ZZ), and 2010 NARSAD Young Investigator Award (to PJ).
Competing interests None.
Ethics approval Not applicable because this study is not classified as human subjects research. For the GAIN dataset, the genotyping of samples was provided through the Genetic Association Information Network (GAIN) and obtained from the database of Genotype and Phenotype (dbGaP) found at http://www.ncbi.nlm.nih.gov/gap through dbGaP accession number [phs000021.v2.p1] (data access request and approval #4532-2). For the ISC dataset, the genotyping of samples was approved by Institutional Review Boards in the original study (Purcell et al. Nature 2009;460:748–52).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement There is no additional unpublished data from this study. All the data used in this work was from previous publications.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.