Method for detecting blood lipid level, method for detecting risk of developing dyslipidemia, and method for determining prevention or treatment plan for dyslipidemia

Abstract

Provided is a method for detecting a blood lipid level or a risk of developing dyslipidemia, or for determining a prevention or treatment plan for dyslipidemia, by detecting at least one lipid trait-related single nucleotide polymorphism and at least one genetic variant selected from among a group of genes consisting of LDLR, APOB, PCS, K9, LDLRAP1, and APOE in a cell collected from a subject, and detecting the level of methylation of at least one CpG site selected from among a group of genes consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1SLC1A5, and SQLE in the cell collected from the subject.

Description

Method for detecting blood lipid levels, method for detecting risk of developing dyslipidemia, and method for determining prevention or treatment policy for dyslipidemia

[0001] The present invention relates to a method for detecting blood lipid concentrations, a method for detecting the risk of developing dyslipidemia, and a method for determining a preventive or therapeutic policy for dyslipidemia, and more specifically to a method for detecting blood lipid concentrations, detecting the risk of developing dyslipidemia, or determining a preventive or therapeutic policy for dyslipidemia, based on a genetic risk score and a methylation risk score associated with lipid traits.

[0002] Dyslipidemia is a significant risk factor for atherosclerotic cardiovascular disease, including coronary artery disease (CAD), and genetic, environmental, and demographic factors interact with each other. Epidemiological studies have shown that appropriate management of dyslipidemia significantly reduces cardiovascular morbidity and mortality. Treatment of atherosclerosis often focuses on reducing low-density lipoprotein cholesterol (LDL-C), a particular type of dyslipidemia, and lifestyle modification is emphasized in all patients with elevated LDL-C. Dietary intervention is essential, particularly because high intake of saturated fatty acids increases LDL-C levels.

[0003] In addition, patients with a strong genetic predisposition, such as those with familial hypercholesterolemia (FH), require lipid-lowering drugs such as statins at an earlier stage. Clinical practice guidelines emphasize the importance of estimating absolute risk of CAD and adjusting the intensity of preventive measures accordingly. The optimal LDL-C target for low-risk individuals is set at <116 mg / dl in the 2019 ESC guidelines, whereas the previous NCEP-ATP III guidelines recommended <160 mg / dl (for risk factors 0-1). Nevertheless, there remains a significant gap between the treatments recommended by the guidelines and the everyday clinical practice of dyslipidemia.

[0004] Although an individual's lipid levels are influenced by the degree of environmental exposure and genetic predisposition, no method has been established to quantitatively characterize the combination of these factors. On the other hand, quantitatively assessing the impact of unhealthy lifestyle habits on lipid levels is difficult. Even so, exposure to external factors such as smoking and hyperlipidemia can cause long-term changes in DNA methylation in blood cells. On the other hand, genetic information as an auxiliary diagnostic tool for dyslipidemia is still under investigation, with a few exceptions, such as the genetic diagnosis of FH (Non-Patent Document 1). In this regard, polygenic risk scores (PRSs), which summarize the cumulative effects of genetic loci successfully identified through genome-wide association studies (GWAS) of lipid traits, have recently become popular (Non-Patent Documents 2, 3). Nevertheless, considerable uncertainty remains in individual PRS estimation for PRS-based risk stratification (Non-Patent Document 4), making it quite difficult to unbiasedly assess the contribution of rare variants to complex traits (Non-Patent Document 5).

[0005] Migliara G,Baccolini V,Rosso A,D‘Andrea E,Massimi A,Villari P,De Vito C.Familial Hypercholesterolemia:A Systematic Review of Guidelines on Genetic Testing and Patient Management.Front Public Health.2017;5:252Jones AC,Irvin MR,Claas SA,Arnett DK.Lipid Phenotypes and DNA Methylation:a Review of the Literature.Curr Atheroscler Rep.2021;23:71.Jhun MA,Mendelson M,Wilson R,Gondalia R,Joehanes R,Salfati E,Zhao X,Braun KVE,Do AN,Hedman AK,et al.A multi-ethnic epigenome-wide association study of leukocyte DNA methylation and blood lipids.Nat Commun.2021;12:3987.Ding Y,Hou K,Burch KS,Lapinska S,Prive F,Vilhjalmsson B,Sankararaman S,Pasaniuc B.Large uncertainty in individual polygenic risk score estimation impacts PRS-based risk stratification.Nat Genet.2022;54:30-39.Evans LM,Romero Villela PN.How rare mutations contribute to complex traits. Nature.2023;614:418-419.

[0006] As a result of intensive research conducted by the inventors to achieve the above-mentioned object, they focused on lipid traits, particularly LDL-C, and conducted genetic analysis (i.e., genotyping using GWAS arrays and rare variant searches in FH-related target genes) and epigenetic analysis (specifically, DNA methylation analysis) to develop a risk index incorporating multiple data formats for individuals with high LDL-C or CAD. They then collected individual-level data to verify the accuracy of the risk index. As a result, they found that the combination of genetic predisposition (genetic risk score) and DNA methylation (methylation risk score) has an additive effect on an individual's lipid levels, leading to improved diagnostic and predictive capabilities, and thus completed the present invention.

[0007] That is, the present invention provides the following aspects.

[0008] [1] A method for detecting blood lipid levels, comprising: (Step A) detecting at least one lipid trait-related single nucleotide polymorphism and at least one genetic variant selected from a group of genes consisting of LDLR, APOB, PCSK9, LDLRRAP1, and APOE in cells collected from a subject; (Step B) detecting a methylation level of at least one CpG site selected from a group of genes consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1, SLC1A5, and SQLE in cells collected from the subject; and (Step C-1) calculating the blood lipid levels of the subject based on the lipid trait-related single nucleotide polymorphism and the genetic variant detected in Step A and the methylation level detected in Step B. A method comprising:

[0009] [2] A method for detecting a risk of developing dyslipidemia, comprising: (Step A) detecting at least one lipid trait-related single nucleotide polymorphism and at least one genetic variant selected from a group of genes consisting of LDLR, APOB, PCSK9, LDLRRAP1, and APOE in cells collected from a subject; (Step B) detecting the methylation level of at least one CpG site selected from a group of genes consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1, SLC1A5, and SQLE in the cells collected from the subject; and (Step C-2) detecting a risk of developing dyslipidemia in the subject based on the lipid trait-related single nucleotide polymorphism and the genetic variant detected in step A, and the methylation level detected in step B.

[0010] [3] A method for determining a strategy for preventing or treating dyslipidemia, comprising: (Step A) detecting at least one lipid trait-related single nucleotide polymorphism and at least one genetic variant selected from a group of genes consisting of LDLR, APOB, PCSK9, LDLRRAP1, and APOE in cells collected from a subject; (Step B) detecting the methylation level of at least one CpG site selected from a group of genes consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1, SLC1A5, and SQLE in the cells collected from the subject; and (Step C-3) determining to administer a drug for dyslipidemia and / or provide advice on lifestyle improvement to the subject based on the lipid trait-related single nucleotide polymorphism and the genetic variant detected in step A and the methylation level detected in step B.

[0011] According to the present invention, it is possible to detect blood lipid concentrations, detect the risk of developing dyslipidemia, or determine a course of prevention or treatment for dyslipidemia based on a genetic risk score and a methylation risk score associated with lipid traits.

