Population-specific reference panel improves imputation quality for genome-wide association studies conducted on the Japanese population
Jack U. Flanagan, Xiaoxi Liu, David Ortega-Reyes, Kohei Tomizuka, Nana Matoba, Masato Akiyama, Masaru Koido, Kazuyoshi Ishigaki, Kyota Ashikawa, Sadaaki Takata, Mingyang Shi, Tomomi Aoi, Yukihide Momozawa, Kaoru Ito, Yoshinori Murakami, Koichi Matsuda, The Biobank Japan Project, Koichi Matsuda, Yuji Yamanashi, Yoichi Furukawa, Takayuki Morisaki, Yoshinori Murakami, Yoichiro Kamatani, Kaori Muto, Akiko Nagai, Wataru Obara, Ken Yamaji, Kazuhisa Takahashi, Satoshi Asai, Yasuo Takahashi, Takao Suzuki, Nobuaki Sinozaki, Hiroki Yamaguchi, Shiro Minami, Shigeo Murayama, Kozo Yoshimori, Satoshi Nagayama, Daisuke Obata, Masahiko Higashiyama, Akihide Masumoto, Yukihiro Koretsune, Yoichiro Kamatani, Andrew P. Morris, Momoko Horikoshi, Chikashi Terao
Abstract
To improve imputation quality for genome-wide association studies (GWAS) conducted on the Japanese population, we developed and evaluated four Japanese population-specific reference panels. These panels were constructed through the augmentation of the 1000 Genomes Project (1KG) panel using Japanese whole genome sequencing (WGS) data, with sample sizes ranging from 1 K to 7 K individuals enrolled through the Biobank Japan (BBJ) project, and sequencing depths ranging from 3× to 30×. Among these panels, an augmented reference panel comprising 7472 WGS samples of mixed depth (1KG+7K) exhibit the greatest improvement in imputation quality relative to the Trans-Omics for Precision Medicine (TOPMed) reference panel. Notably, we observe these improvements primarily for rare variants with a minor allele frequency (MAF) <5%. To demonstrate the benefits of improved imputation quality in association analyses of complex traits, we conducted GWAS for serum uric acid and total cholesterol levels following imputation up to the 1KG+7K panel. The analysis reveals several loci reaching genome-wide significance (P < 5 × 10–8) in the 1KG+7K imputation output yet remaining undetected when the same sample set is imputed up to the TOPMed reference panel. In summary, the 1KG+7K panel demonstrates significant advantages in the discovery of trait-associated loci, particularly those influenced by low-frequency association signals. This study suggests that enhanced imputation quality in genome-wide association studies, using a Japanese population-specific reference panel, leads to the discovery of rare, trait-associated loci.