Cansin Sackesen1,2,3, Mayte Suárez-Fariñas4, Ronaldo Silva4, Jing Lin3,5, Stephanie Schmidt6, Robert Getts6,7, Gustavo Gimenez3, Ebru A Yilmaz2, Ozlem Cavkaytar2, Betul Buyuktiryaki2, Ozge Soyer2, Galina Grishina3, Hugh A Sampson3. 1. Division of Pediatric Allergy, Koc University School of Medicine, Istanbul, Turkey. 2. Division of Pediatric Allergy, Hacettepe University School of Medicine, Ankara, Turkey. 3. Division of Pediatric Allergy and Immunology, Icahn School of Medicine at Mount Sinai, New York, New York. 4. Center for Biostatistics, Icahn Institute for Genomics and Multiscale Biology, Icahn School of Medicine at Mount Sinai, New York, New York. 5. Institute of High Performance Computing, A*STAR, Singapore City, Singapore. 6. Genisphere LLC, Hatfield, Pennsylvania. 7. AllerGenis LLC, Hatfield, Pennsylvania.
Abstract
BACKGROUND: The majority of children with cow's milk allergy (CMA) tolerate baked milk. However, reactivity to fermented milk products such as yogurt/cheese has not been previously evaluated. We sought to determine whether children with CMA could tolerate yogurt/cheese and whether a patient's IgE and IgG4-binding pattern to milk protein epitopes could distinguish clinical reactivity. METHODS: Four groups of reactivity were identified by Oral food challenge: baked milk reactive, fermented milk reactive, whole milk reactive, and outgrown. sIgE and sIgG4 binding to milk protein epitopes were assessed with a novel Luminex-based peptide assay (LPA). Using machine learning techniques, a model was developed to predict different degrees of CMA. RESULTS: The baked milk reactive patients demonstrated the highest degree of IgE epitope binding, which was followed sequentially by fermented milk reactive, whole milk reactive, and outgrown. Data were randomly divided into two groups with 75% of the data utilized for model development (n = 68) and 25% for testing (n = 21). All 68 children used for training were correctly classified with models using IgE and IgG4 epitopes. The average cross-validation accuracy was much higher for models using IgE plus IgG4 epitopes by LPA (84.8%), twice the performance of the serum component proteins assayed by UniCAP (41.9%). The performance of the model on "unseen data" was tested using the 21 withheld patients, and the accuracy of IgE was 86% (AUC = 0.89) while of IgE+IgG4 model was 81% (AUC = 0.94). CONCLUSION: Using a novel high-throughput LPA, we were able to distinguish the diversity of IgE/IgG4 binding to epitopes in the varying CMA phenotypes. LPA is a promising tool to predict correctly different degrees of CMA.
BACKGROUND: The majority of children with cow's milk allergy (CMA) tolerate baked milk. However, reactivity to fermented milk products such as yogurt/cheese has not been previously evaluated. We sought to determine whether children with CMA could tolerate yogurt/cheese and whether a patient's IgE and IgG4-binding pattern to milk protein epitopes could distinguish clinical reactivity. METHODS: Four groups of reactivity were identified by Oral food challenge: baked milk reactive, fermented milk reactive, whole milk reactive, and outgrown. sIgE and sIgG4 binding to milk protein epitopes were assessed with a novel Luminex-based peptide assay (LPA). Using machine learning techniques, a model was developed to predict different degrees of CMA. RESULTS: The baked milk reactive patients demonstrated the highest degree of IgE epitope binding, which was followed sequentially by fermented milk reactive, whole milk reactive, and outgrown. Data were randomly divided into two groups with 75% of the data utilized for model development (n = 68) and 25% for testing (n = 21). All 68 children used for training were correctly classified with models using IgE and IgG4 epitopes. The average cross-validation accuracy was much higher for models using IgE plus IgG4 epitopes by LPA (84.8%), twice the performance of the serum component proteins assayed by UniCAP (41.9%). The performance of the model on "unseen data" was tested using the 21 withheld patients, and the accuracy of IgE was 86% (AUC = 0.89) while of IgE+IgG4 model was 81% (AUC = 0.94). CONCLUSION: Using a novel high-throughput LPA, we were able to distinguish the diversity of IgE/IgG4 binding to epitopes in the varying CMA phenotypes. LPA is a promising tool to predict correctly different degrees of CMA.
Authors: Daniel R Monaco; Brandon M Sie; Thomas R Nirschl; Audrey C Knight; Hugh A Sampson; Anna Nowak-Wegrzyn; Robert A Wood; Robert G Hamilton; Pamela A Frischmeyer-Guerrerio; H Benjamin Larman Journal: Nat Commun Date: 2021-01-22 Impact factor: 17.694