[0012] This figure shows an overview of the integrated genomic analysis. The integrated genomic analysis included GWAS array genotyping, targeted gene resequencing, and DNA methylation measurements at selected CpG sites. This figure shows the dot plots showing the percentage reduction in LDL-C with medication. The percentage reduction in treatment-induced LDL-C was calculated by retrospective analysis of data from 36 Japanese patients with elevated LDL-C levels who had available plasma LDL-C measurements both before and within 18 months of starting statin and ezetimibe medication. The triangles in the figure represent the mean percentage reduction for each treatment group. In European populations, the percentage reduction in treatment-induced LDL-C has been reported to range from 0.2 to >0.5, depending on the statin dose and type (Reference 15). However, a reduction of 0.3 is commonly used to impute LDL-C measurements in patients receiving statin therapy (References 25 and 26). Considering the difference between Asians and Europeans in the LDL-C lowering effect of statins (i.e., the dose is significantly lower in Asians than in Europeans (Reference 44)), we determined that a reduction rate of 0.4 with statins in Japanese patients was appropriate. PRS in the following three cohorts ＬＤＬ－Ｃ This is a histogram showing the PRS for plasma LDL-C levels in Japanese people. ＬＤＬ－Ｃ To estimate the standardized effect of ＬＤＬ－Ｃ Decile classes (also called "standard PRS of LDL-C in Japanese people") were defined. Data from three geographically separated regional cohort studies in Japan (Amagasaki, Ehime, and Kita-Nagoya [KING]) were used for this definition. Next, the PRS of the high LDL-C subgroup and the CAD subgroup (both subjects collected at NCGM) was calculated. ＬＤＬ－Ｃ To convert the PRS into a predicted LDL-C value, we used the average LDL-C measurements of subjects from the NCGM hospital cohort who belonged to the same population as these two subgroups (i.e., the external reference group). ＬＤＬ－Ｃ Box plots showing LDL-C by decile class. PRS of Japanese people in each decile combining data from the three regional population cohorts. ＬＤＬ－ＣGraph showing the LDL-C measurements in subjects from the NCGM Hospital Cohort. Boxplot showing LDL-C measurements in the high LDL-C subgroup and CAD subgroup. Histogram showing plasma LDL-C levels in two research projects: BIO-CVD (top row) and the NCGM Hospital Cohort (bottom row). BIO-CVD (N=353) is a bioresource collection project focusing on cardiovascular disease, and includes 306 CAD and 47 non-CAD subjects (see Figure 1). The NCGM Hospital Cohort (N=351) is a project that includes subjects collected separately without screening for cardiovascular disease. Arrows in the figure indicate the mean LDL-C levels for individual project subjects. PRS in the two projects (BIO-CVD (top row) and NCGM Hospital Cohort (bottom row)) ＬＤＬ－Ｃ This is a histogram showing the degree of increase in LDL-C compared to a reference group (BIO-CVD subjects without notable rare variants, N=305) estimated for three categories of five FH-associated gene variants (functional classification: disruptive, damaging, and nonsynonymous transformation). To analyze the impact of rare variants in FH-associated genes on LDL-C levels and their prevalence in the population, five FH-associated genes were targeted and resequenced as described below. For each of the three categories, the degree of increase in LDL-C compared to the reference group was estimated based on the LDL-C levels associated with rare variants identified by resequencing. This is a plot showing the correspondence between rare variants identified by resequencing of targeted genes and LDL-C levels. This is a pie chart showing the proportion of carriers of "disruptive + damaging" rare variants in Japanese populations with different disease status. The figure shows, from the left, the prevalence rates in the high LDL-C subgroup, the CAD subgroup, and the general population (TMM cohort). ＬＤＬ－Ｃ In order to predict the risk of hyper-LDL cholesterol based on PRS, a conversion formula from PRS to LDL-C values ​​was developed. First, the standardized PRS wasＬＤＬ－Ｃ To define decile classes, the data is aggregated and the PRS is used. ＬＤＬ－Ｃ and the standardized PRS ＬＤＬ－Ｃ The actual distribution of LDL-C in decile classes was obtained. Next, data from another hospital-based cohort study (NCGM Hospital Cohort) were used to convert the PRS of newly collected subjects at NCGM into LDL-C values ​​(see also Figure 3). ＬＤＬ－Ｃ Box plot showing the distribution of actual LDL-C in decile classes. ＬＤＬ－Ｃ The distribution of the standardized PRS was stratified based on the standardized decile classes of the NCGM high LDL-C subgroup (top) and CAD subgroup (bottom). ＬＤＬ－Ｃ1 is a plot showing the correlation between predicted values ​​based on decile class alone and actual LDL-C levels. 2 is a plot showing the correlation between predicted values ​​based on composite genetic risk (PRS plus rare variant genetic risk) and actual LDL-C levels in newly collected subjects with no missing data from NCGM (BIO-CVD and NCGM Biobank). 3 is a plot showing the correlation between predicted values ​​based on composite genetic risk and methylation risk score (MRS) and actual LDL-C levels in newly collected subjects with no missing data from NCGM (BIO-CVD and NCGM Biobank). 4 is a histogram showing the distribution of metabolic traits in three newly collected subgroups from NCGM (BIO-CVD and NCGM Biobank). Histograms show the distribution of lipid traits and body mass index (BMI) for three subgroups: high LDL-C (≥160 mg / dl), CAD (regardless of LDL-C level), and non-CAD (LDL-C <160 mg / dl). LDL-C was imputed for subjects taking lipid-lowering medications (statins or ezetimibe) (see Figure 2), and the imputed values ​​(second column from the left) were used for regression analysis. Arrows indicate the mean values ​​for each metabolic trait in each subgroup. Scatter plots show the correlation between DNA methylation measurements by ddPCR and EPIC array. The scatter plots demonstrate a fair correlation between the two analytical methods, ddPCR (x-axis) and EPIC array (y-axis), at 13 CpG sites. This figure shows the effect sizes of CpG-trait associations for triglycerides (left in the figure; 11 CpGs), HDL-C (center in the figure; 6 CpGs), and LDL-C (right in the figure; 2 CpGs). The top panel shows the results of comparing this example with a previous multi-ethnic EWAS. The bottom panel shows the results of comparing the high LDL-C subgroup with the CAD subgroup. For the Japanese population (high LDL-C subgroup (left in the figure), CAD subgroup (center in the figure), and general population subjects (right in the figure; KING study cohort)), the effects of the two risk scores, PRS and MRS, on lipid traits were measured using the coefficient of determination R ２10 is a graph showing the results of estimation by (the ratio of the variation in predicted values ​​to the variation in actual measured values ​​in multiple regression analysis). *P<0.05, **P<0.001, ***P<10 －４This figure shows histograms of lipid trait MRS calculated for 13 CpG sites for the high LDL-C subgroup, CAD subgroup, and general population subjects. Histograms of lipid trait MRS are shown for two NCGM subgroups (high LDL-C (≥ 160 mg / dl) and CAD subjects) and general population subjects (part of the KING study cohort). For patients taking lipid-lowering medications (statins or ezetimibe), unimputed LDL-C was used to calculate the MRS for the high LDL-C and CAD subgroups to accurately reflect their status at the time of blood sampling. Arrows indicate the mean lipid MRS for each subgroup. Intergroup comparisons of MRS between the high LDL-C subgroup and other subgroups were performed to test for significance. This figure shows scatter plots showing the correlation between predicted and actual lipid levels using the following three risk score models: The strength of correlation is compared between predicted values ​​(x-axis) and actual lipid trait values ​​(y-axis). Each lipid trait (LDL-C, HDL-C, and triglycerides) is shown in the top, middle, and bottom panels, respectively. The left, middle, and right columns plot the predicted lipid levels using each risk score model (PRS, MRS, and PRS + MRS). To convert PRS to mg / dL for lipid traits, PRS was matched to actual lipid measurements by within-subject linear regression of the relevant samples, i.e., newly collected data from NCGM (BIO-CVD + NCGM Biobank). Note that for LDL-C, the conversion method differs between the top panel (this figure) and Figures 6D-6F, as Figures 6D-6F used an additional conversion formula to incorporate rare variant genetic effects to predict LDL-C. 1 is a histogram showing the distribution of LDL-C in the general Japanese population, and a plot diagram schematically showing the correlation between predicted LDL-C values ​​and genetic risk scores in the population. 2 is a schematic diagram for explaining the process of predicting LDL-C based on MRS. 3 is a histogram showing risk classification based on the interquartile range, and a schematic diagram of a matrix table for determining a preventive or therapeutic policy for dyslipidemia by combining genetic risk scores and methylation risk scores.This figure shows an overview of combining genetic risk scores and methylation risk scores to determine a preventive or therapeutic strategy for dyslipidemia. This is a graph showing the results of analyzing the amount of change in blood lipid levels over time. Each bar graph represents, from left to right, the amount of change in blood LDL over time (ΔLDL), the amount of change in blood HDL over time (ΔHDL), and the amount of change in blood triglycerides over time (ΔTG) for each group. This is a graph showing the difference in DNA methylation rate at a CpG site (cg05119988 of SC4MOL) between two time points (before and after medication). This figure shows the prediction accuracy (coefficient of determination (R)) of blood LDL-C levels based on MRS calculated for each combination of CpG sites. ２ 1 is a graph showing the R of blood TG levels based on MRS calculated for each combination of CpG sites. ２ 1 is a graph showing

[0013] As shown in the Examples below, the inventors compared the strength of correlation between predicted and measured values ​​of lipid traits (such as blood lipid concentrations) using various risk score models to evaluate individual predictability. As a result, they found that a model incorporating a genetic risk score and a methylation risk score showed a higher correlation than a model using only one risk score, and was therefore effective in increasing predictability.

[0014] <Method for detecting blood lipid levels> Accordingly, the present invention provides a method for detecting blood lipid levels, comprising: (Step A) detecting at least one lipid trait-related single nucleotide polymorphism and at least one genetic variant selected from a group of genes consisting of LDLR, APOB, PCSK9, LDLRRAP1, and APOE in cells collected from a subject; (Step B) detecting the methylation level of at least one CpG site selected from a group of genes consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1, SLC1A5, and SQLE in the cells collected from the subject; and (Step C-1) calculating the blood lipid concentration of the subject based on the lipid trait-related single nucleotide polymorphism and the genetic variant detected in step A, and the methylation level detected in step B.

[0015] In the present invention, the "lipid" to be detected is not particularly limited. Examples include simple lipids (neutral fats, triglycerides (TG), etc.), complex lipids (phospholipids, glycolipids, etc.), derived lipids (fatty acids, cholesterol, etc.), and lipoproteins, which are complexes of apoproteins and lipids (low-density lipoproteins (LDL) and high-density lipoproteins (HDL)), etc.), but preferably at least one lipid selected from the group consisting of LDL-C, HDL-C, and TG, and particularly preferably LDL-C.

[0016] The "subject" to be subjected to the method of the present invention is not particularly limited and may be male or female. The subject may be of any age, such as a child, a young person, a middle-aged person, or an elderly person. The subject may also be a person suffering from dyslipidemia, as described below, or a person at risk of suffering from dyslipidemia, as described below. The subject may be of any race, but is preferably a non-European person, more preferably an Asian (East Asian, Southeast Asian, South Asian, North Asian, Central Asian, West Asian, or a person of any of these ethnicities), and particularly preferably a Japanese person (Yamato, Ryukyu, Ainu, or a person of any of these ethnicities (Japanese descent)).

[0017] Furthermore, the "cells" that are the target of the method of the present invention may be any cells in which any of the lipid trait-related single nucleotide polymorphisms, genetic variants, and methylation levels described below can be detected, and examples thereof include blood cells. When detecting lipid trait-related single nucleotide polymorphisms and genetic variants, oral mucosal cells may also be used as the target.

[0018] (Regarding Step A) In the present invention, "Step A" is a step of detecting DNA markers associated with lipid traits (lipid trait-associated single nucleotide polymorphisms and genetic variants described below). In this detection, genomic DNA extracted from the cells is usually the target, but mRNA extracted from the cells, cDNA or cRNA prepared from the mRNA, etc. may also be used, as long as the DNA marker can be detected. Note that those skilled in the art can appropriately select known methods suitable for the extraction and preparation of these nucleotides, taking into consideration the type and state of the cells, etc.

[0019] In the present invention, the "detection of DNA markers" is not particularly limited, and a person skilled in the art can appropriately select and perform the detection using a known DNA analysis method (particularly, a genome-wide association study (GWAS) method). Examples of such a "DNA analysis method" include next generation sequencing (NGS) and DNA microarray. More specifically, examples of the "next-generation sequencing method" include synthetic sequencing (sequencing-by-synthesis, for example, sequencing using a Solexa genome analyzer, Hiseq, or Miseq manufactured by Illumina), pyrosequencing (for example, sequencing using a GSLX or FLX sequencer manufactured by Roche Diagnostics (454) (so-called 454 sequencing)), ligase reaction sequencing (for example, sequencing using SoliD or 5500xl manufactured by Life Technologies), and ion semiconductor sequencing (for example, Ion Torrent technology by Thermo Fisher Scientific).

[0020] The "lipid trait-associated single nucleotide polymorphism," which is one of the DNA markers according to the present invention, may be a single nucleotide polymorphism (SNP) that shows a genetic association with a lipid trait, such as the blood concentration of the above-mentioned lipids. For example, as shown in the Examples below, at least one SNP can be selected from a group of SNPs selected from GWAS data by performing statistical analysis using clumping (a statistical process that removes excess SNPs from a group of closely correlated SNPs, and is expressed as the linkage disequilibrium coefficient r2) and thresholding (searching for the optimal value among a series of statistical significance levels [P-values] in observed data / simulated data) to maximize the ability to predict lipid traits (outcomes).

[0021] The "GWAS data" to be selected includes, for example, the NBDC Human Database https: / / humandbs.biosciencedbc. Examples of SNPs for each lipid trait are stored in https: / / humandbs.biosciencedbc.jp / files / hum0014 / hum0014.v8.LDL.zip for LDL-C, https: / / humandbs.biosciencedbc.jp / files / hum0014 / hum0014.v8.HDL.zip for HDL-C, and https: / / humandbs.biosciencedbc.jp / files / hum0014 / hum0014.v8.TG.zip for triglycerides.

