Decoding basic GWAS: The beginner’s guide

Introduction

A genome-wide association study (GWAS) is a research approach used to identify genomic variants that are statistically associated with a particular trait or trait of interest. This method studies the entire set of genomes within a large set of natural populations. If we have the whole genome sequence information of any individual then it becomes easier to identify the function of each gene. So, the next big advancement in the field of genetics after the Human Genome Project (2003) was Genome-Wide Association Studies. As for many studies in the last 40 years the initial applications of GWAS in plants were in Arabidopsis to find the genomic diversity (Borevitz et al., 2007) [2] . Further in cultivated crop Rice was the first to conduct GWAS. This study finds associated genes with traits/diseases throughout the genome using SNP's (Single Nucleotide Polymorphism) to determine the location of these genes. The GWAS studies required many advancements and assistance. Initially in the beginning era of GWAS it required up to $250 million of costing in early 5 years. The first application of GWAS was employed in human beings (Klein et al., 2005) [9] and thereafter in plants (Benjamin et al., 2007; Korte and Farlow, 2013) [1,10] . The genetic sources of phenotypic variation have been a major focus of both plant and animal studies aimed at identifying the causes of disease, improving agriculture and understanding adaptive processes. (Benjamin et al., 2007) [1] . In plants, quantitative trait loci (QTL) were originally mapped in bi-parental crosses, but they were restricted in allelic diversity and in having limited genomic resolution. In such cases the approach GWAS plays a key role to determine allelic diversity. Primary goal of GWAS is to uncover genetic factors that contribute to the risk of developing particular disease or the variation of certain traits. It significantly helps to identify genetic risks of various complex diseases and traits. Also, it gives clues about biological pathways. With the time technique has been advanced with meta-analyses, fine mapping, functional genomics, integration with omics sciences and with machine learning and artificial intelligence (AI) techniques.

Need

The need of GWAS is varies according to crops and breeding objective but some generalize objectives are depicted here: • To use diverse genetic resources is important for breeding crop varieties (Glaszman et al., 2010) [8] . • To exclude crop species with narrow genetic diversity which are susceptible to emerging pathogens or other constraints leading to loss of productivity and this may lead to a serious decline in the areas of adaptation (Dyer et al., 2014) [6] . • To find genes contributing to variation in phenotypes of interest. • Non availability of tightly linked flanking markers to the gene of interest. The NGS (Next Generation Sequencing) facilitates development and use of markers like SNP which overcomes this difficulty.

Features

Single nucleotide polymorphism (SNP) markers are desirable for estimation of genetic diversity because of their abundance in the genome. SNPs have the ability to discriminate between closely related individuals at a higher resolution. SNP markers have been developed and used to study genetic diversity and genetic mapping. Genome-Wide Association Study (GWAS) is an efficient approach to dissect the genetic basis of complex traits using the naturally occurring genetic diversity (Korte and Farlow, 2013) [10] . GWAS provides higher mapping resolution than classical bi-parental populations to detect associations ( [13,3] . GWAS requires an assessment of the population structure to determine the genetic relatedness (Kinship) of individuals and minimize detection of false associations (Korte and Farlow, 2013; Sul et al., 2016) [10] . Steps to detect SNPs Linkage disequilibrium (LD) refers to the non-random association of alleles at different loci in a given population. In genome-wide association studies (GWAS), LD plays a crucial role in identifying genetic variants associated with traits or diseases. Linkage equilibrium (LE) is a random association of alleles at different loci and equal product of allele frequencies within haplotypes.

History

1. Proximity of SNPs-SNPs that are physically close to each other on a chromosome tend to be inherited together. This means that if one SNP is associated with a particular trait, nearby SNPs might also show association due to their linkage.

LD Blocks:

The human genome is organized into blocks of high LD, where recombination events are rare, separated by regions with higher recombination rates. Within these LD blocks, SNPs are strongly correlated with each other.

Tagging SNPs:

In GWAS, not every SNP needs to be genotyped. Instead, a subset of SNPs, known as tagging SNPs, can represent the variation in an LD block. This reduces the genotyping effort while still capturing the genetic diversity of the region. 4. Association Signals: When a SNP shows a significant association with a trait, it suggests that the causal variant (the SNP or gene directly affecting the trait) is nearby, within the same LD block. Researchers then investigate the region around the associated SNP to identify the specific gene or variant responsible. 5. Haplotype Mapping: LD patterns are used to construct haplotypes, which are combinations of alleles at different loci that are inherited together. Haplotypes can provide more information about the genetic architecture of a trait than individual SNPs. After identifying regions of interest in a GWAS, fine-mapping techniques are applied to pinpoint the exact causal variants. This involves studying additional SNPs in high LD with the associated SNPs to narrow down the specific genetic changes responsible for the trait.

