A genome-wide association (GWA) study that aimed at studying loci associted with childhood obesity was carried out with the Singapore Cohort study of the Risk factors for Myopia (SCORM) samples (n=1004, age 9) using the Illumina 550 SNP chip. After initial genotyping and QC of SNPs and samples, imputation based on HapMap East-Asian samples (Chinese and Japanese) was carried out to increase coverage of SNPs detected in this study. Subsequent association analysis on zscores of BMI revealed several interesting SNPs from genes such as the apolipoprotein B mRNA editing enzyme and the tubby gene, with top p-values in the range of 10-6. Association analysis were also conducted on 695 samples who had relevant measures at birth (birth-weight, birth-length and head circumference), to detect genetic loci associated with obesity at birth.

However, as only a modest sample size (n=1004 at age 9) was available from the SCORM dataset it could be likely that important SNPs fail to reach higher levels of significance and are thus missed from the analyses. As such performing a meta-analysis with the SCORM GWA dataset and the ALSPAC GWA dataset would be useful for both increasing power to detect SNPs that are associated with childhood obesity and possibly reveal more relevant SNPs when they turn up in both these children datasets.

The ALSPAC GWA data carried out with individuals at a similiarly young age (age 11, n=1500+) on the Illumina 330 SNP chip, with relevant obesity measures (height and weight to calculate BMI, birth weight and birth length) and suitable covariates (gender, mother's education status, father's education status) would be used for the meta-analysis with the existing SCORM BMI zscore association results. This ALSPAC dataset would be put through quality control measures to filter of samples based on on call-rate (less than 95%) and gender discrepancies. Subsequently, ethinic confounders would be removed using using principal component analyses. Similarly SNPs would be filtered off based on call-rate (less than 95%), minor allele frequency (less than 0.5%) and Hardy-Weinberg equilibrium (pval less than 1O-7). Subsequently, additional SNPs would be imputed based on HapMap samples to increase coverage of SNPs and overlap between the existing SCORM imputed dataset. Genotype calling for impuation would be based on a posterior probability threshold of 90%. To further ensure confidence of selecting imputed SNPs, those that are solely imputed from both datasets and without any typed SNPs in the region (1.25Mb from the imputed SNP in both datasets) would be filtered off.

SNPs that overlap in both the datasets would be used for detecting association with zscores of BMI and relevant measures at birth for the ALSPAC dataset. This analysis would be done using the linear trend test under various models (additive, recessive, dominant and genotypic) and controlled for relevant covariates such as gender, using the PLINK genome-wide association analysis toolset. Subsequently, individual SNP results (beta, standard error, variance and p-values) from SCORM and ALSPAC datasets would be extracted and used to derive pooled estimates and overall p-values assuming a fixed effect model, with the inverse variance-weighted method. The pooled estimate for the 2 studies would be derived using the formula ? = ∑ w*beta / ∑w (where ? = pooled estimate, w = inverse variance weight of each study and beta= individual beta of each study), which incoporates a weightage for each SNP from the individual study into the overall estimate. Measures of heterogeneity, Cochran's Q and I square statistics would also be generated to gauge the level of heterogeneity between the 2 studies for each SNP. Subsequently, gene annotation of top hits that turn up from this meta-analysis exercise would be carried out to reveal any functional effects the SNP may have on the gene product.

As a follow-up step to the meta-analysis, the remaining subset of ALSPAC samples at age 11 could be used for a replication study and confirm obesity association of possible novel hits that are revealed from the meta-analysis. A tagging SNP approach would be used to best cover the top hits (pval less than 10-4 from the meta-analysis) and the suitable set of SNPs could be genotyped in the Illumina Golden-gate mutiplex genotyping platform.

Concept Specific measures Source Time points

Obesity Anthropometry measures Children in Focus Birth to 12 months

Obesity -Anthropometry measures Focus Clinic Sessions Age 7 to 11

- Measures

-Food and activity

-Health utilities index

-nutrition