内蒙古民族大学学报（自然科学版）2025，Vol.40Issue(4)：51-56,6.DOI:10.14045/j.cnki.15-1220.2025.04.008
基于深度时空网络和双重特征融合学习预测转录因子结合位点
Prediction of Transcription Factor Binding Sites Based on Deep Spatial Temporal Network and Dual Feature Fusion Learning
摘要
Abstract
Accurately identifying the transcription factor binding sites(TFBSs)and DNA plays a crucial role in understanding of cell functions and gene regulatory processes.In view of the limitations of existing methods in the mining of deep features of DNA sequences,a deep learning model with a simple-structured is proposed.This model designs a dual feature fusion mechanism to predict the TFBSs by integrating sequence features and hybrid spatial temporal features.Meanwhile,a lightweight three-layer 2D convolutional neural network is designed to be connected after the temporal convolutional network(TCN)to learn more expressive hybrid spatial temporal features,and these hybrid features are fused with the sequence features output by TCN to achieve the TFBSs prediction.Experimental results show that our method achieves good results in various evaluation metrics on 165 ChIP-seq datasets.关键词
转录因子结合位点/时间卷积网络/序列特征/空间特征/特征融合Key words
transcription factor binding site/temporal convolutional network/sequence feature/spatial feature/feature fusion分类
信息技术与安全科学引用本文复制引用
吴志强,罗蕊,姜静清..基于深度时空网络和双重特征融合学习预测转录因子结合位点[J].内蒙古民族大学学报（自然科学版）,2025,40(4):51-56,6.基金项目
国家自然科学基金项目(62162050) （62162050）
内蒙古自治区自然科学基金项目(2021BS03036) （2021BS03036）
蓖麻产业技术创新内蒙古自治区工程研究中心开放课题(MDK2021004,MDK2023012) （MDK2021004,MDK2023012）
内蒙古自治区蓖麻产业协同创新中心开放课题(MDK2022016) （MDK2022016）
内蒙古民族大学博士科研启动基金项目(BS672) （BS672）
