Machine learning-based quantification for disease uncertainty increases the statistical power of genetic association studies.

MOTIVATION: Allowance for increasingly large samples is a key to identify the association of genetic variants with Alzheimer's disease (AD) in genome-wide association studies (GWAS). Accordingly, we aimed to develop a method that incorporates patients with mild cognitive impairment and unknown cognitive status in GWAS using a machine learning-based AD prediction model. RESULTS: Simulation analyses showed that weighting imputed phenotypes method increased the statistical power compared to ordinary logistic regression using only AD cases and controls. Applied to real-world data, the penalized logistic method had the highest AUC (0.96) for AD prediction and weighting imputed phenotypes method performed well in terms of power. We identified an association (P<5.0×10-8) of AD with several variants in the APOE region and rs143625563 in LMX1A. Our method, which allows the inclusion of individuals with mild cognitive impairment, improves the statistical power of GWAS for AD. We discovered a novel association with LMX1A. AVAILABILITY AND IMPLEMENTATION: Simulation codes can be accessed at https://github.com/Junkkkk/wGEE_GWAS.

Multi-ancestry meta-analysis and fine-mapping in Alzheimer's disease.

Genome-wide association studies (GWAS) of Alzheimer's disease are predominantly carried out in European ancestry individuals despite the known variation in genetic architecture and disease prevalence across global populations. We leveraged published GWAS summary statistics from European, East Asian, and African American populations, and an additional GWAS from a Caribbean Hispanic population using previously reported genotype data to perform the largest multi-ancestry GWAS meta-analysis of Alzheimer's disease and related dementias to date. This method allowed us to identify two independent novel disease-associated loci on chromosome 3. We also leveraged diverse haplotype structures to fine-map nine loci with a posterior probability >0.8 and globally assessed the heterogeneity of known risk factors across populations. Additionally, we compared the generalizability of multi-ancestry- and single-ancestry-derived polygenic risk scores in a three-way admixed Colombian population. Our findings highlight the importance of multi-ancestry representation in uncovering and understanding putative factors that contribute to risk of Alzheimer's disease and related dementias.

Heritability of cognitive abilities and regional brain structures in middle-aged to elderly East Asians.

This study examined the single-nucleotide polymorphism heritability and genetic correlations of cognitive abilities and brain structural measures (regional subcortical volume and cortical thickness) in middle-aged and elderly East Asians (Korean) from the Gwangju Alzheimer's and Related Dementias cohort study. Significant heritability was found in memory function, caudate volume, thickness of the entorhinal cortices, pars opercularis, superior frontal gyri, and transverse temporal gyri. There were 3 significant genetic correlations between (i) the caudate volume and the thickness of the entorhinal cortices, (ii) the thickness of the superior frontal gyri and pars opercularis, and (iii) the thickness of the superior frontal and transverse temporal gyri. This is the first study to describe the heritability and genetic correlations of cognitive and neuroanatomical traits in middle-aged to elderly East Asians. Our results support the previous findings showing that genetic factors play a substantial role in the cognitive and neuroanatomical traits in middle to advanced age. Moreover, by demonstrating shared genetic effects on different brain regions, it gives us a genetic insight into understanding cognitive and brain changes with age, such as aging-related cognitive decline, cortical atrophy, and neural compensation.

Unveiling Bladder Cancer Prognostic Insights by Integrating Patient-Matched Sample and CpG Methylation Analysis.

Bladder cancer prognosis remains a pressing clinical challenge, necessitating the identification of novel biomarkers for precise survival prediction and improved quality of life outcomes. This study proposes a comprehensive strategy to uncover key prognostic biomarkers in bladder cancer using DNA methylation analysis and extreme survival pattern observations in matched pairs of cancer and adjacent normal cells. Unlike traditional approaches that overlook cancer heterogeneity by analyzing entire samples, our methodology leverages patient-matched samples to account for this variability. Specifically, DNA methylation profiles from adjacent normal bladder tissue and bladder cancer tissue collected from the same individuals were analyzed to pinpoint critical methylation changes specific to cancer cells while mitigating confounding effects from individual genetic differences. Utilizing differential threshold settings for methylation levels within cancer-associated pathways enabled the identification of biomarkers that significantly impact patient survival. Our analysis identified distinct survival patterns associated with specific CpG sites, underscoring these sites' pivotal roles in bladder cancer outcomes. By hypothesizing and testing the influence of methylation levels on survival, we pinpointed CpG biomarkers that profoundly affect the prognosis. Notably, CpG markers, such as cg16269144 (PRKCZ), cg16624272 (PTK2), cg11304234, and cg26534425 (IL18), exhibited critical methylation thresholds that correlate with patient mortality. This study emphasizes the importance of tailored approaches to enhancing prognostic accuracy and refining therapeutic strategies for bladder cancer patients. The identified biomarkers pave the way for personalized prognostication and targeted interventions, promising advancements in bladder cancer management and patient care.

