Novel, computational IgE-clustering in a population-based cross-sectional study: Mapping the allergy burden

Abstract

Background: Even though the prevalence of allergies is increasing, population-based data are still scarce. As a read-out for chronic inflammatory information, new methods are needed to integrate individual biological measurements and lifestyle parameters to mitigate the consequences and costs of allergic burden for society.

Methods: More than 480.000 data points were collected from 1462 Luxembourg adults during the representative, cross-sectional European Health Examination Survey, spanning health and lifestyle reports. Deep IgE-profiles based on unsupervised clustering were correlated with data of the health survey.

Findings: 42.6% of the participants reported a physician-diagnosed allergy and 44% were found to be IgE-positive to at least one allergen or extract. The main sensitization sources were tree pollens followed by grass pollens and mites (52.4%, 51.8% and 40.3% of sensitized participants respectively), suggesting seasonal as well as perennial burden. The youngest group of participants (25-34 years old) showed the highest burden of sensitization, with 18.2% of them having IgE to 10 or more allergen groups. Unsupervised clustering revealed that the biggest cluster of 24.4% of participants was also the one with the highest medical need, marked by their multi-sensitization to respiratory sources.

Interpretation: Our novel approach to analyzing large biosample datasets together with health information allows the measurement of the chronic inflammatory disease burden in the general population and led to the identification of the most vulnerable groups in need of better medical care.

Keywords: European Health Examination Survey; allergy burden; multiplex IgE-profiles; population-based cross-sectional study.

FIGURE 1

(A) Workflow of the study and overview of the results. (B) The percentage of participants (N = 1462) reporting physician‐diagnosed allergy (“diagnosed group”) with recent symptoms (half‐tone colors) and symptoms ever in life (bright colors). Indicated are reported allergies of any kind (purple), nasal allergy (blue), skin allergy (green), eye allergy (yellow), food allergy (orange) and asthma (red). (C) The correlation matrix comparing allergy report and sensitization (diagnosed group, N = 623). Correlation of categorical values was calculated using Cramer's V. Correlation coefficients between 0 and 0.19 indicate no correlation and 0.2–0.29 weak, 0.30–0.49 moderate, 0.5–0.69 strong, and 0.7–1 very strong correlation. (D) Total IgE levels of participants reporting recent allergy symptoms (light gray; N = 332) versus those reporting allergy symptoms a long time ago (dark gray; N = 153), p‐value calculated using Mann‐Whitney testing. For all significance tests, the p‐values were set to: **** <0.0001, *** <0.001, ** <0.01, * <0.05 and ns ≥ 0.05.

FIGURE 2

The heatmap of most eliciting sources based on specific IgE levels and population reached, selected for specific IgE >5 kU_A/L for all participants with at least 1 allergen/extract >5 kU_A/L. Color grading within the heatmap indicates levels of specific IgE; darkest orange color signifies 5 kU_A/L, with the color getting lighter the higher the value is. Colored columns signify exposure route (green = outdoor exposure, yellow = indoor exposure, orange = food, blue = other exposure routes). Every column represents one allergen and every row indicates one participant, reaching this level of sensitization (N = 346). IgE, immunoglobulin E.

FIGURE 3

(A) Circus plot showing co‐sensitization to all allergen groups for different ages. Black boxes indicate sensitization to at least one allergen or extract in the group >0.3 kU_A/L, and light blue boxes indicate no specific IgE against any source in that group. Every radial line indicates one participant. Colored rings signify exposure route (green = outdoor exposure, yellow = indoor exposure, orange = food, blue = other exposure routes). (B) Bar plot of co‐sensitization to multiple allergen source groups divided by age. Purple indicates mono‐sensitization, dark blue sensitization to 2 different source groups, blue to 3, green to 4–9 sources and yellow indicates strong multi‐sensitization to 10 or more sources. IgE, immunoglobulin E.

FIGURE 4

(A) Unbiased clustering of sensitized participants (N = 643) based on their IgE signatures over 298 allergens/extracts by UMAP. Differently colored dots indicate different clusters. Percentage represents the percent of sensitized participants clustered together. (B) IgE binding characteristics differentiating between the seven IgE clusters. Columns indicate participants, rows are sorted and selected by a maximum of the 10 top most signifying allergens. Right: indicated allergens are the ones most significantly different between the cluster and all other clusters. Left: biological origins and groups of structural homologous allergens. Color code of the heatmap indicates normalized specific IgE levels, with higher levels being brighter yellow. IgE, immunoglobulin E; NPC2, Niemann pick type C2 protein; PR‐10, pathogenesis related class‐10 like protein family; UMAP, Uniform Manifold Approximation and Projection.

FIGURE 5

(A) Significant differences between IgE cluster 0 and the non‐sensitized population cohort on selected variables concerning health and lifestyle. Allergy diagnosis refers to all participants reporting physician‐diagnosed allergy. (B) Significant differences between IgE cluster 0 and the other sensitized cohort on variables concerning health and lifestyle. IgE, immunoglobulin E. For all significance tests, the p‐values were set to: **** <0.0001, *** <0.001, ** <0.01, * <0.05 and ns ≥ 0.05.