基于XGBoost与拓扑结构信息的蛋白质复合物识别算法

蛋白质相互作用(PPI)网络中存在大量不确定性及已知蛋白质复合物数据的不完整性,单独地根据结构信息进行搜索或对已知复合物进行监督学习的方法在识别蛋白质复合物的准确性上存在不足。对此,提出一种XGBoost模型与复合物拓扑结构信息相结合的搜索方法(XGBP)。首先,根据复合物拓扑结构信息进行特征提取;然后,把所提取的特征用XGBoost模型进行训练;最后,将拓扑结构信息与监督学习方法相结合,建立特征与复合物之间的映射关系以提高蛋白质复合物预测的准确性。该算法分别与目前流行的马尔可夫聚类算法(MCL)、极大团聚类方法(CMC)、基于核心-附属结构算法(COACH)、快速层级聚类算法(HC-PIN)、基于重叠邻居的扩展聚类(ClusterONE)、分子复合物检测算法(MCODE)、基于不确定图模型的蛋白质复合物检测方法(DCU)和加权核心-附属算法(WCOACH)这八种非监督学习算法和三种监督学习方法贝叶斯网络(BN)、支持向量机(SVM)、回归模型(RM)进行比较,所提方法在精准度、敏感度、F-measure方面显示出良好的性能。     

尿嘧啶和它的分子识别膜之间相互作用机理的理论研究

运用密度泛函方法研究尿嘧啶分子识别膜的作用机理,提出了分子识别膜的4种可能的分子间相互作用方式,对基于分子间氢键相互作用构建的复合物结构进行了优化,求得了相应的结合能;同时,对这4种可能存在的复合物的振动模式实施了理论计算预测,并与实验数据相对比,最终推断出该分子识别膜内的主要作用方式。     

人工代谢算法在故障诊断中的应用

从生物体新陈代谢的生理机能出发,建立了基于酶催化的人工代谢算法模型．通过对代谢算子功能的 分析提出了模式定理的概念,并证明了算法的收敛性．利用酶与底物相匹配的原理,通过代谢操作找出酶与底物的 最优匹配,从而找出对应故障现象的原因．在寻优过程中,不断地通过凋亡算子将不可能成立的原因筛除,降低了 故障诊断的成本．实例分析表明,该算法能有效地进行故障定位,尽可能地缩小了不必要的搜索空间．     

间歇硫酸盐法蒸煮过程纸浆卡伯值的近红外光谱法在线测量的研究

研究发现，硫酸盐法蒸煮液在800-900nm近红光谱波段间的吸收主要来自其中的碱木素，而其中的D-葡萄糖、D-甘露糖、D-木糖、D-半乳糖和糖醛酸等碳水化合物的降解产物及二氯甲烷和苯醇抽出物在此波段基本上没有吸收。并在这段波段建立起的纸浆卡伯值近红外光谱测量模型。该模型显示蒸煮液的吸光度与纸浆卡伯值具有较好的线性关系（相关系数R^2为0．996），并有较高的预测精度。因此，建议采用近红外光谱法实现硫酸盐法蒸煮过程纸浆卡伯值的在线测量，并有望开发出蒸煮过程纸浆卡伯值近红外光谱法在线测量仪，真正实现蒸煮过程的纸浆卡伯值的闭环反馈控制。     

基于免疫原理的信号调理系统的故障自愈分析

在机电装备信号调理系统的故障自愈研究中,通过对信号调理系统的结构特点、参数特性及典型故障特征等进行分析,并建立健康因子群,对系统抗衰性能进行评估．在深入分析生物免疫机理的基础上,建立自愈系统与免疫系统映射关系,构建基于免疫原理的故障自愈系统结构模型,并给出模型中主要模块结构和实现算法．根据系统故障自愈模型建立的方法,建立信号调理系统的模型,分析系统的故障,并对此模型进行仿真实验,验证故障自愈系统模型的可行性和有效性．经仿真实验证明,基于免疫原理构建的信号调理故障自愈系统可以实现其主要故障的自愈．     