BALLI: Bartlett-adjusted likelihood-based linear model approach for identifying differentially expressed genes with RNA-seq data.

BACKGROUND: Transcriptomic profiles can improve our understanding of the phenotypic molecular basis of biological research, and many statistical methods have been proposed to identify differentially expressed genes (DEGs) under two or more conditions with RNA-seq data. However, statistical analyses with RNA-seq data are often limited by small sample sizes, and global variance estimates of RNA expression levels have been utilized as prior distributions for gene-specific variance estimates, making it difficult to generalize the methods to more complicated settings. We herein proposed a Bartlett-Adjusted Likelihood-based LInear mixed model approach (BALLI) to analyze more complicated RNA-seq data. The proposed method estimates the technical and biological variances with a linear mixed-effects model, with and without adjusting small sample bias using Bartlkett's corrections. RESULTS: We conducted extensive simulations to compare the performance of BALLI with those of existing approaches (edgeR, DESeq2, and voom). Results from the simulation studies showed that BALLI correctly controlled the type-1 error rates at various nominal significance levels and produced better statistical power and precision estimates than those of other competing methods in various scenarios. Furthermore, BALLI was robust to variation of library size. It was also successfully applied to Holstein milk yield data, illustrating its practical value. CONCLUSIONS;: BALLI is statistically more efficient and valid than existing methods, and we conclude that it is useful for identifying DEGs in RNA-seq analysis.

Protective Effects of Alpha-Lipoic Acid on Glutamate-Induced Cytotoxicity in C6 Glioma Cells.

Glutamate-mediated cytotoxicity has been implicated in the pathogenesis of neurological diseases, including Parkinson's disease, Alzheimer's disease, and stroke. In this study, we investigated the protective effects of alpha-lipoic acid (ALA), a naturally occurring thiol antioxidant, on glutamate-induced cytotoxicity in cultured C6 astroglial cells. Exposure to high-dose glutamate (10 mM) caused oxidative stress and mitochondrial dysfunction through the elevation of reactive oxygen species, depletion of glutathione, and loss of the mitochondrial membrane potential (ΔΨm). Pretreatment with ALA (200 µM), however, significantly inhibited the glutamate-induced oxidative stress and mitochondrial dysfunction. ALA pretreatment dose-dependently suppressed glutamate-induced apoptotic events including altered nuclear morphology and activation of caspase-3. In addition, ALA significantly attenuated glutamate-induced endoplasmic reticulum (ER) stress markers; namely, glucose-regulated protein 78 (GRP78), activating transcription factor 6 (ATF6), protein kinase regulated by RNA (PKR)-like ER-associated kinase (PERK), eukaryotic translation initiation factor 2 alpha (eIF2α), inositol-requiring enzyme 1 (IRE1), CCAAT/enhancer binding protein homologous protein (CHOP), and caspase-12. We confirmed that CHOP and caspase-12 are key mediators of glutamate-induced ER stress. Furthermore, exposure of the cells to a caspase-12-specific inhibitor and CHOP small interfering RNAs (siRNAs) led to restoration of the ΔΨm that was damaged by glutamate treatment. These results suggest that ALA can effectively suppress oxidative stress, mitochondrial dysfunction, and ER stress in astroglial cells.

Targeted Proteomics Predicts a Sustained Complete-Response after Transarterial Chemoembolization and Clinical Outcomes in Patients with Hepatocellular Carcinoma: A Prospective Cohort Study.

