Protein language models improve acid-loving protein identification efficiency
MI-PEFT: Mixture-of-Experts Integrated Parameter-Efficient Fine-Tuning Protein Language Models Improves Acidophilic Proteins Classification
Machine Learning
Summary
Identifying acid-loving proteins, which work well in very acidic environments, is important but usually slow and experimental. The authors developed a new method called MI-PEFT that uses advanced machine learning on protein sequences to quickly and accurately find these special proteins. Their method combines clever tuning of a large protein model with a flexible approach that handles the difficulty of unbalanced data classes. This helps make better predictions while preserving useful learned information from big protein datasets.
What this means in practice
- •For biotech engineers: Enhance the speed and accuracy of identifying acid-stable enzymes suitable for industrial processes using computational protein analysis.
- •For bioinformatics teams: Implement efficient fine-tuning methods for protein classification tasks especially when data is imbalanced, improving model performance and resource use.
Authors
Honghan Shen
Abstract
Acidophilic proteins that remain stable and functional under highly acidic conditions, are important for industrial biocatalysis, acid-related bioprocessing, and the discovery of acid-stable enzymes. However, their identification relies heavily on time-consuming experimental screening methods. With the rapid growth of protein sequence databases, the need for computational identification methods that are both accurate and efficient has become stronger. The emergence of protein language models (PLMs) has significantly improved the sequence representation of downstream biological prediction tasks. This paper proposes MI-PEFT, a mixture-of-experts integrated parameter-efficient fine-tuning framework. Built on the ESM C-600M backbone, the framework incorporates LoRA-based PEFT methods and a DeepSeekMoE-based classification head to resolve the limitations of PEFT and significantly improve computational efficiency. Notably, this task is characterized by a significant class imbalance in the dataset, making high specificity particularly challenging. The experimental results demonstrate that MI-PEFT on PLMs, especially {\text{C}}^{\text{3}}\text{A}, serves as an efficient tool for identifying acidophilic proteins and a constrained pathway that helps resolve class-imbalance by preserving the pretrained representations.