计算机算法结合实验技术进行T细胞表位筛选

计算机算法与实验技术相结合的方法,已广泛用于各种免疫相关抗原T细胞表位的鉴定,对现在可获得的计算机预测方法,以及计算机算法与实验相结合鉴定T细胞表位的方法进行了综述,主要内容包括表位预测及抗原加工处理的预测。     

人ZP3B-细胞表位嵌合肽DNA的设计和合成

目的：设计和合成适合在昆虫细胞中表达的一个人ＺＰ３ Ｂ细胞表位的ＤＮＡ序列。方法：根据国外的研究基础，设计由人ＺＰ３的Ｂ细胞表位肽段和外源Ｔｈ细胞表位肽段串联而成的４５肽嵌合肽序列，并根据昆虫细胞对密码子的偏爱性设计出该嵌合肽的ＤＮＡ序列。然后人工合成该嵌合肽的ＤＮＡ链，并克隆到ＰＵＣ１８载体上。结果：通过酶切分析和测序证明，合成的ＤＮＡ与所设计的嵌合肽ＤＮＡ序列一致。结论：按设计合成了人ＺＰ３的     

基于并行的非支配排序遗传Ⅱ算法优化双聚类

双聚类是微阵列基因表达数据分析中很实用的一种数据挖掘技术,它是一种同时对微阵列基因和条件进行聚类的方法,用来挖掘基因子集在条件子集下所体现出来的生物模式。传统的双聚类算法对于庞大的基因表达数据处理效率很弱,考虑在j Metal平台上实现基因表达数据的双聚类的一种新的研究方法及思路。同时考虑加入并行策略,提高算法的效率。在酵母啤酒细胞基因表达集和人类B-细胞两个标准数据集上对两个算法进行实验验证,表明所提出算法比其他多目标双聚类算法呈现出更好的优越性。     

Antigenic epitope peptides of influenza H3N2 virus neuraminidase gene based on experiments

The neuraminidase (NA) in viral surface is one of the main subtype-specific antigen of influenza type A viruses.Neuraminidase is an enzyme to break the bonds between hemagglutinin (HA) and sialic acid to release newly formed viruses from infected cells.In this study,the H3N2 subtype virus NA genes were sequenced and NA proteins were screened for B-cell epitopes and assessed based on immunoinformatics.Based on this results,four peptides DR6,EY7,VG8 and RE8 (covering amino acid residues 151-156,368-374,398-405 and 428-435,respectively) of the NA protein were synthesized artificially.These peptides were used to immunize New Zealand rabbits subcutaneously to raise antisera.Experimental results showed that these four peptides were capable of eliciting antibodies against H3N2 viruses in a specific and sensitive feature,detected in vitro by enzyme-linked immunosorbent assay.Moreover,hemadsorption anti-releasing effects took place in three three-antisera-mixtures at a dilution of 1:40.Alignment using NA gene database showed that amino acid residues in these four epitope peptides were substituted at specific sites in all the NAs sequenced in this study.It was suggested that these NA epitope peptides might be used in combination with HA proteins as vaccine antigens.     

ECOC多分类算法在慕课数据挖掘中的应用

收集并整合多所高校学生的慕课学习行为数据,设计基于数据复杂度的纠错输出编码(ECOC)多分类算法.该算法利用数据复杂度降低多类之间的分类难度,从而提高算法的预测准确度.实验结果表明,在不同高校的慕课数据集的测试中,所设计基于数据复杂度的ECOC分类算法比传统的ECOC算法具有更高的分类准确度和鲁棒性,实现了学生学习成绩...     

展示白色念珠菌热休克蛋白表位的杂合噬菌体诱导小鼠产生的体液免疫应答反应研究

白色念珠菌HSP90 LKVIRK是一段保守的B细胞线性表位.利用展示有LKVIRK表位的杂合噬菌体作为抗原,免疫C57BL/6J小鼠,然后在小鼠的系统性念珠菌感染模型中评价其免疫保护作用.结果证明该杂合噬菌体能够诱导C57BL/6J小鼠产生高滴度的抗HSP90抗体,而且对系统性念珠菌感染的小鼠具有免疫保护作用.说明展示有白色念珠菌LKVIRK表位的杂合噬菌体有可能成为一种候选的真菌疫苗.     

HPV-58型E7蛋白抗原模拟表位的筛选及分析

利用该实验室成熟的噬菌体随机十二肽库筛选平台,对HPV-58型E7蛋白单克隆抗体株4C4a、6C7a进行3轮亲和筛选,得到2组、各8个阳性噬菌体克隆.阳性噬菌体氨基酸序列用以模拟单抗(靶蛋白)特异性识别的HPV-58型E7蛋白的抗原表位.将测得的阳性噬菌体多肽序列及HPV-58型E7蛋白氨基酸序列进行生物信息学分析后,进行线性序列比对及构象性表位预测及分析.构象性表位的预测借力于EpiSearch Server和Pepitope Server,对构象性表位匹配及簇的寻找提供有力支持,也为早期诊断及研制多肽疫苗提供参考.     

