基于Transformer的促炎肽预测  

Prediction of Pro-Inflammatory Peptides Based on Transformer

作　　者：孙一平 Sun Yiping(School of Information and Electronic Technology,Jiamusi University,Jiamusi,Heilongjiang 154007,China)

机构地区：[1]佳木斯大学,黑龙江佳木斯154007

出　　处：《计算机时代》2025年第4期11-14,19,共5页Computer Era

摘　　要：提出了一种基于改进Transformer架构的深度学习模型,用于预测促炎肽,以解决传统实验方法周期长及特征表征不足的问题。模型通过整合多维度3-mer特征编码,构建改进的Transformer深度学习框架,基于去冗余数据集的五折交叉验证表明,改进模型在关键性能指标上显著优于基准方法MultiFeatVotPIP:准确率(ACC)提升11.4%至76.9%,曲线下面积(AUC)提高12.2%至80.8%。特征分析验证了3-mer特征在捕捉局部残基模式和跨域相互作用中的优势,为高通量促炎肽筛选和炎症疾病机制研究提供了新工具和新思路。This study proposes a deep learning model based on an improved Transformer architecture for predicting proinflammatory peptides,addressing the limitations of traditional experimental methods such as lengthy cycles and insufficient feature representation.The model integrates multi-dimensional 3-mer feature encoding and constructs an improved Transformer-based deep learning framework.Evaluated through five-fold cross-validation on a non-redundant dataset,the enhanced model demonstrates significant superiority over the benchmark method MultiFeatVotPIP in key metrics:accuracy(ACC)increases by 11.4%to 76.9%,and AUC improves by 12.2%to 80.8%.Feature analysis confirms the advantages of 3-mer features in capturing local residue patterns and cross-domain interactions.This work provides a novel tool and methodology for high-throughput screening of proinflammatory peptides and advances research into inflammatory disease mechanisms.

关 键 词：TRANSFORMER 促炎肽 深度学习 3-mer 特征编码 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程] R373[医药卫生—病原生物学]