采用隐马尔科夫模型的蛋白质复合物识别研究

动态蛋白质网络的构建和复合物识别问题是生物信息学领域目前研究的热点.针对现有的算法在解决前述问题上的不足,提出了一种基于隐马尔科夫模型的蛋白质复合物识别算法(HMM-PC).首先基于蛋白质的基因共表达特性构建初始蛋白质网络,然后利用蛋白质的共享功能注释、共享结构域和连接强度等信息来对网络进行加权,得到动态蛋白质网络.在此基础上,考虑前一时刻蛋白质网络拓扑结构信息对当前时刻蛋白质网络拓扑结构信息的影响,采用隐马尔科夫模型描述蛋白质复合物与网络个体间的相互关系,进而将动态蛋白质网络中的复合物识别问题建模为隐马尔科夫模型中的最优状态序列发现问题,并采用维特比算法识别得到蛋白质复合物.最后通过理论分析证明了所提算法的复杂度较低.采用DIP数据集和MIPS数据集中的酵母蛋白质网络作为测试对象,大量的仿真实验结果也表明,HMM-PC算法的鲁棒性较强,在查全率、查准率、F-measure和效率等方面的性能都要优于现有的复合物识别算法.     

拓扑距离特征指数对多氯联苯的QSPR研究

根据分子拓扑理论,提出1种新的多氯联苯(PCBs)化合物分子中,顶点原子(基团)点价值δ_i的计算方法,据以构建分子的结构参数——拓扑距离特征指数(TDEI);同时考虑分子体积以及氯原子在苯环上的位置和数目对PCBs性质的影响,引入分子体积参数(V_R)和基团定位指数(D),然后,采用多元线性回归法,分别建立了PCBs的正辛醇-水分配系数(lgK_(OW))和水溶解度(-lgS_w)的定量结构-性质相关模型,其相关系数分别达到0.9972和0.9730,均优于前人的工作。利用模型预测另外7个PCBs分子的lgK_(OW)和-lgS_W,预测值与实验值的一致性令人满意。     

一种改进主动形状模型的研究

主动形状模型（Active Shape Model,ASM）是一种较为流行的用于特征定位的统计学形状模型,主要应用于灰度图像中物体的跟踪与定位。在传统主动形状模型基础上,研究了一种构建局部纹理模型的新方法。该模型充分利用特征点之间的联系,构建加权模型。在ORL人脸数据库、JAFFE人脸数据库中进行特征定位实验,并进行了ASM性能分析。结果表明,改进ASM方法提高了搜索速度与特征点定位的精度。     

生物信息学中蛋白质结构的B样条结构曲线计算模型研究

蛋白质结构建模是生物信息学中一个基础而重要的问题．根据蛋白质分子自身的链式结构特点,基于蛋白质分子主链(即骨架)上的Ca原子,运用B样条空间曲线理论,提出了一种蛋白质骨架结构生物信息学分析模型——B样条结构曲线(BSSC)模型．利用BSSC模型的连续性、可控性、可扩展性等特点,可对蛋白质结构相关的诸多生物信息学问题进行分析和研究．     

基于贝叶斯网络的分层网络故障诊断

本文以SNMP网络管理模型的管理信息库（MIB）为基础,在不同层次上构建了用于故障判别与定位的贝叶斯网络。对MIB变量采用自适应自回归（AAR）模型建模分析,构建与其相关协议之间的贝叶斯网络,推断协议功能是否发生异常。分析各个协议之间的功能依赖关系,构建协议间的贝叶斯网络,定位协议间的故障根源。考虑网络中故障传播构建了基于网络拓扑的贝叶斯网,定位故障根源节点。最后,对构建的模型进行了实验仿真,并分析了模型的优点和缺点。     

Computational study of human phosphomannose isomerase: Insights from homology modeling and molecular dynamics simulation of enzyme bound substrate

Phosphomannose isomerase is a zinc metalloenzyme that catalyzes the reversible isomerization of mannose-6-phosphate and fructose-6-phosphate, and the three-dimensional (3D) structure of human phosphomannose isomerase has not been reported. In order to understand the catalytic mechanism, the 3D structure of the protein is built by using homology modeling based on the known crystal structure of mannose-6-phosphate isomerase from (PDB code 1PMI). The model structure is further refined by energy minimization and molecular dynamics methods. The mannose-6-phosphate-enzyme complex is developed by molecular docking and the key residues involved in the ligand binding are determined, which will facilitate the understanding of the action mode of the ligands and guide further genetic studies. Our results suggest a hydride transfer mechanism of alpha-hydrogen between the C1 and C2 positions but do not support the cis-enediol mechanism. The detailed mechanism involves, on one side, Zn2+ mediating the movement of a proton between O1 and O2, and, on the other side, the hydrophobic environment formed in part by Tyr278 promoting transfer of a hydride ion.     

