Driving Accurate Allergen Prediction with Protein Language Models and Generalization-Focused Evaluation

2025-08-14

AI Summary

This study addresses three practical challenges in allergen protein identification: (1) zero-shot recognition of novel allergens absent from training data; (2) fine-grained discrimination between allergenic and non-allergenic proteins exhibiting high sequence similarity; and (3) assessing the impact of single-point mutations on allergenicity. To this end, we propose Applm, a computational framework that pioneers the integration of the billion-parameter xTrimoPGLM protein language model into allergen prediction—leveraging its strong generalizable representations learned from trillion-token pretraining—via transfer learning and sequence feature fusion. We further construct the first comprehensive benchmark tailored to real-world scenarios, covering zero-shot prediction, fine-grained classification, and mutation impact analysis. Extensive experiments demonstrate that Applm significantly outperforms seven state-of-the-art methods across seven challenging evaluation metrics. The code and benchmark dataset are publicly released.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Applications
• Multiagent Systems: Adversarial Agents

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Allergens, typically proteins capable of triggering adverse immune responses, represent a significant public health challenge. To accurately identify allergen proteins, we introduce Applm (Allergen Prediction with Protein Language Models), a computational framework that leverages the 100-billion parameter xTrimoPGLM protein language model. We show that Applm consistently outperforms seven state-of-the-art methods in a diverse set of tasks that closely resemble difficult real-world scenarios. These include identifying novel allergens that lack similar examples in the training set, differentiating between allergens and non-allergens among homologs with high sequence similarity, and assessing functional consequences of mutations that create few changes to the protein sequences. Our analysis confirms that xTrimoPGLM, originally trained on one trillion tokens to capture general protein sequence characteristics, is crucial for Applm's performance by detecting important differences among protein sequences. In addition to providing Applm as open-source software, we also provide our carefully curated benchmark datasets to facilitate future research.

Problem

• Accurate prediction of allergen proteins using language models
• Identifying novel allergens without similar training examples
• Differentiating allergens from non-allergens in similar sequences

Innovation

• Leverages xTrimoPGLM protein language model
• Outperforms seven state-of-the-art methods
• Detects important protein sequence differences