[0022] Specific examples of single nucleotide polymorphisms associated with LDL-C traits include at least one SNP selected from the 5,194 SNPs shown in Tables 1 to 17 below.

[0023]

[0024]

[0025]

[0026]

[0027]

[0028]

[0029]

[0030]

[0031]

[0032]

[0033]

[0034]

[0035]

[0036]

[0037]

[0038]

[0039]

[0040] In each column of Tables 1 to 17, the rs number, coding allele, and β value of each SNP are shown from left to right. The β value is the value by which the number of coding alleles for each SNP is multiplied in the PRS calculation described below.

[0041] The number of lipid trait-related single nucleotide polymorphisms detected by the method of the present invention may be at least 1, but may be, for example, 100 SNPs or more, 200 SNPs or more, 300 SNPs or more, 400 SNPs or more, 500 SNPs or more, 600 SNPs or more, 700 SNPs or more, 800 SNPs or more, 900 SNPs or more, 1000 SNPs or more, 2000 SNPs or more. Examples of SNPs include 3000 SNPs or more, 3000 SNPs or more, 4000 SNPs or more, 5000 SNPs or more, 6000 SNPs or more, 7000 SNPs or more, 8000 SNPs or more, 9000 SNPs or more, 10000 SNPs or more, 20000 SNPs or more, 30000 SNPs or more, 40000 SNPs or more, 50000 SNPs or more, or 60000 SNPs or more.

[0042] As shown in the Examples below, the "genetic variant," which is one of the DNA markers according to the present invention, may be at least one genetic variant selected from five FH-related genes (LDLR, APOB, PCSK9, LDLRAP1, and APOE), and the type of variant is not particularly limited, and examples include disruptive variants (frameshift or splice donor), damaging variants, and nonsynonymous variants. More specifically, examples include at least one variant selected from the 2 disruptive variants, 13 damaging variants, and 43 nonsynonymous variants shown in Tables 23 to 25 below.

[0043] The number of genetic variants detected by the method of the present invention may be at least one variant, but may also be, for example, 5 or more, 10 or more, 20 or more, 30 or more, 40 or more, or 50 or more (51, 52, 53, 54, 55, 56, 57, or 58) variants.

[0044] (Regarding Step B) In the present invention, Step B is a step of detecting a methylation marker associated with lipid traits (methylation level of a CpG site, which will be described later). In this detection, genomic DNA extracted from the cells is usually the target, and a person skilled in the art can appropriately select a known method suitable for extraction and preparation of such genomic DNA, taking into consideration the type and state of the cells, etc.

[0045] In the present invention, the "detection of a methylation marker" is not particularly limited, and a person skilled in the art can appropriately select and perform the detection using a known DNA methylation analysis method. Examples of such "DNA methylation analysis methods" include bead array methods (e.g., Infinium (registered trademark) assay), DNA methylation analysis methods using a mass spectrometer (e.g., MassARRAY (registered trademark), see Jurinke C et al., Mutat Res, 2005, vol. 573, pp. 83-95), pyrosequencing methods (see Anal. Biochem. (2000) 10: 103-110), methylation-sensitive high-resolution melting curve analysis (MS-HRM, see Wojdacz TK et al., Nat Protoc., 2008, vol. 3, pp. 1903-1908), methylation-specific quantitative PCR methods (MS-PCR) such as the MethyLight method using TaqMan probes (registered trademark), bisulfite direct sequencing, bisulfite cloning sequencing (see Kristensen et al., J. Immunol. 2004, vol. 10, pp. 1903-1908), and the like. LS et al., Clin Chem, 2009, vol. 55, pp. 1471-83), and COBRA (analysis using a combination of bisulfite and restriction enzymes).

[0046] As shown in the Examples below, the "methylation marker" according to the present invention may be any methylation level of at least one CpG site selected from the group of genes consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1, SLC1A5, and SQLE, and specifically, 13 CpG sites shown in Table 22 (cg06500161 of ABCG1, cg179 of DHCR24, etc.) 01584, cg10073091 of DHCR24, cg05119988 of SC4MOL, cg03725309 of SARS, cg14476101 of PHGDH, cg19693031 of TXNIP, cg06690548 of SLC7A11, cg21429551 of GARS, cg00574958 of CPT1A, cg11024682 of SREBF1, cg02650017 of PHOSPHO1 and cg02711608 of SLC1A5), and at least one methylation level selected from SQL cg09984392.

[0047] Furthermore, as will be described in the Examples below, when the target lipid trait is LDL-C, it is preferable to detect the methylation level of at least one CpG site selected from the CpG site of DHCR24 (e.g., cg10073091 of DHCR24), the CpG site of SC4MOL (e.g., cg05119988 of SC4MOL), and the CpG site of SQLE (e.g., cg09984392 of SQLE), and it is preferable to detect the methylation level of at least the CpG site of DHCR24 (e.g., cg10073091 of DHCR24). is more preferred, and detecting the methylation levels of at least the CpG site of DHCR24 (e.g., cg10073091 of DHCR24) and the CpG site of SC4MOL (e.g., cg05119988 of SC4MOL), or detecting the methylation levels of at least the CpG site of DHCR24 (e.g., cg10073091 of DHCR24), the CpG site of SC4MOL (e.g., cg05119988 of SC4MOL), and the CpG site of SQLE (e.g., cg09984392 of SQLE) is even more preferred.

[0048] As will be described in the Examples below, when the lipid trait of interest is HDL-C, it is preferable to detect the methylation level of at least one CpG site selected from the CpG site of ABCG1 (e.g., cg06500161 of ABCG1), the CpG site of DHCR24 (e.g., cg17901584 and / or cg10073091 of DHCR24), the CpG site of SC4MOL (e.g., cg05119988 of SC4MOL), the CpG site of SREBF1 (e.g., cg11024682 of SREBF1), and the CpG site of PHOSPHO1 (e.g., cg02650017 of PHOSPHO1).

[0049] As shown in the Examples below, when the target lipid trait is TG, the CpG site of ABCG1 (e.g., cg06500161 of ABCG1), the CpG site of DHCR24 (e.g., cg17901584 of DHCR24), the CpG site of SC4MOL (e.g., cg05119988 of SC4MOL), the CpG site of SARS (e.g., cg03725309 of SARS), the CpG site of PHGDH (e.g., cg14476101 of PHGDH), the CpG site of TXNIP (e.g., cg19693031 of TXNIP), the CpG site of SL It is preferable to detect the methylation level of at least one CpG site selected from the CpG site of C7A11 (e.g., cg06690548 of SLC7A11), the CpG site of GARS (e.g., cg21429551 of GARS), the CpG site of CPT1A (e.g., cg00574958 of CPT1A), the CpG site of SREBF1 (e.g., cg11024682 of SREBF1), and the CpG site of SLC1A5 (e.g., cg02711608 of SLC1A5), and at least the CpG site of TXNIP (e.g., cg02711608 of TXNIP). It is more preferable to detect the methylation level of at least the CpG site of TXNIP (e.g., cg19693031 of TXNIP) and SLC1A5 (e.g., cg02711608 of SLC1A5), and it is even more preferable to detect the methylation level of at least the CpG site of TXNIP (e.g., cg19693031 of TXNIP), SLC1A5 (e.g., cg02711608 of SLC1A5), and ABCG1 (e.g., cg06500161 of ABCG1), or at least It is more preferable to detect the methylation levels of at least the CpG sites of TXNIP (e.g., cg19693031 of TXNIP), SLC1A5 (e.g., cg02711608 of SLC1A5), ABCG1 (e.g., cg06500161 of ABCG1), and SREBF1 (e.g., cg11024682 of SREBF1).It is more preferable to detect the methylation levels of SREBF1 (e.g., cg11024682 of SREBF1) and GARS (e.g., cg21429551 of GARS), and at least the CpG sites of TXNIP (e.g., cg19693031 of TXNIP), SLC1A5 (e.g., cg02711608 of SLC1A5), ABCG1 (e.g., cg06500161 of ABCG1), SREBF1 (e.g., cg11024682 of SREBF1), GARS (e.g., It is more preferable to detect the methylation levels of CpG sites of at least TXNIP (e.g., cg19693031 of TXNIP), SLC1A5 (e.g., cg02711608 of SLC1A5), ABCG1 (e.g., cg06500161 of ABCG1), SREBF1 (e.g., cg11024682 of SREBF1), GARS (e.g., cg21429551 of GARS), and CPT1A (e.g., cg00574958 of CPT1A). It is more preferred to detect the methylation levels of ABCG1 (e.g., cg06500161 of ABCG1), CPT1A (e.g., cg00574958 of CPT1A), and SC4MOL (e.g., cg05119988 of SC4MOL), and also to detect the methylation levels of at least ABCG1 (e.g., cg06500161 of ABCG1), DHCR24 (e.g., cg17901584 of DHCR24), SC4MOL (e.g., cg05119988 of SC4MOL), SARS (e.g., cg03725309 of SARS), PHGDH ( For example, it is preferable to detect the methylation levels of PHGDH cg14476101), TXNIP (e.g., TXNIP cg19693031), SLC7A11 (e.g., SLC7A11 cg06690548), GARS (e.g., GARS cg21429551), CPT1A (e.g., CPT1A cg00574958), SREBF1 (e.g., SREBF1 cg11024682), and SLC1A5 (e.g., SLC1A5 cg02711608).

[0050] Furthermore, the number of methylation markers detected by the method of the present invention may be the methylation level of at least one CpG site, but may also be, for example, the methylation levels of 2 or more, 3 or more, 4 or more, 5 or more, 6 or more, 7 or more, 8 or more, 9 or more, 10 or more, 11 or more, 12 or more, or 13 CpG sites.

[0051] (Regarding Step C-1) As will be described in Examples below, in the method of the present invention, it is possible to calculate (predict) the blood lipid concentration of the subject based on the lipid trait-related single nucleotide polymorphism and the genetic variant detected in Step A and the methylation level detected in Step B.

[0052] As a specific example, first, a polygenic risk score (PRS) is calculated from the lipid trait-related single nucleotide polymorphisms detected in step A using the following formula:

[0053]

[0054] β is the partial regression coefficient of linear regression or logistic regression, X is the number of alleles (0, 1, 2) at each SNP detected in the subject, and n is the number of SNPs. The specific value of β at each SNP can be found, for example, by referring to the NBDC Human Database https: / / humandbs.biosciencedbc.jp / en / hum0014. Furthermore, when targeting LDL-C, the β values ​​at each SNP shown in Tables 1 to 17 can be used.

[0055] Next, based on the PRS calculated in this way, the lipid concentration can be determined as shown in the Examples below. There are roughly two types of methods for determining this.

[0056] One method is to calculate a scaling factor by linear regression using a PRS model, similar to the MRS described below, and calculate lipid values ​​based on the PRS for each subject. This method can be used when the analysis population is selected without bias for the outcome of interest.