Hybrid crystals based on thiophene/phenylene co-oligomers

We have grown crystals in which two kinds of thiophene/phenylene co-oligomers (TPCOs) are hybridized. Biphenyl-capped thiophene (BP1T) and biphenyl-capped terthiophene (BP3T) were chosen from among the TPCOs. The hybrid crystals were grown in both the vapor phase and the liquid phase. These hybrid crystals showed the emission colors intermediate between the two components. Correspondingly maximum peak positions of the emissions from the hybrid crystals were located halfway between those from single-component crystals of BP1T and BP3T. We made hybrid thin films by co-deposition of the two TPCOs in vacuum. The thin films exhibited both emission colors and emission peak positions similar to those of the hybrid crystals. The X-ray diffraction measurements indicate that the crystallographic structure of the hybrid crystals resembles that of the BP1T crystal. Also we made field-effect transistors using the hybrid crystals and measured their hole mobilities. We briefly discuss the implications of the X-ray diffraction and electrical data.     

Chemical fragmentation in quantum mechanical methods

We give a survey on the application of the chemical fragmentation concept in computer modelling of extended covalent systems. It will be stressed that information on molecular topology, as well as location and composition of the reaction centre allows the construction of a reasonable initial guess for the wave function and thus facilitates the solution of the Schr?dinger equation. For systems, where the chemical changes are localised to a few atoms, while others play the role of essentially electrostatic perturbation, a partition into active site and environment is possible providing a background to hybrid quantum mechanical/molecular mechanical (QM/MM) methods. Full molecular orbital treatment of large covalent systems at the minimal basis, semiempirical level becomes possible in the frame of the fragment self-consistent field (FSCF) method which was developed in the past two decades in our laboratory. As an application, we discuss the hydride shift reaction step in xylose isomerase catalysis.     

木糖醇发酵过程的软测量研究

针对木糖醇发酵过程中木糖醇浓度不能在线测量和影响发酵过程控制的情况,使用软测量技术来估算木糖醇的浓度和底物浓度,使用动态BP网络作为软测量模型,并确定了10个隐含层节点的网络拓扑结构,使用LM算法训练网络。用未经训练的数据检验软测量模型,取得了满意的逼近效果。实现了木糖醇发酵过程木糖醇浓度和底物浓度的间接实时测量。     

Structure-based analysis of Bacilli and plasmid dihydrofolate reductase evolution

Dihydrofolate reductase (DHFR), a key enzyme in tetrahydrofolate-mediated biosynthetic pathways, has a structural motif known to be highly conserved over a wide range of organisms. Given its critical role in purine and amino acid synthesis, DHFR is a well established therapeutic target for treating a wide range of prokaryotic and eukaryotic infections as well as certain types of cancer. Here we present a structural-based computer analysis of bacterial (Bacilli) and plasmid DHFR evolution. We generated a structure-based sequence alignment using 7 wild-type DHFR x-ray crystal structures obtained from the RCSB Protein Data Bank and 350 chromosomal and plasmid homology models we generated from sequences obtained from the NCBI Protein Database. We used these alignments to compare active site and non-active site conservation in terms of amino acid residues, secondary structure and amino acid residue class. With respect to amino acid sequences and residue classes, active-site positions in both plasmid and chromosomal DHFR are significantly more conserved than non-active site positions. Secondary structure conservation was similar for active site and non-active site positions. Plasmid-encoded DHFR proteins have greater degree of sequence and residue class conservation, particularly in sequence positions associated with a network of concerted protein motions, than chromosomal-encoded DHFR proteins. These structure-based were used to build DHFR specific phylogenetic trees from which evidence for horizontal gene transfer was identified.     