This study was aimed to identify blood-based biomarkers to predict a sustained complete response (CR) after transarterial chemoembolization (TACE) using targeted proteomics. Consecutive patients with HCC who had undergone TACE were prospectively enrolled (training (n = 100) and validation set (n = 80)). Serum samples were obtained before and 6 months after TACE. Treatment responses were evaluated using the modified Response Evaluation Criteria in Solid Tumors (mRECIST). In the training set, the MRM-MS assay identified five marker candidate proteins (LRG1, APCS, BCHE, C7, and FCN3). When this five-marker panel was combined with the best-performing clinical variables (tumor number, baseline PIVKA, and baseline AFP), the resulting ensemble model had the highest area under the receiver operating curve (AUROC) value in predicting a sustained CR after TACE in the training and validation sets (0.881 and 0.813, respectively). Furthermore, the ensemble model was an independent predictor of rapid progression (hazard ratio (HR), 2.889; 95% confidence interval (CI), 1.612-5.178; P value < 0.001) and overall an unfavorable survival rate (HR, 1.985; 95% CI, 1.024-3.848; P value = 0.042) in the entire population by multivariate analysis. Targeted proteomics-based ensemble model can predict clinical outcomes after TACE. Therefore, this model can aid in determining the best candidates for TACE and the need for adjuvant therapy.

Family-Based Rare Variant Association Analysis: A Fast and Efficient Method of Multivariate Phenotype Association Analysis.

Family-based designs have been repeatedly shown to be powerful in detecting the significant rare variants associated with human diseases. Furthermore, human diseases are often defined by the outcomes of multiple phenotypes, and thus we expect multivariate family-based analyses may be very efficient in detecting associations with rare variants. However, few statistical methods implementing this strategy have been developed for family-based designs. In this report, we describe one such implementation: the multivariate family-based rare variant association tool (mFARVAT). mFARVAT is a quasi-likelihood-based score test for rare variant association analysis with multiple phenotypes, and tests both homogeneous and heterogeneous effects of each variant on multiple phenotypes. Simulation results show that the proposed method is generally robust and efficient for various disease models, and we identify some promising candidate genes associated with chronic obstructive pulmonary disease. The software of mFARVAT is freely available at http://healthstat.snu.ac.kr/software/mfarvat/, implemented in C++ and supported on Linux and MS Windows.

Network analysis for count data with excess zeros.

BACKGROUND: Undirected graphical models or Markov random fields have been a popular class of models for representing conditional dependence relationships between nodes. In particular, Markov networks help us to understand complex interactions between genes in biological processes of a cell. Local Poisson models seem to be promising in modeling positive as well as negative dependencies for count data. Furthermore, when zero counts are more frequent than are expected, excess zeros should be considered in the model. METHODS: We present a penalized Poisson graphical model for zero inflated count data and derive an expectation-maximization (EM) algorithm built on coordinate descent. Our method is shown to be effective through simulated and real data analysis. RESULTS: Results from the simulated data indicate that our method outperforms the local Poisson graphical model in the presence of excess zeros. In an application to a RNA sequencing data, we also investigate the gender effect by comparing the estimated networks according to different genders. Our method may help us in identifying biological pathways linked to sex hormone regulation and thus understanding underlying mechanisms of the gender differences. CONCLUSIONS: We have presented a penalized version of zero inflated spatial Poisson regression and derive an efficient EM algorithm built on coordinate descent. We discuss possible improvements of our method as well as potential research directions associated with our findings from the RNA sequencing data.

Propensity score matching for comparative studies: a tutorial with R and Rex.

Recently, there has been considerable progress in developing new technologies and equipment for the medical field, including minimally invasive surgeries. Evaluating the effectiveness of these treatments requires study designs like randomized controlled trials. However, due to the nature of certain treatments, randomization is not always feasible, leading to the use of observational studies. The effect size estimated from observational studies is subject to selection bias caused by confounders. One method to reduce this bias is propensity scoring. This study aimed to introduce a propensity score matching process between two groups using a practical example with R. Additionally, Rex, an Excel add-in graphical user interface statistical program, is provided for researchers unfamiliar with R programming. Further techniques, such as matching with three or more groups, propensity score weighting and stratification, and imputation of missing values, are summarized to offer approaches for more complex studies not covered in this tutorial.

GraphMHC: Neoantigen prediction model applying the graph neural network to molecular structure.