[0057] The other is a method of converting PRS into an absolute scale (absolute scale of lipid measurement values) based on the distribution of the mean and standard deviation of the outcome (continuous variable called blood lipid concentration) corresponding to the PRS quantile in a population (see Pain et al. Eur J Hum Genet. 2022, PMID 34983942). When applying this conversion method, an appropriate reference population is required to calculate the mean value for each PRS quantile. If the size of this reference population is sufficiently large, the number of quantiles can be increased (for example, in the case of percentiles, separation into 1% units) to confirm the distribution of the mean values ​​in detail and convert them into an absolute scale.

[0058] More specifically, for example, decile classes (strata) can be set according to the PRS threshold values ​​shown in Table 18 below, and the lipid concentration (average value) corresponding to each strata can be calculated.

[0059]

[0060] In the present invention, the lipid concentration calculated based on PRS in this manner may be corrected based on the genetic variants. For example, as shown in the Examples below, when the target lipid is LDL-C, such correction may involve adding a correction value for each genetic variant category (disruptive variant: +106 mg / dl, damaging variant: +42 mg / dl, nonsynonymous variant: +13 mg / dl) to the lipid concentration calculated based on PRS.

[0061] In the present invention, the genetic risk score (GRS) for blood lipid concentrations differs between TG and HDL-C and LDL-C depending on the presence or absence of genetic variant information. That is, the PRS itself is used for TG and HDL-C, while a "composite (fused) genetic risk score" based on the lipid trait-related single nucleotide polymorphisms and the genetic variants can be used for LDL-C. However, genetic variant information can also be collected for TG and HDL-C, and in such cases, a fused genetic risk score can be utilized, similar to LDL-C. In the present invention, the term "GRS" is used in a broad sense, including both the use of the PRS itself and the use of the fused genetic risk score. Below, an example of a fused genetic risk score for LDL-C, which has attracted the most clinical attention from the perspective of arteriosclerosis, is described.

[0062] First, as per usual, PRS was obtained from the GWAS array data on LDL-C. ＬＤＬ－Ｃ As shown in Figure 6A, as the population size increases, the PRS ＬＤＬ－Ｃ The histogram of PRS approaches a normal distribution. If we transform this so that the X axis represents percentile and the Y axis represents PRS, we get a distribution of the Provid function (shaped like an "S" flipped 180 degrees horizontally), as shown in Figure 3C. ＬＤＬ－Ｃ Although there is considerable variability in the data due to factors other than PRS (including the genetic variants mentioned above and environmental exposures), ＬＤＬ－Ｃ Since LDL-C is positively correlated with IL-1 (see Figures 3B and 3D), the PRS on the Y-axis is hypothetically ＬＤＬ－Ｃ It is possible to replace the above with predicted LDL-C values. The LDL-C distribution in a population is influenced by various factors, such as age, sex, and habits (lifestyle), but as a reference population, for example, the 2015 National Health and Nutrition Survey data can be used to represent the general Japanese population. If predicted LDL-C values ​​are generally close to this survey data, the first quartile is estimated to be 100 mg / dl and the third quartile is estimated to be 140 mg / dl.

[0063] For genetic variant carriers, the individual PRS ＬＤＬ－ＣThe LDL-C estimate based on the fused genetic risk score can be calculated by adding the correction values ​​for each of the above-mentioned genetic variant categories to the LDL-C estimate. A simple method is to consider individuals with an LDL-C estimate above the third quartile (≥ 140 mg / dL) as being at high risk for the genetic risk score. As shown in Figure 12, even if the PRS itself is medium risk (between the first and third quartiles), adding the correction value for a rare genetic variant may shift the percentile of the fused genetic risk score to high risk. If a reference population with detailed quantile information is available, the predicted LDL-C value with the correction value for the genetic variant added can also be displayed as a percentile.

[0064] Furthermore, in the present invention, the lipid concentration calculated based on the DNA marker may be corrected based on the methylation level. As an example of such correction, a methylation risk score (MRS) calculated using a weighted sum of beta values ​​at each CpG site may be added to the lipid concentration calculated based on the DNA marker, as shown in the Examples below.

[0065] More specifically, first, for each CpG site, the beta value is standardized across the entire cohort population, and then a weighted sum of the standardized beta value and weighting (effect size BETA expressed as a Z-score) is calculated for each individual subject. Note that the "beta value" here is an index representing the proportion of methylation occurring at a specific CpG site in a subject sample in DNA methylation analysis. Although confusingly similar in name, it is different from the BETA used for weighting in the weighted sum. Next, this weighted sum is standardized across the entire cohort population to obtain the MRS prototype. Finally, the standardized MRS prototype is multiplied by a scaling factor to convert it to mg / dL units, resulting in the MRS. This scaling factor can be obtained, for example, by linear regression, as shown in Figure 13, and is exemplified as 1.83 for LDL-C, 4.33 for HDL-C, and 19.0 for TG. The MRS calculated in this manner is expressed in units of mg / dL as a lipid measurement value, and is adjusted so that the average is zero.

[0066] <Method for detecting risk of developing dyslipidemia> The present invention also provides a method for detecting a risk of developing dyslipidemia, comprising: (Step A) detecting at least one lipid trait-related single nucleotide polymorphism and at least one genetic variant selected from a group of genes consisting of LDLR, APOB, PCSK9, LDLRRAP1, and APOE in cells collected from a subject; (Step B) detecting a methylation level of at least one CpG site selected from a group of genes consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1, SLC1A5, and SQLE in the cells collected from the subject; and (Step C-2) detecting a risk of developing dyslipidemia in the subject based on the lipid trait-related single nucleotide polymorphism and the genetic variant detected in step A, and the methylation level detected in step B.

[0067] Steps A and B in this method are as described above. In step C-2, for example, the risk of developing dyslipidemia can be detected based on the blood lipid concentration detected in step C-1 described above, as follows.

[0068] Here, "dyslipidemia" according to the present invention can also be referred to as abnormal blood lipid concentrations or hyperlipidemia (other than HDL-C). For example, as shown below (based on the 2022 edition of the Japan Atherosclerosis Society's Atherosclerotic Disease Prevention Guidelines), states above or below the thresholds set for the blood concentrations of various lipid traits can be cited. When the blood concentrations approach or fall outside these thresholds, it can be determined that the patient is at high risk of developing dyslipidemia or already has the condition. LDL-C: 140 mg / dl or higher (120 to 139 mg / dl is the borderline level), HDL-C: less than 40 mg / dl, TG: 150 mg / dl or higher in fasting blood draw, or 175 mg / dl or higher in casual blood draw.

[0069] Furthermore, such abnormalities in blood lipid levels, combined with other risk factors (smoking, high blood pressure, diabetes, etc.), lead to the onset of arteriosclerotic diseases. Therefore, dyslipidemia in the present invention includes arteriosclerotic diseases. Examples of such "arteriosclerotic diseases" include coronary artery disease (CAD), ischemic heart disease (angina pectoris, myocardial infarction, etc.), cerebrovascular disorders (cerebral infarction, etc.), and arteriosclerosis obliterans.

[0070] <Method for determining a strategy for preventing or treating dyslipidemia> The present invention also provides a method for determining a strategy for preventing or treating dyslipidemia, comprising: (Step A) detecting at least one lipid trait-related single nucleotide polymorphism and at least one genetic variant selected from a group of genes consisting of LDLR, APOB, PCSK9, LDLRRAP1, and APOE in cells collected from a subject; (Step B) detecting the methylation level of at least one CpG site selected from a group of genes consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1, SLC1A5, and SQLE in the cells collected from the subject; and (Step C-3) and determining to administer a drug for dyslipidemia and / or provide advice on lifestyle improvement to the subject based on the lipid trait-related single nucleotide polymorphism and the genetic variant detected in step A and the methylation level detected in step B.

[0071] (Regarding Step C-3) Step A and Step B in the method are as described above. In Step C-3, for example, a strategy for preventing or treating dyslipidemia can be determined based on a genetic risk score (GRS) calculated based on the lipid trait-related single nucleotide polymorphism and the genetic variant, and a methylation risk score (MRS) calculated based on the methylation level.

[0072] More specifically, for example, by using a matrix table combining the risk scores as a diagnostic support tool, the degree of contribution of constitution and poor lifestyle habits to arteriosclerotic diseases such as dyslipidemia can be estimated, thereby standardizing and refining policy decisions (Figure 14). A high GRS, which reflects constitution, indicates a high need for drug treatment, while a high MRS, which reflects poor lifestyle habits, indicates a high need for lifestyle improvement. Furthermore, the distribution of each risk score within a population is broadly classified into a "low risk group" below the first quartile, a "medium risk group" within the interquartile range (from the first quartile to the third quartile), and a "high risk group" above the third quartile. Specific examples of the GRS and MRS ranges for the "low risk group," "medium risk group," and "high risk group" for each lipid trait are shown in Table 19.

[0073]

[0074] The risk classification "high" is above the 75th percentile (third quartile), "medium" is between the 25th percentile (first quartile) and the 75th percentile, and "low" is below the 25th percentile. * The genetic risk scores in Table 19 use only PRS, and the PRS ranges are based on the combined data of three regional populations (Figure 3C). ** The MRS ranges are based on the two subgroups in Figure 10 [the middle row (CAD subjects) and the bottom row (KING subjects)].

[0075] As shown in Figure 15 , when explaining test results during a medical checkup or physical examination, based on the subject's risk score, those with a low GRS risk and a high MRS risk are initially recommended to make lifestyle modifications, while those with a high GRS risk and a low MRS risk are recommended to start drug treatment early. For those who do not belong to these two unbalance groups (those with two risk scores at opposite ends of the spectrum), lifestyle modifications are first recommended, but if these are ineffective, additional drug treatment is considered. Treatment goals are set in accordance with clinical guidelines, depending on the coexistence of atherosclerosis risk factors. Similarly, as shown in Figure 15 , during outpatient follow-up of individuals undergoing drug treatment, the balance between lifestyle modifications and drug treatment is evaluated based on the subject's risk score. While GRS remains constant throughout life and only needs to be measured once, MRS is variable and must be measured as needed.

[0076] More specifically, the drug treatment according to the present invention involves the administration of a drug for dyslipidemia. Examples of drugs for dyslipidemia include cholesterol-lowering agents: statin preparations (HMG-CoA reductase inhibitors, for example, Mevalotin, Lipovas, Livalo, Lipitor, Crestor, Lochol), small intestinal cholesterol transporter inhibitors (for example, ezetimibe), anion exchange resin preparations (for example, Questran, Cholebain), etc. Triglyceride-lowering agents: fibrate preparations (for example, Vinoglac, Bezatol SR, Lipocrine, Tricor), EPA preparations (for example, Epadel, Lotriga), sulfated polysaccharide preparations (dextran sulfate), etc. Cholesterol and triglyceride-lowering agents: nicotinic acid derivative preparations (for example, Yubela N, Cholexamin, Pericit), etc. HDL-increasing agents: PPARα agonists such as Palmodia, polyunsaturated fatty acid preparations (polyene phosphatidylcholine), vitamin-related substances (pantothenic acid), Bezatol SR, Tricor, Pantosin, etc.

