선택한 단어 수는 입니다.
최소 단어 이상 선택하여야 합니다.
최소 단어 이상 선택하여야 합니다.
선택한 단어 수는 30입니다.
최대 10 단어까지만 선택 가능합니다.
최대 10 단어까지만 선택 가능합니다.
문제 정의
- In this study, we focus on the prediction of quantitative traits using common genetic variants. We systematically compared the performance of prediction models through real data from the Korea Association Resource (KARE).
제안 방법
- At this time, the number of SNPs was 51. Also, for the model with the fewest SNPs, we selected variables using SLR from ALL-200 and created a model using SLR. The number of SNPs was 38, and the MSE value was 9.
- We then constructed commonly used risk prediction models, such as SLR, LASSO, and EN. Finally, we compared the predictive accuracy by calculating the mean square error (MSE) value for predicting body mass index (BMI). Overall, our results show that LASSO and SLR provide the smallest MSE value among the compared methods.
- In this study, we used statistical methods (SLR, LASSO, and EN) to select variables and build quantitative prediction models. Then, we compared the performance of the quantitative prediction models by each SNP set (ASIAN-100, KOREAN-100, ALL-200, ASIAN-200, KOREAN-200, GWASALL, and GWAS-ASIAN).
- MSE, mean square error; SNP, single nucleotide polymorphism; SLR, stepwise linear regression; LASSO, least absolute shrinkage and selection operator; EN, Elastic-Net; ASIAN-100, GWAS catalog (Asia) ＋ single-SNP analysis; KOREAN-100, GWAS catalog (Korea) ＋ single-SNP analysis; ALL-200, GWAS catalog (All) ＋ single-SNP analysis; ASIAN-200, GWAS catalog (Asia) ＋ single-SNP analysis; KOREAN-200, GWAS catalog (Korea) ＋ single-SNP analysis; GWAS-ALL, GWAS catalog (All); GWAS-ASIAN, GWAS catalog (Asia).
- First, based on three different populations—overall population, Asian-only population, and Korean-only population—we collected the SNPs registered in the GWAS catalog for BMI. Second, the SNPs were selected by single-SNP analysis using linear regression with adjustments for sex, age, and area. We chose the SNPs based on the p-values.
- In this study, we used statistical methods (SLR, LASSO, and EN) to select variables and build quantitative prediction models. Then, we compared the performance of the quantitative prediction models by each SNP set (ASIAN-100, KOREAN-100, ALL-200, ASIAN-200, KOREAN-200, GWASALL, and GWAS-ASIAN). As a result, the performance of the prediction models using the GWAS catalog and KARE data was better than that of the prediction models using only SNPs reported in the GWAS catalog.
- We selected SNPs from the KARE data analysis based on single-SNP analysis and collected SNPs in the GWAS catalog [24]. Then, we performed two steps to make quantitative prediction models. First, we selected the variables by using SLR, LASSO, and EN and then built quantitative prediction models by using the same methods.
- To build a quantitative prediction model, we used the same prediction methods that were applied for the variable selection step for the comparison of these three methods in the variable selection and quantitative prediction. Each prediction model was created by using 7,071 training individuals via 5-fold CV.
대상 데이터
- Overall, SLR selected fewer SNPs than LASSO and EN. All SNPs were selected when EN was used in ASIAN-100, ASIAN-200, and KOREAN-200.
- To build a quantitative prediction model, we used the same prediction methods that were applied for the variable selection step for the comparison of these three methods in the variable selection and quantitative prediction. Each prediction model was created by using 7,071 training individuals via 5-fold CV. To compare the performance of the quantitative prediction models, we calculated the MSE by applying each quantitative prediction model using the test set (n = 1,767).
- In the KARE data, out of 8,838 individuals, we randomly selected 1,767 for test sets and composed the training set with the rest of the 7,071 participants. We selected SNPs using 5-fold cross-validation (CV) of the training set.
데이터처리
- To create the SNP sets associated with BMI, single-SNP analysis was performed by linear regression with adjustments for sex, age, and area. As shown in Supplementary Fig.
이론/모형
- The stepwise procedure was performed using the R package “MASS” [25].
- γ1, γ2, and γ3 represent the sex, age, and area of the i-th individual, respectively. Variable selection was performed by a MSE-based stepwise procedure. The stepwise procedure was performed using the R package “MASS” [25].
성능/효과
- Then, we compared the performance of the quantitative prediction models by each SNP set (ASIAN-100, KOREAN-100, ALL-200, ASIAN-200, KOREAN-200, GWASALL, and GWAS-ASIAN). As a result, the performance of the prediction models using the GWAS catalog and KARE data was better than that of the prediction models using only SNPs reported in the GWAS catalog. For the case that selected variants using LASSO in ASIAN-100 and created a prediction model using SLR, the MSE value was the smallest, 9.
후속연구
- It seems that LASSO and EN tended to select SNPs with little contribution to BMI. For further research, we plan to perform simulation studies and a real-data analysis with other continuous traits.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.