基于结构统计的纹理分类

针对纹理图像在视觉上的结构表现，选用能说明任意二值纹理的基元序列，将基元所对应的区域进行平移或伸张以覆盖整个纹理图像，完成所有基元对应区域的象素统计，获得能说明纹理中各结构比例及粒度的统计曲线。用此方法重新分类了Brodatz的111个纹理图片。实验证明视觉上相似的纹理有相似的统计曲线(相同或接近的峰、谷或转折位置)。比较其他分类结果，该方法更具有视觉接近性。     

An automatic homology modeling method consisting of database searches and simulated annealing

We introduce a method of homology modeling consisting of database searches and simulated annealing. All processes involving searches for homologous proteins, alignment, the construction of Calpha atoms, construction of main-chain atoms, and the construction of side-chain atoms are performed automatically. In this method, main-chain conformations are generated from the weighted average of mainchain coordinates in reference proteins. The weight is defined by the local space homology representing the similarity of environmental residues at topologically equivalent positions in reference proteins. Side-chain conformations are generated for constructed main-chain atoms by database searches, and main-chain atoms are optimized for the fixed side-chain conformations. These two processes, i.e., the side-chain generation and main-chain optimization, are repeated several times. This type of construction provides a structure similar to the x-ray structure, in particular, for main-chain and side-chain atoms in the residues belonging to structurally conserved regions (SCRs). The accuracy of our method was evaluated for 14 proteins whose structures are known. The average root mean square deviation between models and x-ray structures was 2.29 A for all atoms, and the percentage of chi1 angles within 30 degrees was 72.6% for SCRs residues. Some models were in good agreement with their respective x-ray structures. Our method, which has the advantage of being automated, gives results similar to, or better than, published results for three widely used test proteins. Our software, FAMS, is available on the World Wide Web.     

Structures and spectroscopic properties of nonperipherally and peripherally substituted metal-free phthalocyanines: a substitution effect study based on density functional theory calculations

The molecular structures, molecular orbitals, atomic charges, electronic absorption spectra, and infrared (IR) and Raman spectra of a series of substituted metal-free phthalocyanine compounds with four (1, 3, 5, 7) or eight (2, 4, 6, 8) methoxyl (1, 2, 5, 6) or methylthio groups (3, 4, 7, 8) on the nonperipheral (1-4) or peripheral positions (5-8) of the phthalocyanine ring are studied by density functional theory (DFT) and time-dependent DFT (TD-DFT) calculations. The calculated structural parameters and simulated electronic absorption and IR spectra are compared with the X-ray crystallography structures and the experimentally observed electronic absorption and IR spectra of the similar molecules, and good agreement between the calculated and experimental results is found. The substitution of the methoxyl or methylthio groups at the nonperipheral positions of the phthalocyanine ring has obvious effects on the molecular structure and spectroscopic properties of the metal-free phthalocyanine. Nonperipheral substitution has a more significant influence than peripheral substitution. The substitution effect increases with an increase in the number of substituents. The methylthio group shows more significant influence than the methoxyl group, despite the stronger electron-donating property of the methoxyl group than the methylthio group. The octa-methylthio-substituted metal-free phthalocyanine compounds have nonplanar structures whose low-lying occupied molecular orbitals and electronic absorption spectra are significantly changed by the substituents. The present systematical study will be helpful for understanding the relationship between structures and properties in phthalocyanine compounds and designing phthalocyanines with typical properties.     

基于局部特征分析的人脸识别方法

在传统的弹性图匹配基础上，提出一种基于局部特征分析的人脸识别算法。该方法利用人脸的先验结构和人脸图像的灰度分布知识，首先粗略地找出人脸图像的特征点，然后利用人脸弹性图对特征点的位置进行调整。最后在各个特征点处计算Gabor变化的系数，人脸相应被表示为特征点处的Gabor系数集合，对提取的特征向量用几种不同的度量距离来进行分类，并给出测试结果。实验表明，该方法优于传统的Eigenface方法，特别适用于训练图像样本较少的情况。相对于传统的弹性图匹配方法，该方法由于人脸特征点预先被估算出，而不是在整个图像上搜索，所以大大减少了计算量。     

FUS: a system to simulate conformational changes in biological macromolecules

In order to study the dynamics of protein and nucleic acid conformations, a molecular folding-unfolding system (FUS written in Lisp) has been developed. Secondary structure features of protein and nucleic acids are graphically represented by cubes in a modified 'Blocks World' paradigm. Modeling of protein and nucleic acid unfolding (denaturation) and folding of their three-dimensional structure is possible by the use of high level 'block' operators which allow displacement of these structural features in space. Due to the flexible nature of this program, FUS is a useful tool for the rapid evaluation of user-defined rules governing conformational changes. The use of FUS to unfold three common proteins (prealbumin, flavodoxin and triose phosphate isomerase) and a tRNA is presented.