Citation: Platts-Mills TAE, Perzanowski M (2018) The use of machine learning to understand the relationship between IgE to specific allergens and asthma. PLoS Med 15(11): e1002696. https://doi.org/10.1371/journal.pmed.1002696
Published: November 20, 2018
Copyright: © 2018 Platts-Mills, Perzanowski. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Abbreviations: IgE, Immunoglobulin E; RAST, Radioallergosorbent test
Provenance: Commissioned; not externally peer reviewed.
In a study recently published in PLOS Medicine, Custovic and colleagues  report an elegant analysis of the results of specific Immunoglobulin E (IgE) assays on sera from a birth cohort that they have studied in detail over the last 15 years. The IgE assays used a microchip with >100 purified allergens, and their machine learning approach used a hypothesis-free statistical approach to group children with similar allergen sensitization profiles. The primary objective was to evaluate the relationship of the results to the diagnosis of current asthma.
The goal of employing a “machine learning” approach is both as a form of data reduction to avoid spuriously identified associations resulting from multiple comparisons and to identify biologically relevant groupings that may not be recognized with conventional approaches. The use of “hypothesis free” clustering methods with individual allergen components in itself is not novel [2,3]. However, the authors have employed a new network analysis method, which is improved over previous methods because it can capture nonlinear relationships and does not rely on assumption of a parametric probability. As the authors show, this approach can lead to improved sensitivity and specificity. With this model, they were able to show that pairings of allergens (e.g., the cat allergen Fel d 1 and the peanut allergen Ara h 1) were more predictive of the child having asthma than individual components, although we must be cautious in the generalizability of the specific patterns identified, in this single cohort in the United Kingdom, because the pairing may represent host susceptibility to making an IgE response or common co-exposures, which could be allergens or adjuvants. Still, the concept of complicated connections between individual allergenic proteins and allergic disease are likely to apply more broadly and are important to consider in future studies.
In vitro assays for IgE go back to 1967, when the Radioallergosorbent test (RAST) was published within a year of the discovery of IgE. [4–6] The great-grandchild of RAST is the current most widely recognized assay for IgE, which uses a high-capacity sponge, can be used either with an allergen extract or a purified allergen, and can bind micrograms of protein. In contrast to the “singleplex” assays, microchip assays have been developed to carry large numbers of purified proteins. The ISAC chip used in the current study has purified proteins on spots with nanogram quantities. The microchip gives results in standardized units that are similar to, but not the same as, the absolute units where one unit is equal to 2.4 ng of IgE. Overall, the microchip is less sensitive and can definitely be blocked by the presence of high-titer IgG antibodies in the serum. This is a real problem in the evaluation of IgE to milk proteins in eosinophilic esophagitis but is unlikely to be relevant in relation to asthma in childhood . In relation to asthma, only occasional allergens will give rise to IgG antibodies sufficient to inhibit the microchip results. However, immunotherapy can certainly achieve high-titer IgG antibodies. The inevitable result of having 100 proteins is that many of them are both structurally similar and also immunologically cross-reactive. An excellent example is the lipocalins, in that the horse protein Equ c 1 can appear to be a dominant allergen when the primary exposure is to the cross-reactive dog allergen Can f 1 . Equally, the pathogenesis-related proteins cross-react extensively, which means that many patients who are primarily allergic to birch pollen (Bet v 1) experience oral symptoms from eating apples (Mal d 1) or hazelnuts (Cor a 1). However, of these, only Bet v 1 has been associated with asthma.
The authors point out that some asthma guidelines do not recommend allergy testing, and they argue that better understanding of the relationship of specific allergen IgE responses to asthma is necessary. They certainly demonstrate that their analysis of the microchip assay results in a striking prediction of current asthma diagnosis in the birth cohort. However, skin testing or IgE assays usually occurs in children when the diagnosis is already clear or very likely from history and examination. The primary objective at that point is to inform the physician, to help refine the diagnosis, and to educate the patient. Of these, educating the patients may be the most important. Two years ago, this same group of investigators in the UK published an important study showing that measures to reduce house dust mite allergens in the home could decrease the risk of acute asthma episodes in childhood . Equally, a major inner city study demonstrated a significant effect of allergen avoidance in asthma . In each case, the measures recommended were based on a specific diagnosis of sensitization. Furthermore, there is increased evidence that the titer of IgE antibodies is an important predictor of the persistence of asthma, and also of the probability of an acute episode of asthma following a rhinovirus infection [8,11]. In the analysis reported by Custovic and colleagues, one Alternaria allergen was included in their Fig 3, but this allergen, Alt a 1, does not appear in the cluster analysis in their Table 1, and no Aspergillus allergens are included. The implication is that no fungal allergens are relevant to asthma at this age. In other analyses at the same age, Aspergillus sensitization—though not common—was shown to have the strongest association with asthma . In older patients, Alternaria sensitization has often been shown to have a strong relationship to severity . In our practice, an initial screen with 10 allergens may indicate the need for more focused evaluation of fungi, or even foods. Although the analysis presented argues cogently that the data can provide better information to help with diagnosis, both the quantity of data and the details of the analysis would not be helpful in educating either patients or their doctors. The microchip assays have the potential to provide extensive information about a wide range of allergic disease using a modest quantity of serum, approximately 80 μL. The authors have shown that careful analysis of the results can provide dramatically enhanced data on the relevance of the results to a major allergic disease. The way forward may depend on designing more specific selections of allergens on a microchip and microchips designed for different diseases.