[0077] Furthermore, the lifestyle habit correction guidance according to the present invention includes, for example, weight control guidance, dietary guidance [for example, control of lipid intake (reducing intake of trans fatty acids, saturated fatty acids and cholesterol, promoting intake of unsaturated fatty acids), intake of dietary fiber and antioxidants, and drinking in moderation], exercise guidance, and smoking cessation guidance.

[0078] The present invention will be described in more detail below based on examples, but the present invention is not limited to the following examples. The examples were carried out using the following <Materials and Methods>.

[0079] Materials and Methods (Study Subjects) The Institutional Review Board approved this study, and participants provided written informed consent. Procedures complied with the ethical standards of the Institutional Committee on Human Research at the National Center for Global Health and Medicine (NCGM). At NCGM, Japanese adult patients were recruited as a target population through two separate projects—BIO-CVD and NCGM Biobank—and were classified into three subgroups based on LDL-C levels and CAD status: a high LDL-C group without CAD, a CAD group, and a non-CAD group with an LDL-C level <160 mg / dL (Table 20). A hospital-based study called BIO-CVD evaluated the CAD status of 353 participants (306 with CAD and 47 without CAD) (Figure 1). CAD was defined as previously reported (Reference 31). Samples from the NCGM Biobank were also examined, particularly those with elevated plasma LDL-C (≥160 mg / dl) (N = 300). These samples were used for genetic and DNA methylation analysis.

[0080]

[0081] The samples in Table 20 were derived from either the BIO-CVD project or the NCGM Biobank and were all used for resequencing of five target genes associated with familial hypercholesterolemia. CAD: coronary artery disease. LDL-C: low-density lipoprotein cholesterol. * The smoking status of two subjects in the CAD group was unknown. ** Hypertension and diabetes were defined as those prescribed antihypertensive and antidiabetic medications, respectively. High LDL cholesterolemia was defined as medical records showing a basal LDL-C ≥ 140 mg / dl (irrespective of treatment with lipid-lowering drugs, statins, and / or ezetimibe). *** Subjects in the high LDL-C group did not have obvious CAD, although detailed examinations (e.g., coronary angiography) were not performed in the majority of cases.

[0082] Additionally, we included a significant number of Japanese individuals from non-target populations. For low-frequency (or rare) genetic variants, we used publicly available whole-genome sequencing data (N = 8.3K). For common single-nucleotide polymorphisms (SNPs), we used GWAS array genotyping data from three population-based cohorts (including the Kitanagoya Genomic Epidemiology [KING] Study Cohort) and another hospital-based sample from NCGM (named the NCGM Hospital Cohort; N = 351) (Reference 17) (Table 21, Characteristics of Subjects in the Reference Group). Furthermore, for DNA methylation, we utilized epigenome-wide association study (EWAS) assay data from a subset of previously reported population-based cohort samples (N = 314) (Reference 14).

[0083]

[0084] In Table 21, *Dyslipidemia is often diagnosed as hyper-LDL cholesterolemia. Hyper-LDL cholesterolemia is defined as a basal LDL-C ≥ 140 mg / dl (irrespective of treatment with lipid-lowering drugs, statins, and / or ezetimibe) as documented by medical records. Detailed information on statin use was not recorded in the Kitanagoya and Amagasaki study cohorts.

[0085] (Imputation of Baseline LDL-C Levels) To impute baseline (or pretreatment) LDL-C levels (i.e., to correct for unusable data items), a retrospective analysis was performed using data from 36 individuals with high LDL-C levels who had plasma LDL-C measurements available before and within 18 months of starting statin and ezetimibe treatment (Figure 2). Similar to the study by Ruel et al. (Reference 15), baseline LDL-C levels were imputed by dividing the treatment (LDL-C) value by the expected treatment reduction (expressed as a ratio) minus 1.

[0086] (Gene Sequencing) Rare variants in five FH-related genes (LDLR, APOB, PCSK9, LDLRRAP1, and APOE) were screened by next-generation sequencing analysis using the KAPA HyperChoice Kit (KAPA BioSystems, Wilmington, MA, USA) on an Illumina MiSeq or NextSeq500 system. This resulted in coverage of target gene variants in the coding DNA sequence, exon-intron junctions, and untranslated regions of the target genes at a median read depth of 50x. Over 90% of the reads exceeded 20x.

[0087] Sequences were aligned to the human genome reference sequence (GRCh37 / hg19). SNPs and insertion / deletion (Indel) variants were called using the software GATK (https: / / gatk.broadinstitute.org / ) and McCortex (https: / / github.com / mcveanlab / mccortex) and annotated using VarSeq (https: / / www.goldenhelix.com / products / VarSeq / ). Next, we searched for rare, putative pathogenic variants, including SNPs that cause nonsynonymous, nonsense, or splice site substitutions and are predicted to be disruptive or damaging by the in silico prediction algorithms of ClinVar, LOVD, or dbNSFP (Reference 16).

[0088] Genotyping: 644 samples were de novo genotyped using Infinium OmniExpress-24 BeadChips (Illumina, San Diego, CA, USA). GWAS array genotyping data from a reference population (panel) was also used. Data cleaning and analysis were performed using PLINK (version 1.07). Genotype imputation for the 1000 Genomes Phase 3 set (a statistical method for estimating genotypes of regions not included in SNP arrays, etc., using the same terminology, but different from the previous description) was performed using Minimac3 software (version 2.0.1) as previously reported (Reference 17).

[0089] (Polygenic Risk Score (PRS)) For each lipid trait, a PRS was constructed using LDpred (https: / / github.com / bvilhjal / ldpred) (Reference 18). Briefly, for each block (a cluster of genomic regions) based on linkage disequilibrium (LD) on the genome, index SNPs that passed a certain threshold of statistical significance (P<0.005) and LD coefficient (r2≧0.5) were extracted from a large-scale GWAS reported in Japanese people (Reference 19). As a result, 5194 to 6616 SNPs were selected for PRS calculation. The number of risk alleles (0, 1, 2) was then weighted by the β value of the association between the SNP and lipids and scored to calculate LDL-C (PRS). ＬＤＬ－Ｃ ) and other lipid traits were used to develop the best polygenic predictors.

[0090] (Conversion of PRS to LDL-C value) PRS to plasma LDL-C value ＬＤＬ－Ｃ To estimate the standardization effect of PRS, we developed a formula to convert PRS to LDL-C values ​​using two different external reference panels. First, we used the standardized PRS in the general Japanese population. ＬＤＬ－ＣDecile classes (strata) were set and used as the standard PRS for LDL-C in Japanese people (Fig. 3A-C). Next, the mean LDL-C values ​​for the standard PRS decile strata for Japanese people were calculated for subjects in the NCGM Hospital cohort, which has no selection bias for cardiovascular disease (Table 21 and Fig. 4). The PRS values ​​for BIO-CVD and NCGM Biobank samples (Fig. 6D and Fig. 3E), which belong to the same population as the NCGM Hospital cohort, were calculated. ＬＤＬ－Ｃ To convert, the mean LDL-C values ​​(shown in Figure 3D) for each PRS decile (of the NCGM hospital cohort) were used.

[0091] Annotation of Rare Variants: We searched for rare [minor allele frequency (MAF) < 0.01], putatively pathogenic variants, including SNVs that cause nonsynonymous, nonsense, or splice site substitutions and are predicted to be lethal by in silico prediction algorithms (SIFT, Polyphen2, MutationTaster, MutationAssessor, FATHMM, and FATHMM-MKL) in ClinVar, LOVD, or dbNSFP (Reference 16). Rare variants that are putatively functionally important were arbitrarily classified into three distinct categories: (1) disruptive variants (frameshift or splice donor), (2) damaging variants, and (3) other nonsynonymous variants. This is similar to previous studies (Reference 20), with some modifications. In particular, damaging variants were defined as missense variants that were annotated as pathogenic / likely pathogenic in reputable databases such as ClinVar, or as disruptive or damaging by four or more (out of six) predictive algorithms in dbNSFP, in addition to nominal significance (P<0.05) in the comparison of MAF between case and reference groups (by chi-square test).

[0092] Fusion Genetic Risk As previously described, rare (MAF<0.01) variants predicted to be functionally important were arbitrarily classified into three distinct categories: (1) disruptive variants (frameshift or splice donor), (2) damaging variants, and (3) other nonsynonymous variants.

[0093] Predicted value from actual measurement value (PRS ＬＤＬ－Ｃ The net increase in LDL-C levels in rare variant carriers was estimated by subtracting the difference (based on the variance estimate). A linear regression model was used to calculate the change in LDL-C levels and its standard error (SE) from those without rare variants in FH-related genes, for categories including disruptive variants, damaging variants, and other nonsynonymous variants.

[0094] To combine the genetic risks of rare variants and lipid trait-associated single nucleotide polymorphisms, the following algorithm was used: [individual LDL-C level] = [PRS ＬＤＬ－Ｃ Predicted LDL-C levels based on [predicted LDL-C levels based on rare variants] + [estimated change from non-carriers (by rare variant category)].

[0095] (Selection of Target CpGs) A literature search for lipid-related EWAS was conducted to generate an initial list of target CpG methylation sites that showed strong and reproducible associations with lipid-related phenotypes, including lipoprotein subfractions (References 32 and 33), plasma lipids (References 34-39), postprandial hyperlipidemia (Reference 40), statin use (Reference 41), and metabolic syndrome components (Reference 42). Here, cg12556569 of the APOA5 gene was excluded from the target list because the probe signal for measuring DNA methylation showed a dense distribution due to the influence of a SNP (rs10750097, a variant relatively common in both Europeans and Asians) in the probe (Reference 43).

[0096] (Methylation Risk Score (MRS)) Genomic DNA was extracted from peripheral blood buffy coat samples and stored at -80°C before DNA methylation analysis. Genome-wide methylation profile analysis was performed using an Illumina EPIC array (Illumina, Inc.) as previously reported (Reference 21). Additionally, methylation levels of 13 CpG sites for which convincing evidence of association with lipid traits has been reported in previous EWAS studies were measured using a droplet digital PCR (ddPCR) assay using a QX200 Droplet Digital PCR system (Bio-Rad, Carlsbad, CA, USA) (Reference 9). These CpG-trait associations are not limited to lipids, and have been reported to overlap significantly with other cardiometabolic phenotypes, such as blood pressure, alcohol intake, and liver fat content (overlap between DNA methylation and cardiometabolic phenotypes at CpG sites in Tables 22 and 13).