Neoantigens are tumor-derived peptides and are biomarkers that can predict prognosis related to immune checkpoint inhibition by estimating their binding to major histocompatibility complex (MHC) proteins. Although deep neural networks have been primarily used for these prediction models, it is difficult to interpret the models reported thus far as accurately representing the interactions between biomolecules. In this study, we propose the GraphMHC model, which utilizes a graph neural network model applied to molecular structure to simulate the binding between MHC proteins and peptide sequences. Amino acid sequences sourced from the immune epitope database (IEDB) undergo conversion into molecular structures. Subsequently, atomic intrinsic informations and inter-atomic connections are extracted and structured as a graph representation. Stacked graph attention and convolution layers comprise the GraphMHC network which classifies bindings. The prediction results from the test set using the GraphMHC model showed a high performance with an area under the receiver operating characteristic curve of 92.2% (91.9-92.5%), surpassing a baseline model. Moreover, by applying the GraphMHC model to melanoma patient data from The Cancer Genome Atlas project, we found a borderline difference (0.061) in overall survival and a significant difference in stromal score between the high and low neoantigen load groups. This distinction was not present in the baseline model. This study presents the first feature-intrinsic method based on biochemical molecular structure for modeling the binding between MHC protein sequences and neoantigen candidate peptide sequences. This model can provide highly accurate responsibility information that can predict the prognosis of immune checkpoint inhibitors to cancer patients who want to apply it.

Multiphasic analysis of whole exome sequencing data identifies a novel mutation of ACTG1 in a nonsyndromic hearing loss family.

BACKGROUND: The genetic heterogeneity of sensorineural hearing loss is a major hurdle to the efficient discovery of disease-causing genes. We designed a multiphasic analysis of copy number variation (CNV), linkage, and single nucleotide variation (SNV) of whole exome sequencing (WES) data for the efficient discovery of mutations causing nonsyndromic hearing loss (NSHL). RESULTS: From WES data, we identified five distinct CNV loci from a NSHL family, but they were not co-segregated among patients. Linkage analysis based on SNVs identified six candidate loci (logarithm of odds [LOD] >1.5). We selected 15 SNVs that co-segregated with NSHL in the family, which were located in six linkage candidate loci. Finally, the novel variant p.M305T in ACTG1 (DFNA20/26) was selected as a disease-causing variant. CONCLUSIONS: Here, we present a multiphasic CNV, linkage, and SNV analysis of WES data for the identification of a candidate mutation causing NSHL. Our stepwise, multiphasic approach enabled us to expedite the discovery of disease-causing variants from a large number of patient variants.

A comparative investigation of single nucleotide variant calling for a personal non-Caucasian sequencing sample.

BACKGROUND: Dropping cost and increasing clinical application of whole genome sequencing (WGS) lead a necessity of efficient (accurate and rapid) variant calling procedures from a personal WGS data (n = 1). A number of variant calling pipelines have been introduced utilizing the human genome reference GRCh38 as a reference and a benchmark dataset called 'NA12878', which are both 'standard' but limited ethnic origin. Considering the nature of variant calling algorithms and recent updates in sequencing protocol, however, it is necessary to revisit the efficiency of the current best pipelines for a personal WGS data from diverse ethnicity. OBJECTIVE: We discuss the most efficient practices for variant calling of a personal WGS reads, with a particular emphasis on whether (1) ethnic match or mismatch between the reference genome and a WGS data produces a distinct result and more importantly (2) there is an ethnic-specific optimal workflow. METHODS: Here, we generate an appropriate WGS data, DNA array, and sufficient number of Sanger validated variants from a single Korean subject to perform such a comprehensive comparison. We applied this WGS reads and the 'NA12878' reads to 8 different variant calling pipelines with 2 different reference genomes (GRCh38 and KOREF, a Korean reference genome) to which the WGS reads from different ethnic origins are aligned. RESULTS: We evaluated the performance of the pipelines with the matched array genotype data and Sanger sequencing validation and demonstrated that: regardless to the ethnic match/mismatch (1) Novoalign-GATK4 showed the most efficient performance with the exceptional calls in MHC region; (2) the overall performance was better with GRCh38, while a significant difference in recall was observed. In addition, we found it is largely reduced computing cost maintaining performance to remove 'markduplication' step with PCR-free WGS data. CONCLUSION: For variant calling of a personal PCR-free WGS data, regardless of ethnicity consideration, we recommend the use of the Novoalign + GATK4 with GRCh38 and without 'markduplication'.

A comparative investigation of variant calling and genotyping for a single non-Caucasian whole genome.

