基于生成对抗网络的噬菌体尾部蛋白序列设计
Design of Phage Tail Protein Sequences Based on Generative Adversarial Network
林楷煌 1杜智华1
作者信息
- 1. 深圳大学计算机与软件学院,广东 深圳 518000
- 折叠
摘要
生成对抗网络(GAN)在图像生成、蛋白质设计领域有着广泛应用,但是对于噬菌体尾部蛋白的生成少有研究.提出一种基于GAN的噬菌体尾部蛋白序列生成方法.首先使用Wasserstein距离作为模型的目标函数.其次采用多层感知机(MLP)作为模型的基本结构.然后将 MLP 扩展为多路径结构.实验结果表明,上述方法取得了 0.9241 的质量得分、0.8498 的多样性得分和1.7739 的总得分,优于其它常用的生成方法.相较于单路径MLP,多路径MLP提高了序列的生成效果.所提方法能够生成高质量噬菌体尾部蛋白序列,同时保证生成序列的多样性.
Abstract
Generative Adversarial Network(GAN)is widely used in the fields of image generation and protein de-sign.However,there are only a few researches on the generation of phage tail proteins.The paper proposes a tail pro-tein sequences generative method based on GAN.Firstly,Wasserstein distance is chosen as the model objective func-tion.Secondly,Multi-Layer Perceptron(MLP)is used as the basic structure in the model.Then,MLP is extended to a multi-path structure.The experimental results show that this method achieves a quality score of 0.9241,a diversity score of 0.8498 and a total score of 1.7739,which is superior to other commonly used generative approaches.Com-pared with single-path MLP,the multi-path MLP improves the sequence generative effect.This method can generate high-quality phage tail protein sequences while ensuring the diversity of the generative sequences.
关键词
深度学习/生成对抗网络/蛋白质设计/多层感知机Key words
Deep learning/Generative Adversarial Network/Protein design/Multi-Layer Perceptron引用本文复制引用
基金项目
国家重点研发计划(2020YFA0908700)
国家自然科学基金面上项目(62176164)
深圳市科技计划(GGFW2018020518310863)
出版年
2024
NETL
NSTL
万方数据