[0097]

[0098] In Table 22, the positive EWAS hits (P<5E-8) are taken from the EWAS catalog (http: / / www.ewascatalog.org) as of August 30, 2023. Positive signals are arbitrarily divided into three classes according to p-value: triangle (Δ), 5E-8>P≧1E-10; circle (○), 1E-10>P≧1E-20; double circle (◎), P<1E-20.

[0099] Because methylation profiles reflect environmental influences, MRSs for lipid traits were constructed using a weighted sum of methylation beta values ​​from 13 CpG sites (Reference 22). Here, effect sizes from a previous multi-ethnic EWAS for lipid traits were used as external weights for each CpG probe. More specifically, weights were determined as effect sizes expressed as Z-scores from a previous multi-ethnic EWAS for lipid traits [SuppData2, MODEL4.ALL from Jhun et al. (Reference 9)]. For each CpG site, methylation beta values ​​were standardized across the entire study population. These standardized values ​​were weighted and summed to derive the raw MRS for each individual. Further standardization was then performed to obtain the standardized MRS. To convert the standardized MRS to mg / dL, linear regression was used to match the standardized MRS to lipid measurements. As a result, the scaling (adjustment for conversion) coefficients were 1.83 for LDL-C, 4.33 for HDL-C, and 19.0 for TG, respectively. The MRS calculated in this way is expressed in mg / dL of lipid measurement value and adjusted to have an average of 0. Therefore, in the prediction of lipid concentrations based only on MRS, the average value of the lipid measurement value in the analyzed population was added to the MRS for calculation.

[0100] (Statistical Analysis) Unless otherwise stated, significant results are expressed as mean ± SE, P < 0.05. Association analysis was performed using a linear regression model. When analyzing LDL-C, imputation values ​​were used for subjects receiving lipid-lowering therapy as follows:

[0101] The results obtained using the above <Materials and Methods> are shown below.

[0102] Imputation of Baseline LDL-C Values ​​In this example, adult patients with hyper-LDL-cholesterolemia and CAD were recruited from NCGM. A significant proportion of participants were prescribed statins, with the proportions being 19% in the high LDL-C subgroup and 70% in the CAD subgroup (Table 20). Regardless of the type and dose of statins, the mean reduction in LDL-C from baseline (or pretreatment) was approximately 40%. Furthermore, ezetimibe reduced LDL-C by 22% (observed in one patient) (Figure 2). Therefore, these values ​​(40% for statins and 20% for ezetimibe) were used to estimate pretreatment LDL-C measurements for patients receiving lipid-lowering therapy.

[0103] <Rare variants in five FH-related genes> We identified 1,762 variants in five FH-related genes, of which 58 were functionally significant rare variants (MAF<0.01). These variants (30 in APOB, 13 in LDLR, 7 in PCSK9, 1 in LDLRAP1, and 7 in APOE) were classified into three categories according to the criteria described above: two disruptive variants (frameshift and splice donor), 13 damaging variants, and 43 other non-synonymous variants (Table 23 (list of functional variants identified by targeted resequencing of FH-related genes) and Tables 24 and 25 (list of non-synonymous variants identified by targeted resequencing of FH-related genes)). A gain-of-function variant of PCSK9 (indel, TGCCAGCGCCT / -) was also identified in CAD patients, which appeared to be protective against CAD by exerting an LDL-C lowering effect (Table 23).

[0104]

[0105] In Table 23, a) Ontology with highest priority for variant-transcript interactions. The terminology used is the standard functional description given by the Sequence Ontology Project. b) Clinical significance is the consensus interpretation of submissions for all pathologies for this variant. Primarily based on ACMG classification: P, pathogenic; LP, likely pathogenic; VUS, variant of uncertain clinical significance; LB, likely benign; B, benign. c) Variant effect prediction tools tested are SIFT, PolyPhen2, Mutation Taster, Mutation Assessor, FATHMM, and FATHMM-MKL. d If the FATHMM-MKL coding score (PMID: 25583119) is >0.5 (or rank score >0.28317), the corresponding nonsynonymous SNP is predicted as "damaging type," otherwise it is predicted as "tolerant type." e The Association for Clinical Genomics Sciences (ACGS) classification and interpretation of the variant is shown where applicable. For abbreviations, see the ACMG classification above. f PopMax indicates the gnomAD subpopulation with the highest allele frequency. *One submission (SCV000503122.1) indicates LB, while other submissions claim this variant is pathogenic with either P (6 submissions) or LP (3 submissions).

[0106]

[0107]

[0108] In Tables 24 and 25, a) The ontology with the highest priority for variant-transcript interactions. The terminology used is the standard functional description given by the Sequence Ontology Project. b) The clinical significance is the consensus interpretation of submissions for all pathologies for this variant. Primarily based on ACMG classification: P, pathogenic; LP, likely pathogenic; VUS, variant of uncertain clinical significance; LB, likely benign; B, benign. c) The variant effect prediction tools used were SIFT, PolyPhen2, Mutation Taster, Mutation Assessor, FATHMM, and FATHMM-MKL. d If the FATHMM-MKL coding score (PMID: 25583119) is >0.5 (or rank score >0.28317), the corresponding nonsynonymous SNP is predicted as "damaging type," otherwise it is predicted as "tolerant type." e The Association for Clinical Genomics Sciences (ACGS) classification and interpretation of the variant is shown where applicable. For abbreviations, see ACMG classification above. f PopMax indicates the gnomAD subpopulation with the highest allele frequency.

[0109] To calculate the net increase in LDL-C due to individual rare variants, we assessed the effect size of FH-associated genetic variants by category compared with a reference group of rare variant non-carriers (see Methods above). LDL-C levels were significantly increased in all three categories compared with the reference group (P<0.05) (Fig. 5A). This increase was consistent with the expected pathogenicity (Fig. 5B). The proportion of "disruptive + damaging" variants was significantly higher in subjects with CAD and high LDL-C (ratios = 0.025 and 0.03, respectively) than in the general population (ratio = 0.002) for TMM. －１５ ) was high (Fig. 5C).

[0110] <Combined genetic risk of PRS and rare variants> Standardized PRS ＬＤＬ－Ｃ Using the decile hierarchy, the PRS in the high LDL-C group ＬＤＬ－ＣThe decile hierarchy of the CAD group shifted to a higher level (from 7th to 10th), whereas the PRS group ＬＤＬ－Ｃ The distribution of PRS was found to be prominent in the 9th decile (Fig. 6C). ＬＤＬ－Ｃ Based on the estimated LDL-C elevations by deciles and rare variant categories in Japanese subjects, we examined the correlation between predicted and actual LDL-C values ​​in newly collected subjects at NCGM (Fig. 6D, E). ＬＤＬ－Ｃ Combined genetic risk prediction using both the phenotype and rare variant effects (middle plot; r = 0.261, P = 1.7 × 10 －１１ ) PRS ＬＤＬ－Ｃ Conventional genetic risk prediction using only (left plot; r = 0.151, P = 1.2 × 10 －４ ) was stronger than

[0111] <Effect of PRS and MRS on lipid profile> In each subgroup (Table 20), the distributions of HDL-C, triglycerides, and BMI were unimodal, which differed from the distribution of LDL-C (Figure 7). There was a significant difference in LDL-C (P = 5.8 x 10 －１５８ ), as well as BMI (P = 0.004) and HDL-C (P = 9.2 × 10 －１８ ) significant differences were also observed.

[0112] The consistency (R ２ = 0.34-0.86). Next, we measured the methylation levels of 13 CpG sites using ddPCR (see Table 22). The effect size of the association between CpG and traits correlated well with the results of previous multi-ethnic EWAS studies (Figure 9A, Tables 26-29). Furthermore, there was a significant difference in MRS distribution between the high LDL-C group and the CAD group (P = 0.0045 for LDL, P = 3.5 x 10 for HDL). －７ ) (Fig. 10), which was consistent with the differences in phenotypic distribution between the subgroups.

[0113]

[0114]

[0115]

[0116]

[0117] When CpGs were analyzed individually, the associations between CpGs and traits were more robust for lipid traits in the CAD group compared with the high LDL-C group (Figure 9A). This was supported by the MRS-trait association results, including the cumulative effect of the top CpGs previously reported for each lipid trait (Figure 9B and Tables 30-32). In contrast, the influence of PRS on lipid traits tended to be smaller in the CAD group than in the high LDL-C group. In the regression model, the two risk scores significantly explained the variance of one or more lipid traits in the general population, as represented by the KING cohort. However, there were some differences in their relative influences when comparing the high LDL-C and CAD groups (with the KING cohort). Overall, within each subgroup or population, the effects of PRS and MRS on lipid traits were additive (Figure 9B).

[0118]

[0119] Regression models were used to analyze the association between independent variables (PRS, MRS, or PRS + MRS) and the dependent variable (triglycerides) adjusted for sex, age, smoking status, and BMI. Independent variables were standardized to have a mean of 0 and a standard deviation of 1. To calculate the MRS for triglycerides, the methylation degree measurements of 11 CpGs were multiplied by the corresponding Z-score for each locus from the literature (Jhun et al., 10.1038 / s41467-021-23899-y SuppData2, MODEL4.ALL) and the products were integrated. Abbreviations used: PRS (polygenic risk score) and MRS (methylation risk score).

[0120]

[0121] Regression models were used to analyze the association between independent variables (PRS, MRS, or PRS + MRS) and the dependent variable (HDL-C) adjusted for sex, age, smoking status, and BMI. Independent variables were standardized to have a mean of 0 and a standard deviation of 1. To calculate the MRS for HDL-C, the methylation degree measurements of the six CpGs were multiplied by the corresponding Z-scores for each locus from the literature (Jhun et al., 10.1038 / s41467-021-23899-y SuppData2, MODEL4.ALL) and the products were integrated. Abbreviations used: PRS (polygenic risk score) and MRS (methylation risk score).

[0122]

[0123] Regression models were used to analyze the association between independent variables (PRS, MRS, or PRS + MRS) and the dependent variable (LDL-C), adjusted for sex, age, smoking status, and BMI. Independent variables were standardized to have a mean of 0 and a standard deviation of 1. Imputed LDL-C values ​​were used in the regression analysis. No imputation was performed for LDL-C from the KING study. To calculate the MRS for LDL-C, the methylation degree measurements of the two CpGs were multiplied by the corresponding Z-score for each locus from the literature (Jhun et al., 10.1038 / s41467-021-23899-y SuppData2, MODEL4.ALL), and the products were integrated. Abbreviations: PRS, polygenic risk score; MRS, methylation risk score.

[0124] <Predictability of lipid traits using risk score models> To evaluate individual predictability, we compared the strength of correlation between predicted and measured values ​​of lipid traits using various risk score models (Figures 6D-F and 11). Models incorporating PRS and MRS (and rare variants of LDL-C) showed higher correlations than models using only one risk score, supporting their effectiveness in improving predictability.

