目的: 正确预测噬菌体病毒蛋白(phage virion proteins, PVPs)不仅可以深入了解噬菌体与宿主之间的关系,而且有助于开发新型抗菌药物。方法: 提取三种特征参数:氨基酸单肽组分信息(amino acid composition, AAC)、氨基酸二肽与预期平均值的偏差信息(dipeptide deviation from expected mean, DDE)和蛋白质相似性测量信息(protein similarity measure, PSM),采用混合采样方法对数据集进行平衡处理,对维度过高的特征通过方差分析进行降维,利用支持向量机算法,在Jackknife检验下预测噬菌体病毒蛋白(phage virion proteins, PVPs)。结果: 噬菌体病毒蛋白(phage virion proteins, PVPs)采用SMOTETomek方法平衡数据集后单特征参数最高预测成功率为96.41%,降维后预测成功率为97.07%,最后对特征参数进行融合,融合后最高预测成功率为97.29%。结论: 对不平衡的数据集采用混合采样方法、利用方差分析方法降维、对特征参数进行适当的融合,均能够有效提高预测成功率。
Objective: Accurately predicting phage viral proteins (PVPs) not only provides insights into the interactions between phages and their hosts, but also contributes to the development of novel antibacterial drugs. Methods: Three characteristic parameters were extracted: amino acid composition (AAC), dipeptide deviation from expected mean (DDE) and protein similarity measure (PSM). A hybrid sampling method was employed to balance the dataset, and dimensionality reduction was performed on high-dimensional features through variance analysis. The support vector machine algorithm was utilized to predict bacteriophage viral proteins (PVPs) under Jackknife testing. Results: For phage viral proteins (PVPs), after balancing the dataset using the SMOTETomek method, the highest prediction accuracy achieved with a single feature parameter was 96.41%. After dimensionality reduction, the prediction accuracy increased to 97.07%. Finally, the feature parameters were fused, resulting in the highest prediction accuracy of 97.29%. Conclusion: Employing a mixed sampling approach for unbalanced datasets, utilizing the analysis of variance method for dimensionality reduction, and appropriately integrating feature parameters can all effectively enhance the success rate of predictions.
[1] Clark JR, March JB. Bacteriophages and biotechnology: vaccines, gene therapy and antibacterials[J]. Trends Biotechnol, 2006, 24(5): 212-218.
[2] Auslander N, Gussow AB, Benler S, et al. Seeker: alignment-free identification of bacteriophage genomes by deep learning[J]. Nucleic Acids Res, 2020, 48(21): e121.
[3] 李新颖, 苏畅, 郭超, 等. CRISPR技术在链霉菌细胞工厂中的应用和优化[J]. 化工学报, 2025, 76(3): 922-932.
[4] Zhang YQ, Li ZY. RF phage virion:Classification of phage virion proteins with a random forest model[J]. Front Genet, 2023, 13: 1103783.
[5] Weinheimer AR, Aylward FO, Leray M, et al. Contrasting drivers of abundant phage and prokaryotic communities revealed in diverse coastal ecosystems[J]. ISME Commun, 2023, 3(1): 102.
[6] Aguilar PV, Adams AP, Wang E, et al. Structural and nonstructural protein genome regions of eastern equine encephalitis virus are determinants of interferon sensitivity and murine virulence[J].J Virol, 2008, 82(10): 4920-4930.
[7] 韩海涛. 整合序列和进化信息预测细菌Ⅳ型效应蛋白和噬菌体病毒蛋白[D]. 上海:上海海洋大学, 2022.
[8] 吴加伟. 噬菌体病毒粒子蛋白识别研究及其基因组注释平台构建[D]. 合肥:安徽大学, 2023.
[9] 张会云, 黄鹤鸣. 面向不平衡数据集的语音情感识别研究[J]. 计算机工程与应用, 2024, 60(4): 122-132.
[10] 张凝一. 基于多特征融合的内分泌疾病细胞功能识别方法研究[D]. 哈尔滨:哈尔滨工业大学, 2023.
[11] Feng ZP. Prediction of the subcellular location of prokaryotic proteins based on a new representation of the amino acid composition[J]. Biopolymers, 2001, 58(5): 491-499.
[12] Carr K, Murray E, Armah E, et al. A rapid method for characterization of protein relatedness using feature vectors[J]. PLoS One, 2017, 5(3): e9550.
[13] Vijayakumar S, Namasivayam G. Harnessing computational biology for exact linear B-cell epitope prediction: a novel amino acid composition-based feature descriptor[J]. Omics, 2015, 19(10): 648-658.
[14] Deng M, Yu CL, Liang Q, et al. A novel method of characterizing genetic sequences: genome space with biological distance and applications[J]. PLoS One, 2011, 6(3): e17293.
[15] Yuan SS, Gao D, Xie XQ, et al. IBPred: a sequence-based predictor for identifying ion binding protein in phage[J]. Comput Struct Biotechnol J, 2022, 20(7): 4942-4951.
[16] Feng PM, Ding H, Chen W, et al. Naive Bayes classifier with feature selection to identify phage virion proteins[J]. Comput Struct Biotechnol J, 2013, (2): e201302012.
[17] Charoenkwan P, Nantasenamat C, Hasan M, et al. Meta-iPVP:a sequence-based meta-predictor for improving the prediction of phage virion proteins using effective feature representation[J]. J Comput Aided Mol Des, 2020, 34(10): 1105-1116.