Most genome benchmark studies utilize hg38 as a reference genome (based on Caucasian and African samples) and 'NA12878' (a Caucasian sequencing read) for comparison. Here, we aimed to elucidate whether 1) ethnic match or mismatch between the reference genome and sequencing reads produces a distinct result; 2) there is an optimal work flow for single genome data. We assessed the performance of variant calling pipelines using hg38 and a Korean genome (reference genomes) and two whole-genome sequencing (WGS) reads from different ethnic origins: Caucasian (NA12878) and Korean. The pipelines used BWA-mem and Novoalign as mapping tools and GATK4, Strelka2, DeepVariant, and Samtools as variant callers. Using hg38 led to better performance (based on precision and recall), regardless of the ethnic origin of the WGS reads. Novoalign + GATK4 demonstrated best performance when using both WGS data. We assessed pipeline efficiency by removing the markduplicate process, and all pipelines, except Novoalign + DeepVariant, maintained their performance. Novoalign identified more variants overall and in MHC of chr6 when combined with GATK4. No evidence suggested improved variant calling performance from single WGS reads with a different ethnic reference, re-validating hg38 utility. We recommend using Novoalign + GATK4 without markduplication for single PCR-free WGS data.

DeepParcellation: A novel deep learning method for robust brain magnetic resonance imaging parcellation in older East Asians.

Accurate parcellation of cortical regions is crucial for distinguishing morphometric changes in aged brains, particularly in degenerative brain diseases. Normal aging and neurodegeneration precipitate brain structural changes, leading to distinct tissue contrast and shape in people aged >60 years. Manual parcellation by trained radiologists can yield a highly accurate outline of the brain; however, analyzing large datasets is laborious and expensive. Alternatively, newly-developed computational models can quickly and accurately conduct brain parcellation, although thus far only for the brains of Caucasian individuals. To develop a computational model for the brain parcellation of older East Asians, we trained magnetic resonance images of dimensions 256 × 256 × 256 on 5,035 brains of older East Asians (Gwangju Alzheimer's and Related Dementia) and 2,535 brains of Caucasians. The novel N-way strategy combining three memory reduction techniques inception blocks, dilated convolutions, and attention gates was adopted for our model to overcome the intrinsic memory requirement problem. Our method proved to be compatible with the commonly used parcellation model for Caucasians and showed higher similarity and robust reliability in older aged and East Asian groups. In addition, several brain regions showing the superiority of the parcellation suggest that DeepParcellation has a great potential for applications in neurodegenerative diseases such as Alzheimer's disease.

A system-level investigation into the cellular toxic response mechanism mediated by AhR signal transduction pathway.

MOTIVATION: Viewing a cellular system as a collection of interacting parts can lead to new insights into the complex cellular behavior. In this study, we have investigated aryl hydrocarbon receptor (AhR) signal transduction pathway from such a system-level perspective. AhR detects various xenobiotics, such as drugs or endocrine disruptors (e.g. dioxin), and mediates transcriptional regulation of target genes such as those in the cytochrome P450 (CYP450) family. On binding with 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), however, AhR becomes abnormally activated and conveys toxic effects on cells. Despite many related studies on the TCDD-mediated toxicity, quantitative system-level understanding of how TCDD-mediated toxicity generates various toxic responses is still lacking. RESULTS: Here, we present a manually curated TCDD-mediated AhR signaling pathway including crosstalks with the hypoxia pathway that copes with oxygen deficiency and the p53 pathway that induces a DNA damage response. Based on the integrated pathway, we have constructed a mathematical model and validated it through quantitative experiments. Using the mathematical model, we have investigated: (i) TCDD dose-dependent effects on AhR target genes; (ii) the crosstalk effect between AhR and hypoxia signals; and (iii) p53 inhibition effect of TCDD-liganded AhR. Our results show that cellular intake of TCDD induces AhR signaling pathway to be abnormally up-regulated and thereby interrupts other signaling pathways. Interruption of hypoxia and p53 pathways, in turn, can incur various hazardous effects on cells. Taken together, our study provides a system-level understanding of how AhR signal mediates various TCDD-induced toxicities under the presence of hypoxia and/or DNA damage in cells.

Dehydroabietic Acid Is a Novel Survivin Inhibitor for Gastric Cancer.