Performance Evaluation of an Anomaly-Detection Algorithm for Keystroke-Typing Based Insider Detection

Keystroke dynamics is the process to identify or authenticate individuals based on their typing rhythm behaviors. Several classifications have been proposed to verify a user's legitimacy, and the performances of these classifications should be confirmed to identify the most promising research direction. However, classification research contains several experiments with different conditions such as datasets and methodologies. This study aims to benchmark the algorithms to the same dataset and features to equally measure all performances. Using a dataset that contains the typing rhythm of 51 subjects, we implement and evaluate 15 classifiers measured by F1-measure, which is the harmonic mean of a false-negative identification rate and false-positive identification rate.We also develop a methodology to process the typing data. By considering a case in which the model will reject the outsider, we tested the algorithms on an open set. Additionally, we tested different parameters in random forest and k nearest neighbors classifications to achieve better results and explore the cause of their high performance. We also tested the dataset on one-class classification and explained the results of the experiment. The top-performing classifier achieves an F1-measure rate of 92% while using the normalized typing data of 50 subjects to train and the remaining data to test. The results, along with the normalization methodology, constitute a benchmark for comparing the classifiers and measuring the performance of keystroke dynamics for insider detection.     

基于改进XGBoost的地震多属性地质构造识别方法

地震属性可以用来解释与预测地质构造,因此地震属性被广泛地运用在煤矿地质构造的识别。但一般情况下,勘探区中无构造区域与有构造区域分布不均衡,无构造区域远远多于有构造区域。机器学习中,传统的分类器更习惯于偏向多数类,这使得如何有效地识别出构造体成为一个难题。为了解决这一问题,提出了一种针对不平衡数据集的改进极限梯度提升(extreme gradient boosting, XGBoost)构造识别方法。该方法的原理是,首先,以基于三维地震勘探成果数据体提取的12种地震属性为数据集特征,以实际揭露后的地质构造为数据集标签构建多属性数据集,然后以特征对标签的相关性为标准,过滤掉冗余的特征;其次,将边界样本分类算法(boundary sample classification, BSC)与合成少数类过采样技术(synthetic minority over-sampling technique, SMOTE)相结合形成BSC-SMOTE算法。用BSC-SMOTE算法对原始数据集进行平衡,再利用平衡后的数据集训练XGBoost分类器,并用贝叶斯优化(Bayesian optimization, B...     

局部二值平均熵模式与深度残差网络的人群密度估计

针对人群密度估计算法中场景的人群遮挡、尺度光照变化、噪声和低分辨率等问题,提出了一种结合局部二值熵值纹理特征(ELBP)与深度残差网络的人群密度估计算法。该算法首先在原始RGB人群图像上提取LBP特征;然后通过计算邻域像素点的平均信息熵模式构建ELBP纹理特征;随后基于ELBP纹理特征构建了一个深度为18层的深度残差网络;最后形成了对人群密度估计的end-to-end模式。为验证算法的可行性和有效性,在开源的人群密度估计数据集上进行实验。首先邀请10位专家对开源的数据集进行有效的人群聚集标注作为真实输出标签;随后采用研究提出的算法对人群密度完成估计,并与真实结果进行比较。另外,在三种不同的特征和三种不同的机器学习模型上进行了横向比较。实验结果表明,提出的ELBP纹理特征能够很好地应对噪声和低分辨率问题;深度残差网络则能够解决人群遮挡、尺度光照变化的问题。与传统算法相比,提出的算法能够提升人群密度估计的性能。     

社群网络中基于链接分类的研究与应用

由于在互联网、电信、生物信息、社会网络分析等领域可获得的链接丰富的数据日益增多,链接挖掘已经成为数据挖掘的研究热点。基于链接分类是链接挖掘的一个重要方向。在此以电信领域用户通话特征数据为对象,研究了基于链接分类技术及其在电信客户流失预测的应用方法,在提取并分析大规模客户呼叫图的极大团、结点膨胀率、结点聚集度等与节点稳定性相关的链接属性及其时变特征的基础上,提出了一种适合海量数据的基于链接的电信客户流失预测算法。实验结果表明,此算法较传统分类算法能提高客户流失预测性能,实现了基于链接分类方法的成功应用。     

