Enhanced genetic analysis of type 1 diabetes by selecting variants on both effect size and significance, and by integration with autoimmune thyroid disease

Abstract

For polygenic traits, associations with genetic variants can be detected over many chromosome regions, owing to the availability of large sample sizes. The majority of variants, however, have small effects on disease risk and, therefore, unraveling the causal variants, target genes, and biology of these variants is challenging. Here, we define the Bigger or False Discovery Rate (BFDR) as the probability that either a variant is a false-positive or a randomly drawn, true-positive association that exceeds it in effect size. Using the BFDR, we identify new variants with larger effect associations with type 1 diabetes and autoimmune thyroid disease.

Thymocyte regulatory variant alters transcription factor binding and protects from type 1 diabetes in infants

“…Insertion/deletion variants (indels), e.g., rs138300818, are not included in the eQTLgen data base comprising of 31,684 whole blood samples. Nevertheless, 22 out of 24 SNPs in the 96% posterior probability credible set for AAD association, all in high LD with rs138300818 (D’=1, r 2 >0.89 in 1000G GBR individuals), were significant eQTLs for THEMIS ( p -values < 1.9×10 −34 ) and ECHDC1 ( p -values < 1.30×10 −8 ) gene expression in whole blood, such that the minor alleles associated with protection from early onset type 1 diabetes were associated with lower THEMIS and ECHCD1 expression ( Supplementary Table S8 ). Our previous analysis indicated that the AAD association signal co-localises with the THEMIS eQTL signal (5).…”

“…While multiple TFs were predicted to bind to rs113297984 flanking sequence with DeepBind, the rs113297984 alleles did not affect their binding affinity. rs113297984 is in LD with rs138300818 (D’=1, r 2 =0.96 in 1000Genomes GBR population), and thus, has a similar eQTL profile. rs113297984 had a PCHiC interaction with PTPRK (CHiCAGO score 5.12) in thymus, in addition to interaction with THEMIS in GM12878 cell line (CHiCAGO score 11.54).…”

Disentangling the direct and indirect effects of childhood adiposity on type 1 diabetes and immune-associated diseases: a multivariable Mendelian randomization study

“…We firstly applied our multivariable approach using a large number of childhood and adult body size instruments to T1D data analysed previously in the study by Censin et al (n=5,913 cases and n=8,828 controls). Results from this analysis were then validated using a recent large-scale meta-analysis of up to 15,573 cases and 158,408 controls 21 . Analyses were then repeated separately in each contributing cohort from this meta-analysis.…”

“…Additionally, we investigated the opposite direction of effect using the same univariable methods mentioned above to assess whether genetic liability towards T1D risk influences body size in both childhood and adulthood in turn. In this analysis we used a set of genetic instruments for T1D identified from a recent meta-analysis (of up to 15,573 cases and 158,408 controls 21 ) and adult BMI was analysed as a continuous trait to derive a per standard deviation effect estimates.…”

“…We also applied the multivariable MR Egger method to evaluate horizontal pleiotropy for the direct and indirect effects of childhood body size 33 . Furthermore, multivariable analyses were repeated using data from the large-scale T1D meta-analysis 21 , as well as evaluating evidence using data from each contributing cohort to this dataset in turn. Lastly, we repeated our multivariable MR analysis with childhood and adult body size as exposures onto each of the seven different types of immune-associated/autoinflammatory disease in turn.…”