Gastric cancer is a malignant tumor with a high incidence and mortality rate worldwide. Nevertheless, anticancer drugs that can be used for gastric cancer treatment are limited. Therefore, it is important to develop targeted anticancer drugs for the treatment of gastric cancer. Dehydroabietic acid (DAA) is a diterpene found in tree pine. Previous studies have demonstrated that DAA inhibits gastric cancer cell proliferation by inducing apoptosis. However, we did not know how DAA inhibits the proliferation of gastric cancer cells through apoptosis. In this study, we attempted to identify the genes that induce cell cycle arrest and cell death, as well as those which are altered by DAA treatment. DAA-regulated genes were screened using RNA-Seq and differentially expressed genes (DEGs) analysis in AGS cells. RNA-Seq analysis revealed that the expression of survivin, an apoptosis inhibitor, was significantly reduced by DAA treatment. We also confirmed that DAA decreased survivin expression by RT-PCR and Western blotting analysis. In addition, the ability of DAA to inhibit survivin was compared to that of YM-155, a known survivin inhibitor. DAA was found to have a stronger inhibitory effect in comparison with YM-155. DAA also caused an increase in cleaved caspase-3, an apoptosis-activating protein. In conclusion, DAA is a potential anticancer agent for gastric cancer that inhibits survivin expression.

A missense variant in SHARPIN mediates Alzheimer's disease-specific brain damages.

Established genetic risk factors for Alzheimer's disease (AD) account for only a portion of AD heritability. The aim of this study was to identify novel associations between genetic variants and AD-specific brain atrophy. We conducted genome-wide association studies for brain magnetic resonance imaging measures of hippocampal volume and entorhinal cortical thickness in 2643 Koreans meeting the clinical criteria for AD (n = 209), mild cognitive impairment (n = 1449) or normal cognition (n = 985). A missense variant, rs77359862 (R274W), in the SHANK-associated RH Domain Interactor (SHARPIN) gene was associated with entorhinal cortical thickness (p = 5.0 × 10-9) and hippocampal volume (p = 5.1 × 10-12). It revealed an increased risk of developing AD in the mediation analyses. This variant was also associated with amyloid-ß accumulation (p = 0.03) and measures of memory (p = 1.0 × 10-4) and executive function (p = 0.04). We also found significant association of other SHARPIN variants with hippocampal volume in the Alzheimer's Disease Neuroimaging Initiative (rs3417062, p = 4.1 × 10-6) and AddNeuroMed (rs138412600, p = 5.9 × 10-5) cohorts. Further, molecular dynamics simulations and co-immunoprecipitation indicated that the variant significantly reduced the binding of linear ubiquitination assembly complex proteins, SHPARIN and HOIL-1 Interacting Protein (HOIP), altering the downstream NF-κB signaling pathway. These findings suggest that SHARPIN plays an important role in the pathogenesis of AD.

Potential Novel Genes for Late-Onset Alzheimer's Disease in East-Asian Descent Identified by APOE-Stratified Genome-Wide Association Study.

The present study reports two novel genome-wide significant loci for late-onset Alzheimer's disease (LOAD) identified from APOE ε4 non-carrier subjects of East Asian origin. A genome-wide association study of Alzheimer's disease was performed in 2,291 Korean seniors in the discovery phase, from the Gwangju Alzheimer' and Related Dementias (GARD) cohort study. The study was replicated in a Japanese cohort of 1,956 subjects that suggested two novel susceptible SNPs in two genes: LRIG1 and CACNA1A. This study demonstrates that the discovery of AD-associated variants is feasible in non-European ethnic groups using samples comprising fewer subjects from the more homogeneous genetic background.

A Between Ethnicities Comparison of Chronic Obstructive Pulmonary Disease Genetic Risk.

Heterogeneity of lung function levels and risk for developing chronic obstructive pulmonary disease (COPD) among people exposed to the same environmental risk factors, such as cigarette smoking, suggest an important role of genetic factors in COPD susceptibility. To investigate the possible role of different genetic factors in COPD susceptibility across ethnicities. We used a population-stratified analysis for: (i) identifying ethnic-specific genetic susceptibility loci, (ii) developing ethnic-specific polygenic risk prediction models using those SNPs, and (iii) validating the models with an independent dataset. We elucidated substantial differences in SNP heritability and susceptibility loci for the disease across ethnicities. Furthermore, the application of three ethnic-specific prediction models to an independent dataset showed that the best performance is achieved when the prediction model is applied to a dataset with the matched ethnic sample. Our study validates the necessity of considering ethnic differences in COPD risk; understanding these differences might help in preventing COPD and developing therapeutic strategies.