Scholarlink

This study presents a deep learning framework for predicting the inhibitory activity of antimicrobial peptides (AMPs), with a focus on estimating the minimal inhibitory concentration (MIC) against Escherichia coli. To address the growing threat of antibiotic resistance, we integrate ProtBert-BFD—a transformer model pretrained on a large corpus of protein sequences—into a fully connected neural network (FCNN). ProtBert enables the extraction of contextualized embeddings that capture long-range dependencies in peptide sequences. These embeddings are then used to train the FCNN for MIC regression. We validate our model using a curated dataset of MIC and pMIC values for E. coli, achieving improved predictive accuracy compared to traditional machine learning baselines. The results demonstrate the effectiveness of transformer-based representations in capturing biologically relevant sequence features. This approach provides a valuable tool for AMP screening and opens new directions for data-driven antimicrobial peptide discovery.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Novel Deep Learning Framework for Predicting Antimicrobial Peptide Activity Using ProtBert and Neural Networks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Maryam Abbasi,
  • Verónica Vasconcelos,
  • Edgar M. C. O. S. Vicente,
  • Ana L. M. Santos,
  • Joel P. Arrais

摘要

This study presents a deep learning framework for predicting the inhibitory activity of antimicrobial peptides (AMPs), with a focus on estimating the minimal inhibitory concentration (MIC) against Escherichia coli. To address the growing threat of antibiotic resistance, we integrate ProtBert-BFD—a transformer model pretrained on a large corpus of protein sequences—into a fully connected neural network (FCNN). ProtBert enables the extraction of contextualized embeddings that capture long-range dependencies in peptide sequences. These embeddings are then used to train the FCNN for MIC regression. We validate our model using a curated dataset of MIC and pMIC values for E. coli, achieving improved predictive accuracy compared to traditional machine learning baselines. The results demonstrate the effectiveness of transformer-based representations in capturing biologically relevant sequence features. This approach provides a valuable tool for AMP screening and opens new directions for data-driven antimicrobial peptide discovery.