近红外光谱的水体污染指标COD定量预测模型

针对传统化学需氧量（chemical oxygen demand, COD）检测存在检测成本高、耗时、易造成二次污染,以及现有检测模型泛化性较差等不足,难以满足水环境实时监测需求的问题,本研究提出基于近红外光谱技术的COD快速无损定量预测模型。实验结果表明,本模型在污水COD光谱数据集上的预测性能,相较于传统机器学习算法,以及现有其他深度学习算法更优。测试的决定系数（R2）和均方根误差（RMSE）分别达到0.992 1和27.47 mg·L-1,模型卷积层的输出特征可解释强,能有效表征关键波长点。为实际水体COD含量快速检测提供一种新的方法。     

面向同源蛋白质探测的一种新型混合深度学习模型

根据蛋白质氨基酸链探测其同源蛋白质,进而预测蛋白质的功能,是生物信息学研究领域的一个重要挑战,也是众多生物医学研究领域的基础研究内容,有着重要的科研价值和广泛的应用需求。其研究难点在于:(1)如何学习对同源蛋白质预测有效、有用的蛋白质特征信息;(2)如何更好地运用蛋白质特征信息,实现同源蛋白质的探测与识别。为了解决同源蛋白质探测与识别研究中的关键难点,本文提出一种基于混合深度学习架构的同源蛋白质探测与识别模型(HDLM-PHP)。通过采用统一的"管道式"深度学习架构,将蛋白质特征学习和探测识别统一为一个整体,提高同源蛋白质探测与识别的效能。采用多组并行的深度卷积神经网络,学习蛋白质的各种属性信息,以期获得丰富的待检测蛋白质和靶蛋白质的高级相关性特征,并通过全连接方式使用多层RBM结构融合和精炼这些相关性特征为全局相关性特征。通过统一的深度网络连接方式,以探测和识别任务为导向,学习到对于同源蛋白质预测最有效、最全面的蛋白质特征信息。在标准数据集SCOPe上,对所提模型进行性能与效率评测,结果表明:本文提出的模型能有效地学习到符合任务导向的蛋白质特征数据,提升同源蛋白质探测与识别的准确度和召回率,优于现有的模型和算法。     

Repeated epitope in the recombinant epitope-peptide could enhance ELDKWA-epitope-specific antibody response

Based on the hypothesis suggested by us that epitope-vaccine may be a new strategy against HIV mutation, we have studied several neutralizing epitopes on HIV envelope proteins. However we do not know whether a repeated epitope in a recombinant epitope-peptide can enhance epitope-specific antibody response or not. ELDKWA-epitope (aa669-674) on the C-domain of HIV-1 gp41 is a neutralizing epitope defined by the monoclonal antibody (mAb) 2F5 with broad neutralizing activity. In this study, we designed and prepared a series of the recombinant epitope-peptides bearing 1, 4 and 8 copies of ELDKWA-epitope respectively. In the comparison of the antisera induced by the three recombinant antigens, an obviously increased titre of ELDKWA-epitope-specific antibody was observed in the case of four and eight repeated epitopes. In flow cytometry analysis, the epitope-specific antibodies in both antisera showed stronger activity to bind the transfected CHO-WT cells that stably express HIV-1 envelope glycoprotein on the cell surfaces. These experimental results indicated that repeated epitope in the recombinant epitope-peptide could enhance ELDKWA-epitope-specific antibody response, which could contribute to designing an effective recombinant epitope-vaccine.     

鸡新城疫病毒HN蛋白多表位抗原的表达与免疫原性研究

有效抗原及表位的预测和筛选是疫苗研究的基础,在对鸡新城疫病毒HN蛋白抗原表位预测的基础上,对多表位抗原进行表达与免疫原性测定。根据生物信息学表位预测方法获得的家禽新城疫病毒抗原表位,利用PCR技术合成基因,构建pBVIL1-HN重组载体,转化大肠杆菌HB101,进行基因工程表达;经纯化蛋白后免疫小鼠,抗体滴度用酶联免疫吸附方法测定,确定抗原的免疫原活性。结果表明,多表位抗原基因经测序结果正确,融合基因在大肠杆菌得到高效表达,电泳纯融合多表位抗原经三次免疫得到抗血清,抗体滴度为1：8000。鸡新城疫病毒HN蛋白多表位抗原得到高效表达,且具有良好的免疫原性。     

Stepwise prediction and statistical screening: B-cell epitopes on neuraminidase of human avian H5N1 virus

