Identifying patients with primary immunodeficiency using statistical modelling of unique genetic signatures from transcriptomic data

Abstract

Listen

Translate article

Primary immunodeficiencies (PID) are a group of rare disorders, most typically characterised by frequent and/or severe infections that are often caused by a genetic mutation. Early diagnosis is critical for patients as diagnostic delay is associated with increased morbidity and mortality. This study evaluated a novel diagnostic test for primary antibody deficiency, using whole blood from patients diagnosed with predominantly antibody deficiency (n = 15) as well as age-matched healthy donors (n = 18). PID patients comprised both adults (22–65 years, n = 18) and children (ages 6–18, n = 15), with 40% male (n = 13), 59% female (n = 19), and 1% gender not specified (n = 1). Inclusion criteria for patients were as follows: a) had an antibody deficiency b) an underlying in-born error of immunity and c) were under immunoglobulin replacement therapy.We used RNA-seq to characterise the transcriptome of each participant. We identified >2000 genes which were significantly differentially expressed between PID patients and healthy controls, with a p-value < 0.05. We used these signatures to train and test models to differentiate patients with PID from the healthy population. Additionally, in our blinded analysis, we identified the most variable genes in this cohort that were distinct from the differentially expressed genes and also evaluated their predictive utility. Using groups of differentially expressed genes and most variable genes in varying quantities, we generated multiple predictive models with an AUC > 0.9 for detection of primary immunodeficiency using leave-out-one cross validation.Our predictive algorithm using a transcriptomic signature, which we refer to as PrimDx, can help to identify patients with PID. This has the potential to accelerate time to diagnosis and initiation of treatment, thereby leading to better patient health outcomes. Future studies will verify this predictive algorithm in larger patient samples, and expand the application of this diagnostic approach in patients with other immunodeficiency conditions. [Display omitted]