Quality controls and data filtrationIn this study, after quality controls of extracted data, we aimed to reach high genotyping quality, so no samples were filtered out and all 3221 ones (which had information on statin response) remained for further analysis. Out of 893,634 SNP markers, 208893 SNPs and 20719 SNPs were removed due to minor allele threshold (MAF < 0.05) and deviations from Hardy–Weinberg equilibrium test (p < 10−6), respectively. Finally, the 658,084 SNPs were kept for downstream analysis. Based on our outlier detection test through PCA, two samples have been recognized as outliers and excluded during the analysis procedure (Supplementary File 1, Fig. 1). LD pattern investigation in 10, 25, and 50 kb, revealed a smooth decline in LD (0.34, 0.29, and 0.27 on average) and almost remained constant for 100 kb distance which was not represented on the plot (Supplementary File1, Fig. 2).Genome-wide association mappingCalculations of variance components and heritability (h2) for the traits were done using the GEN2VCF programming software. The results illustrated that the heritability (h2) of Statin-LDL and Statin-TC reached 0.175 and 0.067, respectively. The results also revealed additive genetic, residual, and phenotypic variance for Statin-LDL samples, which was expressed as a scattering of data around the mean, estimated at 0.059, 0.277, and 0.336, respectively. Therefore, the additive genetic variance was lower variance and the phenotypic one was upper (Table 2). Our results also illustrated that additive genetic, residual, and phenotypic variance for Statin-TC samples were estimated at 0.025, 0.348, and 0.373, respectively, which as same as Statin-LDL.Table 2 Variance components and genetic parameter estimates for the studied traits.The results represented eight SNPs associated with changing the levels of LDL/TC in response to a statin drug in cardiovascular patients. Also, the results represented the closest protein-coding gene that could be associated with statin response (Table 3). Among the candidate SNPs, three (rs17502794, rs10785232, rs4785621) are located in the intragenic (intronic) area and five (rs10820084, rs44744370, rs102179528, rs1966503, rs484071) are located in intergenic regions (Table3). The p-value for the correlation of all eight SNPs with the response of statin was in the range of 2.79E−07 to 8.14E−06. The best correlation was observed in rs17502794 with the lowest p-value (2.79E−07). The results suggest that the Minor Allele Frequency (MAF) of candidate SNPs varied from 0.093 to 0.245 in Statin-LDL and Statin-TC patient samples. The highest MAF was corresponding to rs484071 (Table 3).Table 3 Significant SNP associated with the statin response.The result for associating SNPs with the statin response showed three SNPs (rs10820084, rs4803750, rs10989887) significantly accompanied the response of decreased LDL levels in cardiovascular patients (Fig. 1A). Also, it revealed that one SNP (rs1966503) is close to significant in association with statin response. Based on the results, two significant SNPs were located in chromosome 9 and one in chromosome 19. In addition, a SNP close to the significant level change was located on chromosome 16.Figure 1(A) Manhattan plot for associations of SNPs with the statin_LDL response. X-axis: SNPs positions on chromosomes, Y-axis: − Log10 p-value. The green and red dashed lines show the genome- and chromosome-wide significance level (3.247 × 10−7 and 7.144 × 10−06) based on the number of independent SNPs identified by the statistical simpleM approach. (B) QQ plot for statin_LDL. The genomic control “inflation factor” lambda = 1.004 for linear regression. The blue area shows a 95% confidence interval.In the same way, the result for the association of SNPs with the statin response showed that three SNPs (rs17502794, rs10785232, rs484071) significantly accompanied the response of decreased total cholesterol (TC) levels in cardiovascular patients (Fig. 2A). Also, it revealed that one SNP (rs4785621) was close to significant in association with statin response. Based on the results, individual significant SNPs were located in chromosomes 12, 18, and 20. In addition, a SNP close to the significant level change was located on chromosome 16.Figure 2(A) Manhattan plot for associations of SNPs with the statin_TC response. X-axis: SNPs positions on chromosomes, Y-axis: − Log10 p-value. The green and red dashed lines show the genome- and chromosome-wide significance level (3.247 × 10−7 and 7.144 × 10–6) based on the number of independent SNPs identified by the statistical simpleM approach. (B) QQ plot for statin_TC. The genomic control “inflation factor” lambda = 1.011 for linear regression. The blue area shows a 95% confidence interval.The results indicated that in the statin-LDL group, two SNPs (rs10820084, rs10989887) caused resistance to statin drugs in cardiovascular patients (\(\beta_{SNP}\) = 0.140997, \(\beta_{SNP}\) = 0.137665), and two SNPs (rs4803750, rs1966503) caused response to statin drugs in cardiovascular patients (\(\beta_{SNP}\) = − 0.133404, \(\beta_{SNP}\) = − 0.121344). In a similar way, the results revealed that in the statin-TC group, two SNPs (rs484071, rs4785621) caused resistance to statin drugs in cardiovascular patients (\(\beta_{SNP}\) = 0.0785785, \(\beta_{SNP}\) = 0.111714), and two SNPs (rs17502794, rs10785232) caused response to statin drugs in cardiovascular patients (\(\beta_{SNP}\) = − 0.132693, \(\beta_{SNP}\) = − 0.0957455).The deviation of the observed P-value from the null hypothesis was illustrated graphically in a QQ plot (Fig. 1B). Plotting the observed P-values versus the anticipated value from the theoretical χ distribution involves sorting each SNP’s P-values from greatest to lowest. All points occur on or close to the midline between the x and y axes if the actual values agree with the predicted values (null hypothesis: blue lines in Figs. 1B and 2B). The data revealed that in an LDL response to statins, certain observed P-values were much less than predicted by the null hypothesis, and the points moved in the x-axis direction. Because of systematic changes in allele frequencies among subpopulations in the collection of individuals tested, a significant percentage of p-values were lower than anticipated.The results illustrated how the total cholesterol (TC) response to the statin shows some observed P-values that were much higher than predicted under the null hypothesis and move the points in the direction of the y-axis (Fig. 2B). Similar to LDL response because of systematic changes in allele frequencies among subpopulations in the collection of individuals tested, a significant percentage of p-values are lower than anticipated.Functional analysisThe results captured from the GWAS Catalogue revealed that just one SNP (rs4803750) among eight candidate SNPs, has 12 reported phenotype associations (Supplmentary File 2). Our results illustrated that the phenotypes were about lipid metabolism, total cholesterol levels, high-density lipoprotein cholesterol levels, low-density lipoprotein cholesterol levels, age-related disease endophenotypes, and Alzheimer’s disease or fasting insulin levels (pleiotropy). Other seven candidate SNPs were introduced in association with a phenotype of statin responses for the first time in this study without any previous reports. The results also showed that the nearest genes with candidate SNPs are involved in several drug responses and molecular pathways based on GLAD4U and KEGG databases (Table 4). Moreover, the results of the gene ontology indicated that all candidate genes were categorized into several groups based on ‘biological process’, ‘cellular components’, and ‘molecular functions’ (Fig. 3).Table 4 Gene enrichment results.Figure 3Gene Ontology Graph. Each Biological Process, Cellular Component, and Molecular Function category is represented by a red, blue, and green bar, respectively. The height of the bar represents the number of gene IDs in the primary list and each category. The genes that were intended for lipid metabolism were located in each group.
