卤代化合物的碳—13化学位移模拟

用人工神经网络和多元回归方法对含２个碳的２１个卤代化合物的３５个化学位移进行计算机图像模拟，结果表明，人工神经网络方法优于多元回归方法，同时此种方法处理这类问题有明显的优势，波谱模拟技术在有机化合物结构解析中是非常有用的方法。     

人工神经网络方法预测气相色谱保留值

本文运用一典型的人工神经网络模型－“反向传播“模型的改进形式，研究了诱导效应指数Ｉ，摩尔折射度Ｒｏ，疏水亲脂参数ＩｇＰ，以及分子联通性指数与气象色谱保留行为的关系，实现了对色谱保留植的预测。神经网络预测模型的最大相对误差不超过８．７％。结果表明，该方法性能良好，可望成为色谱保留值预测的有效手段。     

DNA与目标分子相互作用的预测性研究

以量化的分子结构参数和实验结果为依据, 运用模式识别技术、多元线性回归和人工神经网络研究了目标分子与DNA相互作用的主要影响因素, 建立了准确性较高的2个键合常数预测模型和1个作用模式预测模型. 初次量化的分子结构参数有21种, 经过筛选发现其中的10种参数对相互作用有显著影响. 研究结果可为抗癌药物的分子设计和筛选提供有价值的信息.     

用人工神经网络处理谷物成分分析

本文用人工神经网络处理谷物的付立叶变换近红外漫反射光谱,对谷物中含量在10~(-1)～10~(-3)的蛋白质、脂肪和6种人体必需氨基酸定量分析数据进行了解析,分析结果与经典化学方法没有系统偏差,且优于逐步回归分析法的结果。     

人工神经网络及其在分析化学中的应用

人工神经网络是一种新兴的计算方法，有着广阔的发展前途，目前在分析化学领域已经有了多方面的应用。本文简要介绍了人工神经网络的原理及其在分析化学中的应用。     

高温超导体制备条件的人工神经网络识别

高温超导体的制备条件要求严格,其掺杂量大小、烧结温度控制等都直接影响实验样本的超导性质。近几年高温超导研究已积累大量实验数据,如何应用人工智能方法处理这些数据寻找更好的合成条件颇有意义。人工神经网络方法在许多领域己显示其处理复杂数据和模式识别的优越性,但尚未应用到化合物制备研究方面。本文对此作了初步尝试。     

ANN原子参数法预报合金相晶型及晶格常数

运用人工神经网络方法，以二元合金组成元素的电子结构为基本特征，对二元合金相晶型作了区分，进一步对其晶格常数作了预报，结果令人满意。     

A hydrated amorphous iron oxide nanoparticle as active water oxidation catalyst

Developing efficient water oxidation catalysts (WOCs) with earth-abundant elements still remains a challenging task for artificial photosynthesis. Iron-based WOC is a promising candidate because it is economically cheap, little toxic and environmentally friendly. In this study, we found that the catalytic water oxidation activity on amorphous iron-based oxide/hydroxide (FeO_x) can be decreased by an order of magnitude after the dehydration process at room temperature. Thermogravimetric analysis, XRD and Raman results indicated that the dehydration process of FeO_x at room temperature causes the almost completely loss of water molecule with no bulk structural changes. Based on this finding, we prepared hydrated ultrasmall (ca. 2.2 nm) FeO_x nanoparticles of amorphous feature, which turns out to be extremely active as WOC with turnover frequency (TOF) up to 9.3 s^?1 in the photocatalytic Ru(bpy)₃²⁺-Na₂S₂O₈ system. Our findings suggest that future design of active iron-based oxides as WOCs requires the consideration of their hydration status.     

Repeatability and pattern recognition of bacterial fatty acid profiles generated by direct mass spectrometric analysis of in situ thermal hydrolysis/methylation of whole cells

Direct CI mass spectrometry profiling of fatty acid methyl esters (FAMEs) from in situ thermal hydrolysis/methylation (THM) of whole bacterial cells with tetramethylammonium hydroxide (TMAH) has been demonstrated as a potential method for real time and fieldable detection/identification of microorganisms. Bacillus anthracis (Ames), Yersinia pestis (Nair. Kenya), Vibrio cholerae (E1 Tor), Brucella melitensis (Abortus wild) and Francisella tularensis (LVS vaccine) were profiled by this method during a 10-month period. Repeatability of the in situ FAME data was calculated using one-way analysis of variance (ANOVA) and a t-test. Artificial neural network (ANN) and multivariate statistics of the FAME profiles were also compared for bacterial identification/classification. Equivalent results were obtained with a multivariate rule building expert system (MuRES) and the ANN. However, the ANN analysis required much less computer time and was deemed the best choice for this application. In situ THM FAME profiles of the bacterial samples provided comparable results with those obtained from the Microbial Identification System (MIDI) (Newark, DE) wet chemistry-gas chromatographic based system.     

An investigation into the construction of molecular models by the template joining method

Summary The results of a wide-ranging investigation into some of the different methods available for performing the joining of templates to build molecular models show that the choice of algorithm can significantly affect the quality of the results obtained, and different algorithms are most suited to particular categories of join.