Genome-Wide Association Study of Genetic Variants Associated with Serum Albumin Levels in Chinese Winter Sports Athletes

Figures

• 

  Figure 1. ALB concentration, distribution, and individual variability among winter sports athletes of different competitive levels. (A) ALB levels in male athletes; (B) ALB levels in female athletes; (C) Distribution of ALB levels in male athletes; (D) Distribution of ALB levels in female athletes; (E) Individual variability in ALB levels among male athletes; (F) Individual variability in ALB levels among female athletes. ns: p > 0.05.

• 

  Figure 2. Genome-wide association analysis of ALB levels in winter sports athletes. The x-axis represents chromosomes, distinguished by different colors, while the y-axis represents −log10(p). The dashed line indicates the suggestive significance threshold (p < 1 × 10−5), and the solid line represents the genome-wide significance threshold (p < 5 × 10−8). The top 15 SNPs (as listed in Table 2) are labeled in the Manhattan plot to aid interpretation; full SNP-level details for all suggestive significant variants are provided in Supplementary Table S1.

• 

  Figure 3. Bubble plot of effect sizes (Beta), p-values, and explained variance (R2) for ALB-associated SNPs. The x-axis represents the SNPs identified in the GWAS, while the y-axis denotes their contribution to ALB level variance. The color of the bubbles represents the p-value (displayed as −log10(p)), with a gradient from red to blue indicating decreasing p-values. The bubble size represents the estimated effect size (Beta), with larger bubbles indicating a greater effect. The red dashed line marks an explanatory power threshold of ≥10%. Detailed effect sizes for each SNP are provided in Table S1.

• 

  Figure 4. Lasso regression cross-validation curve and coefficient path plot. In panel (A), the x-axis represents log(λ), and the y-axis represents the mean cross-validation error. The optimal λ is selected at the point of minimum mean error. The numbers shown at the top of the panel indicate the number of predictors/SNPs with nonzero coefficients (degrees of freedom, df) at each λ. Panel (B) illustrates the shrinkage of regression coefficients as log(λ) changes, with nonzero coefficients at the optimal λ retained for further modeling. The numbers at the top of panel B similarly represent the number of predictors/SNPs with nonzero coefficients (df) at each λ.