Classification of adsorbed protein static ToF-SIMS spectra by principal component analysis and neural networks
Classification of adsorbed protein static ToF-SIMS spectra by principal component analysis and neural networks
复制标题
DOI：
10.1002/sia.1438
复制
发表时间：
2002-09-01
影响因子：
1.7
通讯作者：
Vickerman, JC
中科院分区：
文献类型：
--
作者：
Sanni, OD;Wagner, MS;Vickerman, JC
报道了利用主成分分析（PCA）以及一种新的人工神经网络（ANN）方法——NeuroSpectraNet对吸附蛋白质膜的静态飞行时间二次离子质谱（ToF - SIMS）谱进行分析，以对蛋白质膜的谱图进行化学分类。报道并评估了每种方法对13种不同蛋白质吸附膜的正离子谱进行分类的易用性和效率。由于不同蛋白质的谱图中缺乏独特的峰，吸附蛋白质膜的ToF - SIMS谱图尤其难以分析。尽管PCA能够利用氨基酸碎片产生的离子成功区分吸附蛋白质膜的ToF - SIMS谱图，但利用整个谱图进行区分却未成功。尽管仅使用氨基酸特异性离子，但几个蛋白质组中的异常值使得未知谱图的分类变得困难。然而，在将矢量分析增强纳入神经网络后，NeuroSpectraNet利用整个正离子谱图成功地对11种蛋白质膜的谱图进行了分类。通过使用正离子和负离子谱图的组合，实现了对所有13种蛋白质的完全分类。然而，与PCA一样，当输入模式仅包含氨基酸特异性离子时，ANN分类得到了增强。静态SIMS谱图的复杂性和多变量性质是一个非常适合应用神经网络进行模式识别和分类的领域。版权所有（C）2002约翰威立父子有限公司
Analysis of the static time-of-flight secondary ion mass spectrometry (ToF-SIMS) spectra of adsorbed protein films is reported using principal component analysis (PCA) and a novel artificial neural network (ANN) approach, NeuroSpectraNet, to classify chemically the spectra of the protein films. The ease of application and the efficiency with which each approach classified positive ion spectra from adsorbed films of 13 different proteins is reported and assessed. The ToF-SIMS spectra of adsorbed protein films are especially difficult to analyze owing to the absence of unique peaks in the spectra of different proteins. Although PCA was able to differentiate successfully ToF-SIMS spectra of adsorbed protein films using the ions generated from the fragmentation of the amino acids, differentiation of the spectra using the entire spectrum was unsuccessful. Outliers in several of the protein groups make classification of unknown spectra difficult, despite the use of only amino-acid-specific ions. However, NeuroSpectraNet successfully classified the spectra from 11 of the protein films using the whole positive ion spectra after a vector analysis enhancement had been incorporated into the neural network. Full classification of all 13 proteins was achieved by using the combined positive and negative ion spectra. However, as with PCA, ANN classification was enhanced when the input patterns only contained amino-acid-specific ions. The complex and multivariate nature of static SIMS spectra is a domain well suited to the application of neural networks for pattern recognition and classification. Copyright (C) 2002 John Wiley Sons, Ltd.