Framingham FTO Analysis
Mao Ding
2024-03-24
Background:
This analysis is based on the finding that “genetic variants within FTO confer different effects on BMI variation and obesity depending on transmission from the mother or the father” from the prior paper: Indications for Potential Parent-of-Origin Effects within the FTO Gene https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0119206#pone-0119206-t001
The primary essence of parent-of-origin association based on Sorbs data is as follows:
10 variants show differences in association when evaluating paternal or maternal alleles, separately.
Three of the 10 variants (rs1861868, rs1121980, rs9939973; all intron 1) confer different effect directions between paternal and maternal alleles. And the three variants confer different genetic effects (P≤0.05) depending on parental origin.
Two of them (rs1121980 and 9939973) show stronger, albeit still non-significant, relationships when considering alleles paternally transmitted compared to a standard association test setting (P≤0.05).
The strongest association signals for BMI were detected in intron 3 (rs17818920 and rs17818902) involving 948 Sorbs individuals. These two variants also found the strongest association for parent origin.
Among five SNPs in intron 3, we observed the smallest p-values for two variants(rs17818920 and rs17818902) in standard association tests.
rs1121980 and rs9939973 are in high LD with rs8050136, one of the well replicated variants strongly associated with BMI
Objective
Examine the association between parental genotype at ten key FTO variants and offspring BMI, using data from the Framingham cohort.
Table 1. Framingham Cohort Characteristics
The distribution of individual BMI exhibits a slight skew, so use a log
transformation to approximate a normal distribution.
| Characteristic | N = 8,4321 |
|---|---|
| IDTYPE | |
| Generation 3 Cohort | 3,847 (46%) |
| New Offspring Spouse Cohort | 95 (1.1%) |
| Offspring Cohort | 3,536 (42%) |
| Original Cohort | 954 (11%) |
| SEX | |
| Female | 4,555 (54%) |
| Male | 3,877 (46%) |
| AGE | 63.44 (15.01) |
| BMI | 28.12 (5.66) |
| Unknown | 2 |
| 1 n (%); Mean (SD) | |
Summary of Linear Model (log(BMI) ~ SEX + AGE + Individual_Geno)
For each SNP, perform linear regression to assess the relationship between individual log-BMI and individual genotype, without accounting for parent-of-origin effects. The ANOVA table presented below indicates that rs11211980 and rs9939973 exhibit statistically significant p-values.
ANOVA Table (use monozygotes as reference group)
This ANOVA table contrasts the individual genotype effects with the reference level for each SNP. A small p-value highlighted in green indicates that rs1861869, rs1861868, rs1121980, and rs9939973 exhibit one or two robust genotype effects compared to the reference genotype. In other words, the estimated mean difference between each genotype effect and the reference is statistically significantly different from 0.
Post Hoc Test: TukeyHSD Test for Pairwise Differences Comparison between Heterozygotes
To explore the variance in BMI gain among individuals with different heterozygotes, we employed Tukey’s HSD test to compare pairwise differences between heterozygotes for each SNP. Unfortunately, none of the results indicate a significant difference.