An algorithm for the classification of mRNA patterns in eosinophilic esophagitis: Integration of machine learning

Area of Science:

• Computational biology
• Immunology
• Gastroenterology

Background:

• Diagnostic evaluation of eosinophilic esophagitis (EoE) is challenging, particularly assessing allergic status.
• Current diagnostic methods for EoE require improvement in precision and efficiency.

Purpose of the Study:

• To develop an automated medical algorithm using machine learning to aid in EoE diagnosis.
• To establish a diagnostic probability score for EoE (p(EoE)) and an esophageal allergy score (IGHE).

Main Methods:

• Machine learning, including random forest classification, was applied to esophageal mRNA transcript patterns.
• Dimensionality reduction and weighted factor analysis were used to create the p(EoE) score.
• Esophageal IgE production was quantified using epsilon germ line (IGHE) transcripts to develop the IGHE score.

Main Results:

• The p(EoE) score demonstrated high accuracy in identifying EoE (90.9% sensitivity, 93.2% specificity, AUC 0.985).
• The algorithm improved the diagnosis of equivocal cases by 84.6% and showed responsiveness to therapy.
• An IGHE score ≥37.5 identified a subpopulation of EoE patients with increased allergic inflammation.

Conclusions:

• Intelligent data analysis via machine learning offers significant potential for enhancing EoE diagnostic precision.
• The p(EoE) and IGHE scores represent advancements toward decision trees for defining EoE subpopulations.
• These scores facilitate the development of individualized therapy strategies for EoE patients.