[0125] In addition, in the comparison between the high LDL-C group and the CAD group in this example, the high LDL-C group had higher LDL-C, which is one of the risk factors for arteriosclerosis, but lower levels of other risk factors (smoking, hypertension, diabetes, etc.). On the other hand, the coefficient of determination R ２ The CAD group had higher levels of all three lipid traits (Figure 9B, Tables 30-32), and the distribution of methylation scores also shifted (Figure 10). Although this was a cross-sectional analysis, this suggests that the effects of poor lifestyle habits are more pronounced in the CAD group, and that increases or decreases in methylation scores (probably due to the poor lifestyle habits that cause them) are involved in the progression to arteriosclerosis.

[0126] <Evidence of changes in methylation markers due to environmental factors> Typically, lipid EWAS examines the relationship between blood lipid levels and DNA methylation levels collected at a single time point in a population. A significant correlation between variations in lipid levels and DNA methylation levels among different individuals in a population suggests some kind of causal relationship between the two variables (however, it is unclear which is the cause and which is the effect). However, this is merely statistical evidence. More direct biological evidence would be to show a correlation between changes in DNA methylation levels measured in the same individual at two time points, before and after exposure to an environmental factor, and whether or not the individual was exposed.

[0127] Examples of environmental factors that cause changes in DNA methylation include aging and drug treatment. We compared the changes in methylation at specific CpG sites associated with the administration of statins, a lipid-lowering drug that alters the state of hyperlipidemia (especially high LDL-C levels), in a subset of subjects (statin-treated group N=12, non-medicated control group N=9, see Figure 16A). As shown in Figure 16B, we confirmed significant changes in SC4MOL c05119988, a methylation marker for LDL-C.

[0128] <Narrowing down CpG sites using Lasso regression> We further examined whether changes in DNA methylation induced by environmental factors (considering the effects of aging in addition to statin administration) synchronized across different CpG sites. Thirty-four subjects participated in this study (including the 21 subjects mentioned above and the 13 subjects who were followed for the effects of aging over time) and used their EWAS assay data to examine pairwise correlations between top-hit CpG sites for three lipid traits (37 for TG, 36 for HDL-C, and 38 for LDL-C) in the multi-ethnic EWAS (Reference 9).

[0129] As a result, strong correlations were observed not only between adjacent regions on the same chromosome, but also between CpG sites located on different chromosomes. The presence of multicollinearity (a combination of explanatory variables with high correlation coefficients in a multiple regression model) can lead to unstable regression coefficient estimates, potentially adversely affecting the predictive accuracy of the regression model. Therefore, it was deemed appropriate to use Lasso regression (a method for optimizing model performance by reducing the coefficients of unnecessary variables to zero and selecting only important variables) rather than linear regression to evaluate CpG sites to be incorporated into MRS.

[0130] In this example, as described above, a list of target CpG sites (11 for TG, 6 for HDL-C, and 2 for LDL-C) showing a strong and reproducible association with lipid-related phenotypes was initially created without considering the presence or degree of correlation between CpGs located on different chromosomes. Lasso regression was used to verify whether a subset of these CpG sites was selected that was sufficiently strongly associated to represent an MRS model. In the CAD group (N = 315), the LDL-C with the smallest number of features (CpG sites, which are independent variables) and the TG with the largest number were used for verification.

[0131] Regarding LDL-C, considering the possibility that the two sites in the initial list might not be sufficient, we added four sites from the correlation matrix of the 38 top-hit CpG sites described above, and performed Lasso regression on a total of six sites. As a result, three sites (SC4MOL c05119988, DHCR24 cg10073091, and SQL cg09984392) were selected as variables (Figure 17A).

[0132] On the other hand, for TG, Lasso regression was performed on the initially selected 11 locations, and 7 of these locations were selected as variables ( FIG. 17B ).

[0133] Generally, the appropriateness of the MRS model based on the selected variables can be evaluated by the low p-value of the test results for the null hypothesis that the regression coefficient is 0 or the coefficient of determination (R ２ ) can be evaluated by the magnitude of the difference. In the CAD group (N = 315), the p-value was minimized when the CpG sites for LDL-C were the two initially selected (DHCR24 cg10073091 and SC4MOL cg05119988), and when the CpG sites for TG were six (TXNIP cg19693031, SLC1A5 cg02711608, ABCG1 cg06500161, SREBF1 cg11024682, GARS cg21429551, CPT1A cg00574958) (Tables 33 and 34 below, and Figures 17A and 17B). The coefficient of determination (R ２ ) are widely used, so even when the number of TG sites was reduced from the six sites mentioned above to three sites (TXNIP cg19693031, SLC1A5 cg02711608, ABCG1 cg06500161), the model showed a coefficient of determination almost equivalent to that of the 11 sites initially selected.

[0134]

[0135]

