Accurate Diagnosis of Diabetes for Appropriate Management

Aim 1. The investigators will recruit 5,000 cases diagnosed as T1D under the age of 25, from 17 participating clinics across Canada. All cases will be tested for four antibodies (against proinsulin, GAD65, islet antigen 2 (IA-2), and ZnT8). Cases negative for all four will be exome-sequenced.

1. Variant annotation will be focused on known monogenic diabetes genes. Variants rated as pathogenic, likely pathogenic or of unknown significance whose zygosity fits the genetic model, will be confirmed in a clinically certified laboratory and communicated to the treating health care team. End-point is the frequency of such variants compared to their frequency in control, non-T1D exomes.
2. The following variables will be examined for the ability to predict monogenic diabetes: Negativity for all autoantibodies tested, family history, polygenic T1D risk score, age of onset, sex, glycosylated hemoglobin (HbA1c), insulin dose, and presence of syndromic features. Predictors will be analyzed by multiple regression and results subjected to jackknife (leave-one-out) validation. Machine-learning techniques may be used.

Aim 2. Variants outside known genes in non-diagnostic exomes will be annotated and examined under autosomal dominant, recessive, X-linked and mitochondrial inheritance models. Corresponding frequency cutoffs will be 0.0005, 0.01, 0.001 and 0.0005 (if heteroplasmy >70%). Formal mutation-burden analysis will be based on depth-adjusted data from the Genome Aggregation Database (gnomAD). Genes mutated in more than one unrelated proband will be examined by a statistical approach taking into account the presence of a large number of phenocopies (Akawi et al., Nat Genet. 2015;47:1363-1369). Genes that achieve statistical significance will be tested in additional cohorts with international collaborations.