- 1. Fontanella S, Frainay C, Murray CS, Simpson A, Custovic A. Machine learning to identify pairwise interactions between specific IgE antibodies and their association with asthma: A cross-sectional analysis within a population-based birth cohort. PLoS Med. 2018;15(11):e1002691.
- 2. Prosperi MC, Belgrave D, Buchan I, Simpson A, Custovic A. Challenges in interpreting allergen microarrays in relation to clinical symptoms: a machine learning approach. Pediatr Allergy Immunol 2014;25:71–9. pmid:24131308. PMCID: PMC4282342.
- 3. Chen Q, Zhong X, Acosta L, Divjan A, Rundle A, Goldstein IF, et al. Allergic sensitization patterns identified through latent class analysis among children with and without asthma. Ann Allergy Asthma Immunol 2016;116:212–8. pmid:26945495. PMCID: PMC4784103.
- 4. Wide L, Bennich H, Johansson SG. Diagnosis of allergy by an in-vitro test for allergen antibodies. Lancet 1967;2:1105–7. pmid:4168552.
- 5. Bennich HH, Ishizaka K, Johansson SG, Rowe DS, Stanworth DR, Terry WD. Immunoglobulin E: a new class of human immunoglobulin. Immunology 1968;15:323–4. pmid:4176102. PMCID: PMC1409482.
- 6. Platts-Mills TAE. Dr. Kimishige Ishizaka 1926–2018 The discovery of IgE and the revolution in the study of allergic disease. Ann Allergy Asthma Immunol 2018 Oct 1. pii: S1081-1206(18)31287-0. pmid:30287256.
- 7. Schuyler AJ, Wilson JM, Tripathi A, Commins SP, Ogbogu PU, Kruzsewski PG, et al. Specific IgG4 antibodies to cow's milk proteins in pediatric patients with eosinophilic esophagitis. J Allergy Clin Immunol 2018;142:139–148 e12. pmid:29678750.
- 8. Perzanowski MS, Ronmark E, James HR, Hedman L, Schuyler AJ, Bjerg A, et al. Relevance of specific IgE antibody titer to the prevalence, severity, and persistence of asthma among 19-year-olds in northern Sweden. J Allergy Clin Immunol 2016;138:1582–1590. pmid:27430609. PMCID: PMC5405771.
- 9. Murray CS, Foden P, Sumner H, Shepley E, Custovic A, Simspon A. Preventing Severe Asthma Exacerbations in Children. A Randomized Trial of Mite-Impermeable Bedcovers. Am J Respir Crit Care Med 2017;196:150–158. pmid:28282501.
- 10. Morgan WJ, Crain EF, Gruchalla RS, O'Conner GT, Kattan M, Evans R 3rd, et al. Results of a home-based environmental intervention among urban children with asthma. N Engl J Med 2004;351:1068–80. pmid:15356304.
- 11. Soto-Quiros M, Avila L, Platts-Mills TA, Hunt JF, Erdman DD, Carper H, et al. High titers of IgE antibody to dust mite allergen and risk for wheezing among asthmatic children infected with rhinovirus. J Allergy Clin Immunol 2012;129:1499–1505 e5. pmid:22560151. PMCID: PMC3792652.
- 12. Sears MR, Herbison GP, Holdaway MD, Hewitt CJ, Flannery EM, Silva PA. The relative risks of sensitivity to grass pollen, house dust mite and cat dander in the development of childhood asthma. Clin Exp Allergy 1989;19:419–24. pmid:2758355.
- 13. O'Hollaren MT, Yunginger JW, Offord KP, Somers MJ, O'Connell EJ, Ballard DJ, et al. Exposure to an aeroallergen as a possible precipitating factor in respiratory arrest in young patients with asthma. N Engl J Med 1991;324:359–63. pmid:1987459.