Importance of LD in GWAS

It reduces the number of SNPs needed by leveraging LD; fewer SNPs need to be genotyped while still capturing the genetic variation across the genome. LD allows researchers to link associated SNPs to nearby genes or regulatory elements, guiding further functional studies. Understanding LD patterns helps in constructing haplotype.

Manhattan plots

Source: techtarget.com

False Positives

Linkage disequilibrium: LD is non-random association of alleles at different loci, describing the condition of nonequal frequency of haplotype in population. A false positive occurs when a GWAS incorrectly identifies a genetic variant as being associated with a trait when, in fact, there is no real association. This can happen due to random chance, population stratification, or other confounding factors. The consequence of a false positive is that it can lead to pursue non-existent associations, wasting resources and potentially leading to incorrect conclusions.

False Negatives

False negative: The declaration of an outcome as statistically non-significant, when effect is genuine A false negative occurs when a GWAS fails to identify a genetic variant that is truly associated with a trait. This can happen if the study lacks sufficient statistical power, the effect size of the variant is too small to detect, or the variant is not well-represented in the dataset. The consequence of a false negative is that it can prevent the discovery of important genetic associations, leaving gaps in understanding the genetic basis of a trait.

Population Structure

In Genome-Wide Association Studies (GWAS), population structure refers to the presence of systematic differences in allele frequencies between subpopulations within a larger population. This structure can arise due to various factors such as geographical separation, ancestry, or historical migration patterns. If not properly accounted for, population structure can lead to spurious associations between genetic variants and traits, known as false positives, because the allele frequency differences may be confounded with the trait of interest. Population structure can create correlations between genetic variants and traits that are not due to a direct causal relationship but rather due to the shared ancestry within subpopulations. This can lead to the identification of associations that are due to ancestry rather than the trait itself. PCA is commonly used to detect and correct population structure in GWAS. By analyzing the genetic data, PCA identifies principal components that capture the major axes of genetic variation in the study population, which are then included as covariates in the GWAS to control for population structure. Mixed Models is another approach to account for population structure is the use of mixed linear models (MLMs). These models include a random effect to account for genetic relatedness between individuals, which can also capture and correct for population structure. Population stratification is a form of population structure where the population is divided into subgroups with distinct allele frequencies. If stratification is not accounted for, it can lead to biased results in GWAS.

Correcting for population structure is crucial for the accuracy of GWAS results. Failure to do so can lead to incorrect conclusions about the genetic basis of traits and diseases. Therefore, population structure is always a critical consideration in the design and analysis of GWAS.

A Quantile-Quantile (QQ): Plot is a graphical tool used in statistics to compare two probability distributions by plotting their quantiles against each other. In Genome-Wide Association Studies (GWAS), QQ plots are commonly used to assess the presence of population stratification, inflation of test statistics, or other biases in the data. The primary purpose of a QQ plot in GWAS is to visualize the distribution of observed p-values against the expected distribution under the null hypothesis (i.e., no true association between the genetic variants and the trait).

Source: It should be more diverse to have a higher power of GWAS. The increase in sample size may improve detection of causal variants. The ideal size range of population size is 100-500. When population size <100 leads to overestimation of phenotypic variation and gives false association and when Population size beyond 500 will be difficult to handle and manage experimental error.

Step 2: Phenotyping and Collect samples (DNA) Phenotyping: Accurately measure and record the phenotypic traits of interest. This may involve field trials, greenhouse experiments, or laboratory measurements. Population should be grown as in multilocation trials, different seasons in replicated trials. Phenotyping of population as control and cases eg. Drought and irrigated, Infected and healthy, etc further go for isolation of DNA from both types of cases.

Step 3: Genotyping and sequencing DNA Extraction: Extract high-quality DNA from plant tissues, such as leaves. Genotyping: Use high-throughput genotyping platforms like NGS to identify genetic variants (e.g., SNPs) across the genome. Common methods include SNP arrays or sequencing-based approaches.

Data Cleaning and Preparation

Quality Control: Perform quality control on the genotypic data. Filter out low-quality samples and SNPs with low call rates or minor allele frequencies. Remove samples with too many missing SNP (>95%), Remove SNP with very low MAFs (<0.05%), (Uffelmann et al., 2021) [23] . Remove the genotype that does not follow Hardy-Weinberg equilibrium (i.e. allele and genotype frequencies in population will remain constant from generation to generation in absence of evolutionary influences.) Imputation Impute missing genotypes using reference panels or software tools to improve the dataset's completeness like selection, mutation, random mating etc.

Step Functional Annotation: Annotate significant SNPs to identify potential candidate genes or regulatory elements associated with the trait. Pathway Analysis: Perform pathway or network analysis to understand the biological mechanisms underlying the associations.

