PMC:3630385 / 6560-7719 JSONTXT

{"project":"2_test","denotations":[{"id":"23613682-20084277-44836355","span":{"begin":997,"end":998},"obj":"20084277"},{"id":"23613682-21714648-44836356","span":{"begin":1000,"end":1002},"obj":"21714648"},{"id":"23613682-20699438-44836357","span":{"begin":1004,"end":1006},"obj":"20699438"},{"id":"23613682-21680795-44836358","span":{"begin":1008,"end":1010},"obj":"21680795"}],"text":"Statistical analysis\nWe performed logistic regression analyses between the somatic mutations and NK-AML using PLINK/SEQ v0.08 (http://atgu.atgu.mgh.harvard.edu/plinkseq), which provides powerful utilities in variant call format (vcf) for analyzing whole-exome and -genome data. Further, we verified the odd ratios and p-values estimated from PLINK/SEQ using Stata, v11.2 (Stata Corp., College Station, TX, USA).\nWe selected the somatic nsSNVs with complete call rates and evaluated the GRS models composed of the variants associated with NK-AML. The GRS was calculated for each individual by accumulating the number of risk alleles (0, 1, or 2) of the SNVs. We created stepwise GRS models, comprised of the selected SNVs, according to their significance level; if the significance level was equal between two or more SNVs, we selected the SNVs in the order of their chromosomal position. In addition, we evaluated a GRS model that consisted of gene variants reported in previous leukemia studies [9, 10, 18, 19]. We compared the area under the receiver operating characteristic curve (AUC) of each GRS model using the \"roctab\" and \"roccomp\" commands in Stata."}