[0136] <References> 1. Authors / Task Force Members; ESC Committee for Practice Guidelines (CPG); ESC National Cardiac Societies. 2019 ESC / EAS guidelines for the management of dyslipidaemias: Lipid modification to reduce cardiovascular risk. Atherosclerosis. 2019;290:140 - 205. 2. Migliara G, Baccolini V, Rosso A, D’Andrea E, Massimi A, Villari P, De Vito C. Familial Hypercholesterolemia: A Systematic Review of Guidelines on Genetic Testing and Patient Management. Front Public Health. 2017;5:252 3. Chiu S, Williams PT, Krauss RM. Effects of a very high saturated fat diet on LDL particles in adults with atherogenic dyslipidemia: A randomized controlled trial. PLoS One. 2017;12:e0170664. 4. Stone NJ, Robinson JG, Lichtenstein AH, Bairey Merz CN, Blum CB, Eckel RH, Goldberg AC, Gordon D, Levy D, Lloyd-Jones DM, et al. American College of Cardiology / American Heart Association Task Force on Practice Guidelines. 2013 ACC / AHA guideline on the treatment of blood cholesterol to reduce atherosclerotic cardiovascular risk in adults: a report of the American College of Cardiology / American Heart Association Task Force on Practice Guidelines. Circulation. 2014;129:S1-45. 5. Dixon DL, Sharma G, Sandesara PB, Yang E, Braun LT, Mensah GA, Sperling LS, Deedwania PC, Virani SS. Therapeutic Inertia in Cardiovascular Disease Prevention: Time to Move the Bar. J Am Coll Cardiol. 2019;74:1728-1731. 6. Dugue PA, Jung CH, Joo JE, Wang X, Wong EM, Makalic E, Schmidt DF, Baglietto L, Severi G, Southey MC, et al. Smoking and blood DNA methylation: an epigenome-wide association study and assessment of reversibility. Epigenetics. 2020;15:358-368. 7. Dugue PA, Wilson R, Lehne B, Jayasekara H, Wang X, Jung CH, Joo JE, Makalic E, Schmidt DF, Baglietto L, et al. Alcohol consumption is associated with widespread changes in blood DNA methylation: Analysis of cross-sectional and longitudinal data. Addict Biol. 2021;26:e12855 8. Jones AC, Irvin MR, Claas SA, Arnett DK. Lipid Phenotypes and DNA Methylation: a Review of the Literature. Curr Atheroscler Rep. 2021;23:71. 9. Jhun MA, Mendelson M, Wilson R, Gondalia R, Joehanes R, Salfati E, Zhao X, Braun KVE, Do AN, Hedman AK, et al. A multi-ethnic epigenome-wide association study of leukocyte DNA methylation and blood lipids. Nat Commun. 2021;12:3987. 10. Ding Y, Hou K, Burch KS, Lapinska S, Prive F, Vilhjalmsson B, Sankararaman S, Pasaniuc B. Large uncertainty in individual polygenic risk score estimation impacts PRS-based risk stratification. Nat Genet. 2022;54:30-39. 11. Evans LM, Romero Villela PN. How rare mutations contribute to complex traits. Nature. 2023;614:418-419. 12. Torkamani A, Wineinger NE, Topol EJ. The personal and clinical utility of polygenic risk scores. Nat Rev Genet. 2018;19:581-590. 13. Huls A, Czamara D. Methodological challenges in constructing DNA methylation risk scores. Epigenetics. 2020;15:1-11. 14. Nakatochi M, Ichihara S, Yamamoto K, Naruse K, Yokota S, Asano H, Matsubara T, Yokota M. Epigenome-wide association of myocardial infarction with DNA methylation sites at loci related to cardiovascular disease. Clin Epigenetics. 2017;9:54. 15. Ruel I, Aljenedil S, Sadri I, de Varennes E, Hegele RA, Couture P, Bergeron J, Wanneh E, Baass A, Dufour R, et al. Imputation of Baseline LDL Cholesterol Concentration in Patients with Familial Hypercholesterolemia on Statins or Ezetimibe. Clin Chem. 2018;64:355-362. 16. Liu X, Li C, Mou C, Dong Y, Tu Y. dbNSFP v4: a comprehensive database of transcript-specific functional predictions and annotations for human nonsynonymous and splice-site SNVs. Genome Med. 2020;12:103. 17. Takeuchi F, Akiyama M, Matoba N, Katsuya T, Nakatochi M, Tabara Y, Narita A, Saw WY, Moon S, Spracklen CN, et al. Interethnic analyses of blood pressure loci in populations of East Asian and European descent. Nat Commun. 2018;9:5052. 18. Vilhjalmsson BJ, Yang J, Finucane HK, Gusev A, Lindstrom S, Ripke S, Genovese G, Loh PR, Bhatia G, Do R, et al. Modeling Linkage Disequilibrium Increases Accuracy of Polygenic Risk Scores. Am J Hum Genet. 2015;97:576-592. 19. Kanai M, Akiyama M, Takahashi A, Matoba N, Momozawa Y, Ikeda M, Iwata N, Ikegawa S, Hirata M, Matsuda K, et al. Genetic analysis of quantitative traits in the Japanese population links cell types to complex human diseases. Nat Genet. 2018;50:390-400. 20. Tajima T, Morita H, Ito K, Yamazaki T, Kubo M, Komuro I, Momozawa Y. Blood lipid-related low-frequency variants in LDLR and PCSK9 are associated with onset age and risk of myocardial infarction in Japanese. Sci Rep. 2018;8:8107. 21. Takeuchi F, Takano K, Yamamoto M, Isono M, Miyake W, Mori K, Hara H, Hiroi Y, Kato N. Clinical Implication of Smoking-Related Aryl-Hydrocarbon Receptor Repressor (AHRR) Hypomethylation in Japanese Adults. Circ J. 2022;86:986-992. 22. Shah S, Bonder MJ, Marioni RE, Zhu Z, McRae AF, Zhernakova A, Harris SE, Liewald D, Henders AK, Mendelson MM, et al. Improving Phenotypic Prediction by Combining Genetic and Epigenetic Associations. Am J Hum Genet. 2015;97:75-85. 23. Lacaze P, Manchanda R, Green RC. Prioritizing the detection of rare pathogenic variants in population screening. Nat Rev Genet. 2023;24:205-206. 24. Lee A, Mavaddat N, Wilcox AN, Cunningham AP, Carver T, Hartley S, Babb de Villiers C, Izquierdo A, Simard J, Schmidt MK, et al. BOADICEA: a comprehensive breast cancer risk prediction model incorporating genetic and nongenetic risk factors. Genet Med. 2019;21:1708-1718. 25. Khera AV, Won HH, Peloso GM, Lawson KS, Bartz TM, Deng X, van Leeuwen EM, Natarajan P, Emdin CA, Bick AG, et al. Diagnostic Yield and Clinical Utility of Sequencing Familial Hypercholesterolemia Genes in Patients With Severe Hypercholesterolemia. J Am Coll Cardiol. 2016;67:2578-2589. 26. Natarajan P, Peloso GM, Zekavat SM, Montasser M, Ganna A, Chaffin M, Khera AV, Zhou W, Bloom JM, Engreitz JM, et al. Deep-coverage whole genome sequences and blood lipids among 16,324 individuals. Nat Commun. 2018;9:3391. 27. Sharifi M, Futema M, Nair D, Humphries SE. Genetic Architecture of Familial Hypercholesterolaemia. Curr Cardiol Rep. 2017;19:44. 28. van Dijk SJ, Tellam RL, Morrison JL, Muhlhausler BS, Molloy PL. Recent developments on the role of epigenetics in obesity and metabolic disease. Clin Epigenetics. 2015;7:66. 29. Wahl S, Drong A, Lehne B, Loh M, Scott WR, Kunze S, Tsai PC, Ried JS, Zhang W, Yang Y, et al. Epigenome-wide association study of body mass index, and the adverse outcomes of adiposity. Nature. 2017;541:81-86. 30. Battram T, Gaunt TR, Relton CL, Timpson NJ, Hemani G. A comparison of the genes and genesets identified by GWAS and EWAS of fifteen complex traits. Nat Commun. 2022;13:7816. 31. Takeuchi F, Yokota M, Yamamoto K, Nakashima E, Katsuya T, Asano H, Isono M, Nabika T, Sugiyama T, Fujioka A, et al. Genome-wide association study of coronary artery disease in the Japanese. Eur J Hum Genet. 2012;20:333-340. 32. Frazier-Wood AC,Aslibekyan S,Absher DM,Hopkins PN,Sha J,Tsai MY,Tiwari HK,Waite LL,Zhi D,Arnett DK.Methylation at CPT1A locus is associated with lipoprotein subfraction profiles.J Lipid Res.2014;55:1324-1330. 33. Gomez-Alonso MDC,Kretschmer A,Wilson R,Pfeiffer L,Karhunen V,Seppala I,Zhang W,Mittelstraβ K,Wahl S,Matias-Garcia PR,et al.DNA methylation and lipid metabolism: an EWAS of 226 metabolic measures.Clin Epigenetics.2021;13:7. 34. Irvin MR, Zhi D,Joehanes R,Mendelson M,Aslibekyan S, Claas SA,Thibeault KS,Patel N,Day K,Jones LW,Liang L,Chen BH,Yao C,Tiwari HK,Ordovas JM,Levy D,Absher D,Arnett DK.Epigenome-wide association study of fasting blood lipids in the Genetics of Lipid-lowering Drugs and Diet Network study.Circulation.2014;130:565-572. 35. Pfeiffer L,Wahl S,Pilling LC,Reischl E,Sandling JK,Kunze S,Holdt LM,Kretschmer A,Schramm K,Adamski J,Klopp N,Illig T,Hedman AK,Roden M,Hernandez DG,Singleton AB,Thasler WE,Grallert H,Gieger C,Herder C,Teupser D,Meisinger C,Spector TD,Kronenberg F,Prokisch H,Melzer D,Peters A,Deloukas P,Ferrucci L,Waldenberger M.DNA methylation of lipid-related genes affects blood lipid levels.Circ Cardiovasc Genet.2015;8:34342 36. Thanksgiving s, sightseeing No,fingers s,fingers 0,figs 10 smiling five,fingering five D,AsceticAsceticism 1,AscensionS No, snowy no, snowy no, snowflakes s,brush s,brush s,slim Aesthetic emotion is the feeling s s s s s s s s s s s s s emotional emotionally emotionally Emotion: A scientific feeling. THIS SHY THIS THING FACT201625445564564 37. Note yes, snowflakes no, snowflakes AC, scientific s, scientific s, scientific No,State 0,000,000,000,000,000,000,000,000,000,0 THIS TH, THIS THINGS, THIS POETRY 1,POETRY D,SPECIAL CHRISTIANITY 2, PHYSIOLOGY A, FACILITY D, CHRISTIAN 2,Yeah 0,sweet smile Aesthetics a aesthetic aesthetic sight scientific scientific scientific scientific scientific scientifically scientifically scientifically scientifically The snowflake aesthetic feelings. 100000000000000000000000000000000000000000000000000000000 388 sz s, sb s, s s s, s.s.s THIS YOU ARE YOU, YOU ARE YOUR LIFE snowflake s.Hofman A, Hu FB,Franco OH, Dehghan A. Epigenome-wide association study (EWAS) on lipids: the Rotterdam Study.Clin Epigenetics.2017;9:15. 39. Xie T, Gorenjak V,G Stathopoulou M,Dade S,Marouli E,Masson C,Murray H,Lamont J,Fitzgerald P,Deloukas P,Visvikis-Siest S.Epigenome-Wide Association Study (EWAS) of Blood Lipids in Healthy Population from STANISLAS Family Study(SFS).Int J Mol Sci.2019;20:1014. 40. Lai CQ,Wojczynski MK,Parnell LD,Hidalgo BA, Irvin MR,Aslibekyan S,Province MA, Absher DM,Arnett DK,Ordovas JM.Epigenome-wide association study of triglyceride postprandial responses to a high-fat dietary challenge.J Lipid Res.2016;57:2200-2207. 41. Ochoa-osaless C,ortillaaFeeraanzdee ,Nano J,Wilson n ,Lehne ™MMishra ,ｰoo X,ｧhanbari M,uueda-ｯchoa OL!uvinaooQuintero ,Loh MZhanggggg Kooner Jooyrabe e 24,0elix Soeechotttk B,Zhang ,,iieger 4MMullermNurasyyii) M,Heier M,eters A,ehtimakki ,,､ ABrenner r,6aldebbrrger M,Ikrram ｡ Meurs JBB,Franco O2,6oortmaaoortmaan 4,3tricker Bｨ, Muka 4.･pigeneeticciik Between Statin Therapy and Type e Diabetess､iabetes Careee.2020; 42. Nuotio ML,ervjakova Nuotisuu A,Karhunen V,2iekkalinna 4,ｭilaani L,Kettunen J,!arvelin Mｲ:o ,Metspalu A,ｳalomaa 6,K K,Perola M. study of metabolic syndrome and itts componentss.3ci Rep.2020;100:205467. 43. Andrews SV,addmAcosta C,ｦeibberrg . characterizeclustered probe siiggaa f Illumina methylation array data!･pigenetics Chromatiinn(2016{9:56. 44. Li YF, Feng QZ, Gao WQ, Zhang XJ, Huang Y, Chen YD. The difference between Asian and Western in the effect of LDL-C lowering therapy on coronary atherosclerotic plaque: a meta-analysis report. BMC Cardiovasc Disord. 2015;15:6. ;

[0137] As described above, according to the present invention, it is possible to detect blood lipid concentrations, detect the risk of developing dyslipidemia, and determine a course of prevention or treatment for dyslipidemia. Therefore, the present invention is useful in the medical field related to dyslipidemia.

[0138] In particular, the genetic risk score and methylation score of the present invention are useful for standardizing and refining treatment strategies for individuals requiring treatment for risk factor diseases who are regularly followed up at a medical institution as a result of a medical checkup or other medical examination. By measuring MRS in addition to blood tests during regular checkups, (1) compliance with lifestyle modification can be visualized (numerically) based on information on CpG methylation sites, which change relatively quickly in response to changes in environmental exposure. (2) Comparing the test values ​​of target traits with fluctuations in MRS can help predict whether target values ​​can be achieved through lifestyle modification alone (e.g., it can confirm whether a target trait (e.g., LDL-C level) has stopped decreasing due to lifestyle modification). (3) When assessing the effectiveness of a specific therapeutic drug, prescription optimization can be achieved by referring to increases or decreases in CpG markers [drug-responsive CpG methylation sites] along with increases or decreases in target traits. In particular, in the case of multidrug therapy, utilizing methylation information in addition to target trait measurements to assess the effectiveness and side effects of specific drugs improves prescription accuracy.

Claims

1. A method for detecting blood lipid concentration, comprising: (Step A) detecting at least one lipid trait-related single nucleotide polymorphism and at least one genetic variant selected from the gene group consisting of LDLR, APOB, PCSK9, LDLRAP1, and APOE in cells collected from a subject; (Step B) detecting the methylation level of at least one CpG site selected from the gene group consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1, SLC1A5, and SQLE in cells collected from the subject; and (Step C-1) calculating the blood lipid concentration of the subject based on the lipid trait-related single nucleotide polymorphism and the genetic variant detected in Step A and the methylation level detected in Step B.

2. A method for detecting the onset risk of dyslipidemia, comprising: (Step A) detecting at least one lipid trait-related single nucleotide polymorphism and at least one genetic variant selected from the gene group consisting of LDLR, APOB, PCSK9, LDLRAP1, and APOE in cells collected from a subject; (Step B) detecting the methylation level of at least one CpG site selected from the gene group consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1, SLC1A5, and SQLE in cells collected from the subject; and (Step C-2) detecting the onset risk of dyslipidemia of the subject based on the lipid trait-related single nucleotide polymorphism and the genetic variant detected in Step A and the methylation level detected in Step B.

3. A method for determining a prevention or treatment strategy for dyslipidemia, comprising: (Step A) detecting at least one lipid trait-related single nucleotide polymorphism and at least one genetic variant selected from the gene group consisting of LDLR, APOB, PCSK9, LDLRAP1, and APOE in cells collected from a subject; (Step B) detecting the methylation level of at least one CpG site selected from the gene group consisting of ABCG1, DHCR24, SC4MOL, SARS, PHGDH, TXNIP, SLC7A11, GARS, CPT1A, SREBF1, PHOSPHO1, SLC1A5, and SQLE in cells collected from the subject; and (Step C-3) determining to administer a drug for dyslipidemia and / or provide guidance on improving lifestyle habits to the subject based on the lipid trait-related single nucleotide polymorphism and the genetic variant detected in Step A and the methylation level detected in Step B.

Images