Step 6: Replication and Validation

Replication helps ensure that a genotype-phenotype association observed in a genome-wide association study represents a credible association and is not a chance finding or an artifact due to uncontrolled biases. Functional Validation: Use experimental methods, such as gene expression studies or gene editing (e.g., CRISPR, TALEN, ZNF), to validate the functional role of candidate genes.

Source: Okutman et al., (2018) [16] Table 1: Difference between QTL and GWAS

QTL GWAS Contrasting, crossable parents and multiple generation required

Natural population, unrelated individuals required Method of mapping linked genes with a particular phenotype with polygenic inheritance It analyzes complete genetic background behind a particular trait Uses particular locus Uses whole genome Less complex More complex No analysis of SNP (specific genomic loci) Analysis of SNP Analysis of polygenic inheritance takes place No analysis of polygenic inheritance Low resolution power due to limited genetic diversity of parents. High resolution power due to more possible recombination in the population.

Studies trait with continuous variation eg. Plant height

Studies association between alleles and binary traits. Linear regression (association between dependent and independent variables) Logistic regression for binary traits with other independent traits

Some of the common models used in GWAS: Simple Linear Regression (or Logistic Regression)

This is also known as Single-locus, single trait (SLST) mixed models. This is the most basic model used in initial GWAS for a single SNP-trait association test and single trait at a time. Linear Regression is used when the trait is continuous (e.g., height, blood pressure). Logistic Regression is used when the trait is binary (e.g., presence or absence of a disease).

Generalized Linear Mixed Models (GLMM)

It extends mixed linear models to handle non-normal traits (e.g., binary, count, categorical traits). GLMM is flexible in handling various types of traits while accounting for relatedness and population structure. The software like SAIGE is often used for binary traits in large cohorts.

Bayesian Methods

Incorporates prior information into the GWAS analysis, which can be helpful in settings with limited data or when combining GWAS with other types of data. This is generally used effective fine mapping in candidate regions. It can handle complex models, incorporate prior knowledge, and provide full posterior distributions of effect sizes. Bayesian software for GWAS includes BIMBAM, PIQUE, and FINEMAP etc.

Multivariate GWAS

Analyzes multiple traits simultaneously, which can increase power to detect pleiotropic effects (where a single genetic variant affects multiple traits). It is more powerful than univariate GWAS when the traits are correlated. Tools like MTAG*(Multi-Trait Analysis of GWAS) and PLINK can be used for this purpose.

Meta-Analysis Models

This model is used to combine results from multiple GWAS studies to increase statistical power and detect associations that might not be evident in individual studies. It increases power, particularly for rare variants, and allows for crossstudy comparisons. The common tools include METAL and GWAMA.

Rare Variant Association Models

This focuses on the association of rare variants with traits, often using [19] similarly the study carried out for cassava mosaic disease and cassava green mite severity were carried out identifying several novel and previously reported associations (Sonnewald et al., 2020) [24] . [7] . Similarly, Two rice diversity panels, indica and aus, containing a total of 553 genotypes, are phenotyped in waterlogged conditions. (Tamimi et al., 2016) [15] 3. Yield associated traits: The GWAS for the yield and yield contributing character to enhance the yield eg. In rice demonstrating subtle links between panicle size and yield performance (Crowell et al., 2015) [21] . In upland cotton a GWAS identified two ethylene pathway related genes as associated with increased lint yield. (Fang et al., 2017) [11] . 4. Metabolic composition: GWAS was used for metabolite profiling and taste panels to characterize the genetic architecture of tomato fruit taste (Tieman et al., 2017) [4] . In maize identified 74 loci associated with kernel oil levels and fatty acid composition. (Hui Li et al., 2019).

Limitations

One of the disadvantages of GWAS is it is unable to give true results in the case of traits controlled by polygenic mechanisms because of the marked influence of the environment on allelic frequency. There is less power of GWAS to study the variation due to natural factors hence there is need to improve it.

Challenges in GWAS

Complex Traits: Many traits are influenced by multiple genes and environmental factors, making it challenging to identify all contributing variants.

Population Stratification: Differences in ancestry among study participants can lead to false associations if not properly controlled.

Data Interpretation: Distinguishing between causative variants and those merely linked to the trait due to their proximity in the genome.

Key Considerations in Plant GWAS

Population Structure: Plants often have complex population structures due to breeding practices and natural selection. Use statistical methods to control for this structure.

Environmental Interactions: Consider the influence of environmental factors on trait expression and ensure phenotyping is consistent across different conditions.

Genomic Resources: Utilize available genomic resources, such as reference genomes and annotated gene databases, to aid in the interpretation of GWAS results.

Conclusion

The larger the population and the larger the number of SNPs the greater the chance of a successful result. The developments of molecular biology techniques (e.g., CRISPR/Cas9, over-expression, or genetic complementation) have greatly accelerated the biological functions of the causative genes behind the GWAS hits. GWAS is its lack of power to detect rare alleles that are involved in natural variation.

