1. Obtain statistical significance levels for associations between
traits and SNPs by applying mixed linear models (MLM),
including both fixed and random effects. We usually take individuals as random effects, allowing MLM to incorporate information about relationships among individuals. The results of
associations are usually demonstrated by Manhattan plots.
2. Identify candidate associations between traits and genes by
comparing with other orthogonal approaches (e.g., correlation
network analysis, eQTL) or with other genetic populations
(RIL and IL). By doing this, we can find the overlapping
associations by different methods or populations, increasing
the chance of detecting true positive associations.
3. After narrowing down candidate associations, we can check
SNP markers in the candidate genes in order to identify functional amino-acid substitution caused by polymorphism
variants.
4. Perform linkage disequilibrium (LD) analysis of the mapped
genomic region for the target trait, in order to reveal the
relationships between the causal SNPs with the lead SNP.
5. Check whether the trait levels are significantly different
between different informative haplotypes, potentially facilitating the identification of causal SNPs.
6. Generate a short list of candidate genes by incorporating
biological knowledge.
7. Experimentally validate candidate genes by transgenic plants or
in other genetic populations such as: (A) RIL mapping (see
Note 12), (B) IL mapping (see Note 13), or (C) analysis of
T-DNA insertional mutants (see Note 14).
In the following, we present several examples in order to demonstrate the analysis workflow of GWAS and its wide application in
exploring causal metabolic genes, as well as the merit of combined
use of network analysis with quantitative genetics:
A. Two loci (AOP and MAM) predominantly contribute to the
observed variations in aliphatic-glucosinolate formation in
Arabidopsis [32, 33]. The AOP locus on chromosome 4 regulates side-chain modifications while the MAM locus on chromosome 5 controls chain elongation in the aliphaticglucosinolate biosynthesis pathway. For the 18 annotated aliphatic glucosinolates in Arabidopsis rosettes, subsequent
GWAS results showed that almost all upstream and downstream aliphatic glucosinolates mapped to the MAM and/or
AOP loci in control (21
C and 150 μE m
À2 s
À1 ) and stress
conditions (32
C and darkness) (Fig. 1, see Note 15), confirming previous findings [32, 34]. In addition, the mapping
results indicate that most MAM-regulated upstream
Metabolomic Analysis of Natural Variation in Arabidopsis
399
traits and SNPs by applying mixed linear models (MLM),
including both fixed and random effects. We usually take individuals as random effects, allowing MLM to incorporate information about relationships among individuals. The results of
associations are usually demonstrated by Manhattan plots.
2. Identify candidate associations between traits and genes by
comparing with other orthogonal approaches (e.g., correlation
network analysis, eQTL) or with other genetic populations
(RIL and IL). By doing this, we can find the overlapping
associations by different methods or populations, increasing
the chance of detecting true positive associations.
3. After narrowing down candidate associations, we can check
SNP markers in the candidate genes in order to identify functional amino-acid substitution caused by polymorphism
variants.
4. Perform linkage disequilibrium (LD) analysis of the mapped
genomic region for the target trait, in order to reveal the
relationships between the causal SNPs with the lead SNP.
5. Check whether the trait levels are significantly different
between different informative haplotypes, potentially facilitating the identification of causal SNPs.
6. Generate a short list of candidate genes by incorporating
biological knowledge.
7. Experimentally validate candidate genes by transgenic plants or
in other genetic populations such as: (A) RIL mapping (see
Note 12), (B) IL mapping (see Note 13), or (C) analysis of
T-DNA insertional mutants (see Note 14).
In the following, we present several examples in order to demonstrate the analysis workflow of GWAS and its wide application in
exploring causal metabolic genes, as well as the merit of combined
use of network analysis with quantitative genetics:
A. Two loci (AOP and MAM) predominantly contribute to the
observed variations in aliphatic-glucosinolate formation in
Arabidopsis [32, 33]. The AOP locus on chromosome 4 regulates side-chain modifications while the MAM locus on chromosome 5 controls chain elongation in the aliphaticglucosinolate biosynthesis pathway. For the 18 annotated aliphatic glucosinolates in Arabidopsis rosettes, subsequent
GWAS results showed that almost all upstream and downstream aliphatic glucosinolates mapped to the MAM and/or
AOP loci in control (21
C and 150 μE m
À2 s
À1 ) and stress
conditions (32
C and darkness) (Fig. 1, see Note 15), confirming previous findings [32, 34]. In addition, the mapping
results indicate that most MAM-regulated upstream
Metabolomic Analysis of Natural Variation in Arabidopsis
399