The colon cancer genetics group (CCGG) lead by Professor Malcolm Dunlop and Harry Campbell has performed a genome-wide association study (WGAS) of colorectal cancer (CRC) risk using the Illumina HumanHap300 and HumanHap240S chips. In addition, we have performed a WGAS for obesity (BMI) on a subset of the individuals used in the CRC study. A total of 1,268 unrelated individuals were tested for ~550,000 tag SNPs. The top hit from the GWAS had an additive genetic effect of ~1kg/m2 (P = 2 x 10-6). We have performed a replication phase and a fine-mapping phase of the region on a different colorectal cancer case-control cohort comprising over 3,400 unrelated individuals. The top SNP and others from the replication phase were significant (P less than 0.05) and had an additive genetic effect of 0.34 kg/m2 which explains 0.2% of the phenotypic variant. Joint analysis of phase 1 and 2 showed that the variant was also associated with weight and waist circunference but not with height (Table 1).

Table 1. Effect estimates and significance levels of the association for the CCGG.

Trait

P

Additive effect

(SE,P)

Dominant effect

(SE,P)

% phenotypic variance explained

Weight (kg)

3.1 x 10-5

1.6

(0.41, 0.0001)

0.31 (0.52,0.56)

0.45

Height (m)

0.77

-

-

-

Waist (cm)

2.2 x 10-5

1.28

(0.33, 8.7 x 10-5)

0.22

(0.41,0.60)

0.47

We have already genotyped two replication cohorts from Croatia (N=483) and the WTCCT2D study (N=10,278). The Croatian and WTCCT2D cohorts had an additive genetic effect ranging from 0.27 to 0.43 although the association was only replicated (P less than 0.05) in the second cohort.

We would like to genotype our top hit SNP on the ALSPAC 9,000 adult women and 9,000 children, obtain access to gender, age, height, weight, DXA body fat, lean BMI, waist circunference. If repeated measures are available (specially for children) we would very much like to access them and try to model differences in patterns of growth of the different genotypes using random regression models.

We would like the genotyping to be carried out at K Biosciences, as suggested by you, since this would undoubtely be cheaper, faster and more efficient.

The main analysis will be an ANOVA comparing the genotypic mean of the three genotypes after correcting for the relevant fixed effect (ie. Gender, age, etc) but more sofisticated analysis could be performed as suggested above.

Table 2 shows that if our finding is real the ALSPAC cohort has sufficient power to replicate it.

Finally, we hope we have provided you with enough details and that you will consider our proposal favorably.

Table 2. Power to detect a QTL with population allele frequency p = 0.74. The QTL heritability was assumed to be h2QTL=0.002. The genetic model was assumed dominant for the allele that increases genotypic value. Differences in mean genotypic values were tested using an ANOVA.

Significance level

Sample Size

0.05

0.005

0.0005

10,000

greater than 0.99

greater than 0.99

greater than 0.99

7,000

greater than 0.99

greater than 0.99

0.98

5,000

greater than 0.99

0.96

0.88

3,000

0.94

0.76

0.52

1,000

0.49

0.19

0.06

500

0.27

0.07

0.02