The B-cell epitopes of virus are associated with the antiviral drug and the vaccine screening. As the nucleotide sequences of neuraminidase （NA） of stain GD-01-06 were sequenced, we predicted the a-helix and β-fold structure and the indexes of the flexible regions of secondary structure of NA with methods of the Hydrophilicity plot by Kyte-Doolittle, the Surface probability plot by Emini and the An- tigenic index by Jameson-Wolf, and then screened statistically the parameters to predict B-cell epi- topes by the Hierarchical cluster and the Bivariate correlation and the quartiles with SPSS 13.0. The impact of variation of amino acids in NA on its epitopes was analyzed. The predictive results were evaluated by Wu＇s Antigenic Index and SWISS-MODEL. We found that the most possible epitopes on NA were located within or nearby its N-terminal Nos. 120--137, 81--84, 408--415, 273--282, 429--432, 356--368, 46--55, 146--155, 341 --350 and 198--209, which were the dominant regions of NA epitopes. Peptide 120--137 including the glycoprotein domain （NGT126 128） was first chosen as the B-cell epitopes on NA. NA in H5N1 strain isolated after 2003 lacked in No. 53 amino acid （I）, resulting in an increase in the surface flexible region of NA in GD-01-06 and an enlargement to their epitope regions （VEP48-48→ VEPISNTNFL46-55）. Conclusively, prediction of the B-cell epitopes on the NA based on multiple pa- rameters is useful for researches on the molecular immunology and drug screening and immuno-prophylaxis. A deletion of No. 53 amino acid （I） in NA in strain GD-01-06 might increase its anti- genicity.     

基于频率比-随机森林模型的滑坡易发性评价

本文以陕西省洋县为研究区,通过搜集资料、实地调查获得研究区滑坡分布状况。结合研究区地质环境特征与前人研究经验,初步选取海拔、坡度、坡向、地形起伏度、曲率、距水系距离、距道路距离、降雨量及岩土体类型,共九种滑坡影响因子展开滑坡易发性研究。首先,采用皮尔森相关系数法对各因子间的相关性进行分析。其次,按照70/30的比例将滑坡数据随机划分为模型训练集与模型验证集。然后,采用模型训练集对频率比模型(FR)、随机森林模型(RF)及两者的耦合模型(FR-RF)进行训练,利用模型验证集对模型训练结果进行检验,并绘制ROC曲线。最后,利用验证后的模型绘制研究区滑坡易发性分区图。结果表明：(1)所选取的9个滑坡影响因子是相互独立的;(2)本研究所采用的三个模型均表现良好,其中FR-RF模型预测准确度最高(0.901),其次为RF模型(0.863),最后为FR(0.833);(3)本研究所绘制的滑坡易发性分区图可为当地政府制定土地利用规划、预防滑坡等方案提供参考借鉴。     

Chronic lymphocytic leukaemia is driven by antigen-independent cell-autonomous signalling

B-cell antigen receptor (BCR) expression is an important feature of chronic lymphocytic leukaemia (CLL), one of the most prevalent B-cell neoplasias in Western countries. The presence of stereotyped and quasi-identical BCRs in different CLL patients suggests that recognition of specific antigens might drive CLL pathogenesis. Here we show that, in contrast to other B-cell neoplasias, CLL-derived BCRs induce antigen-independent cell-autonomous signalling, which is dependent on the heavy-chain complementarity-determining region (HCDR3) and an internal epitope of the BCR. Indeed, transferring the HCDR3 of a CLL-derived BCR provides autonomous signalling capacity to a non-autonomously active BCR, whereas mutations in the internal epitope abolish this capacity. Because BCR expression was required for the binding of secreted CLL-derived BCRs to target cells, and mutations in the internal epitope reduced this binding, our results indicate a new model for CLL pathogenesis, with cell-autonomous antigen-independent signalling as a crucial pathogenic mechanism.     

博文盆地M煤层气田含气量主控因素分析及预测

精准预测煤层含气量是煤层气资源量评价及开发方案制定的关键之一，传统预测方法应用范围有限且预测误差较大。为了进一步提高含气量预测精度，以博文盆地M煤层气田GM煤层为研究对象，基于皮尔逊相关分析法对含气量主控因素进行定量分析，优选出了9个煤层含气量影响主控因素，提出基于高斯过程回归、支持向量机及集成装袋树等机器学习算法的煤层含气量预测模型，对比分析各方法优缺点，进而优选出最佳的含气量预测模型。研究结果表明：高斯过程回归精确度最高，预测结果平均相对误差为4.5%，适用于本气田煤层气的含气量预测；支持向量机泛化能力强，但是超平面的绝对化导致预测精度降低；集成装袋树并不侧重于训练数据集中的任何特定实例，对于噪声数据其基本不受过分拟合的影响。     

数据仓库的子立方体及其相应算法

数据仓库是研究大数据集的快速、灵活查询的一种技术,系统的执行效率是致关重要的.本文提出了一种新的概念--OLAP子立方体,并且给出应用这一概念改进数据仓库查询性能的相应算法.作者结合铁道部货运数据仓库及数据挖掘系统,深入的研究了这一技术,实验证明这种方法是有效的.