Supplemental Material for Kwong et al., 2021

online resource

posted on 2021-03-09, 16:05 authored by Alan Kwong, Thomas W. Blackwell, Jonathon LeFaive, Mariza de Andrade, John Barnard, Kathleen C. Barnes, John Blangero, Eric Boerwinkle, Esteban G. Burchard, Brian E. Cade, Daniel I. Chasman, Han Chen, Matthew P. Conomos, L. Adrienne Cupples, Patrick T. Ellinor, Celeste Eng, Yan Gao, Xiuqing Guo, Marguerite Ryan Irvin, Tanika N. Kelly, Wonji Kim, Charles Kooperberg, Steven A. Lubitz, Angel C. Y. Mak, Ani W. Manichaikul, Rasika A. Mathias, May E. Montasser, Courtney G. Montgomery, Solomon Musani, Nicholette D. Palmer, Gina M. Peloso, Dandi Qiao, Alexander P. Reiner, Dan M. Roden, M. Benjamin Shoemaker, Jennifer A. Smith, Nicholas L Smith, Jessica Lasky Su, Hemant K. Tiwari, Daniel E. Weeks, Scott T. Weiss, Laura J. Scott, Albert V. Smith, Gonçalo R. Abecasis, Michael Boehnke, Hyun Min Kang

Fig S1 contains the RUTH algorithm.

Fig S2 contains a comparison of using 20 vs. 100 iterations for the E-M algorithm.

Fig S3 contains ROC and PRC for simulated single-ancestry data.

Fig S4 contains PRC for simulated data with multiple ancestries.

Fig S5 contains PRC for simulated data cross different allele frequencies.

Fig S6 contains PRC for 1000G and TOPMed variants.

Fig S7 contains results for testing TOPMed variants found in the 1000G variant list.

Fig S8 contains ROC curves for TOPMed variants found in the 1000G variant list.

Fig S9 contains PRC curves for TOPMed variants found in the 1000G variant list.

Fig S10 contains the results of testing 1000G and TOPMed variants with RUTH using two vs. four PCs.

Fig S11 contains the effect of ancestry estimation accuracy on PRCs.

Fig S12 contains PC plots and group assignments for 1000G and TOPMed samples.

Fig S13 contains results of testing 1000G and TOPMed variants with meta-analysis using K-means to generate ancestry groups.

Table S1 contains simulation results for the unadjusted test, meta-analysis, RUTH, and PCAngsd for HWE.

Table S2 contains results from using lower quaity ancestry estimations on meta-analysis and RUTH.

Table S3 contains the performance of the unadjusted test, meta-analysis, and RUTH on the subset of TOPMEd freeze 5 chromosome 20 variants that are also found in 1000G.

Table S4 contains simulation results for RUTH tests using 2 vs 4 PCs.

Table S5 contains the effect of high vs. low quality subpopulation classification on meta-analysis in simulated samples.

Table S6 contains a comparison of runtimes and memory requirements for RUTH and PCAngsd in simulated and 1000G data.

Table S7 contains sample contributions from each of the participating TOPMed studies.

Table S8 contains TOPMed acknowledgements for omics support.

File S1 contains TOPMed study acknowledgements.