References

1. Benjamin B, Borevitz JO, Morris P. Genome-wide patterns of single-feature polymorphism in Arabidopsis thaliana. Proc Natl Acad Sci USA. 2007;104(29):12057-62.
2. Borevitz JO. Genome-wide patterns of single-feature polymorphism in Arabidopsis thaliana. Proc Natl Acad Sci USA. 2007;104(29):12057-62.
3. Cui C, Mei H, Liu Y, Zhang H, Zheng Y. Genetic diversity, population structure, and linkage disequilibrium of an association-mapping panel revealed by genome-wide SNP markers in sesame. Front Plant Sci. 2017;8:1189.
4. Tieman D, Zhu G, Resende Jr MF, Lin T, Nguyen C, Bies D, et al. A chemical genetic roadmap to improved tomato flavor. Science. 2017;355:391-4.
5. Duc G, Agrama HA, Bao S, Berger J, Bourion V, De Ron AM. Breeding annual grain legumes for sustainable agriculture: New methods to approach complex traits and target new cultivar ideotypes. Crit Rev Plant Sci. 2015;34:381-411.
6. Dyer GA, Lopez-Feldman A, Yunez-Naude A, Taylor JE. Genetic erosion in maize's center of origin. PNAS. 2014;111:14094-9.
7. Rabieyan E, Bihamta MR, Esmaeilzadeh Moghaddam M, Alipour H, Mohammadi V, et al. Analysis of genetic diversity and genome-wide association study for drought tolerance related traits in Iranian bread wheat. BMC Plant Biol. 2023;23:431.
8. Glaszmann JC, Kilian B, Upadhyaya HD, Varshney RK. Accessing genetic diversity for crop improvement. Curr Opin Plant Biol. 2010;13:167-73.
9. Klein R. Complement factor H polymorphism in agerelated macular degeneration. Science. 2005;308(5720):385-9.
10. Korte A, Farlow A. The advantages and limitations of trait analysis with GWAS: a review. Plant Methods. 2013;9:29.
11. Li H, Wang X, Zhuo Y, Chen S, Lin J, Ma H, et al. Molecular characterization and expression analysis of the remorin genes in tomato (Solanum lycopersicum L.). Front Plant Sci. 2023;14:1175153.
12. Liu N, Xue Y, Guo Z, Li W, Tang J. Genome-wide association study identifies candidate genes for starch content regulation in maize kernels. Front Plant Sci. 2016;7:1046.
13. Fosgerau M, Paulsen M, Kjaer Rasmussen T. A perturbed utility route choice model. Transp Res Part C. 2022;136:103514.
14. Al-Tamimi N, Brien C, Oakey H, Berger B, Saade S, Ho YS, et al. Salinity tolerance loci revealed in rice using high-throughput non-invasive phenotyping. Nat Commun. 2016;7:13342.
15. Okutman O, Rhouma MB, Benkhalifa M, et al. Genetic evaluation of patients with non-syndromic male infertility. J Assist Reprod Genet. 2018;35:1939-51.
16. Kulwal P, Ishikawa G, Benscher D, Feng Z, Yu LX, Jadhav A, et al. Association mapping for pre-harvest sprouting resistance in white winter wheat. Theor Appl Genet. 2012;125:793-805.
17. Gupta PK, Kulwal PL, Jaiswal V. Association mapping in plants in the post-GWAS genomics era. Adv Genet. 2018;1-80.
18. Joukhadar R, Hollaway G, Shi F, Kant S, Forrest K, Wong D, et al. Genome-wide association reveals a complex architecture for rust resistance in 2300 worldwide bread wheat accessions screened under various Australian conditions. Theor Appl Genet. 2020;133:3626-9.
19. Saleh A, Dimitrina K, Mustafa B, Alisdair RF. Genome-wide association studies: Assessing trait characteristics in model and crop plants. Cell Mol Life Sci. 2021;78:5743-54.
20. Crowell S, Korniliev P, Falca˜o A, Ismail A, Gregorio G, Mezey J, et al. Genome-wide association and highresolution phenotyping link Oryza sativa panicle traits to numerous trait-specific QTL clusters. Nat Commun. 2015;6:10527.
21. Uffelmann E, Huang QQ, Nchangwi SM, Jantina V, Yukinori O, Alicia R, et al. Genome-wide association studies. Primer. 2021;1:59.
22. Sonnewald U, Fernie AR, Gruissem W, Chang SH, Ludewig F, Rascher U, et al. The Cassava Source-Sink project: opportunities and challenges for crop improvement by metabolic engineering. Plant J. 2020;146:65.
23. Xu Y, Li P, Yang Z, Xu C. Genetic mapping of quantitative trait loci in crops. Crop J. 2017;5:175-84.