关键词： Radiotherapy  

摘要： Among cancer victims, lung cancer accounts for most fatalities inmen and women. Patients at advanced stages of lung cancer suffer from poor survival rate. Majority of these patients are not candidates for surgery and receive radiation therapy (radiotherapy) as their main course of treatment. Despite effectiveness of radiotherapy against many cancers, more than half of these patients are unfortunately expected to fail. Recent advances in biotechnology have allowed for an unprecedented ability to investigate the role of gene regulation in lung cancer development and progression. However, limited studies have provided insight into lung tumor response to radiotherapy. The inherent complexity and heterogeneity of biological response to radiation therapy may explain the inability of existing prediction models to achieve the necessary sensitivity and specificity for clinical practice's or trial's design. In this study, we briefly review the current knowledge of genetic and signaling pathways in modulating tumor response to radiotherapy in non-small cell lung cancer as a case study of data mining application in the challenging cancer treatment problem. We highlight the role that data mining approaches, particularly machine learning methods, can play to improve our understanding of complex systems such as tumor response to radiotherapy. This can potentially result in identification of new prognostic biomarkers or molecular targets to improve response to treatment leading to better personalization of patients' treatment planning by reducing the risk of complications or supporting therapy that is more intensive for those patients likely to benefit. (C) 2012 Wiley Periodicals, Inc.

关键词： Sequence motif analysis   Drosophila melanogaster   Gene expression   Gene regulation   Enhancer elements   Invertebrate genomics   Transcriptional control   Machine learning  

摘要： Transcriptional enhancers integrate the contributions of multiple classes of transcription factors (TFs) to orchestrate the myriad spatio-temporal gene expression programs that occur during development. A molecular understanding of enhancers with similar activities requires the identification of both their unique and their shared sequence features. To address this problem, we combined phylogenetic profiling with a DNA-based enhancer sequence classifier that analyzes the TF binding sites (TFBSs) governing the transcription of a co-expressed gene set. We first assembled a small number of enhancers that are active in Drosophila melanogaster muscle founder cells (FCs) and other mesodermal cell types. Using phylogenetic profiling, we increased the number of enhancers by incorporating orthologous but divergent sequences from other Drosophila species. Functional assays revealed that the diverged enhancer orthologs were active in largely similar patterns as their D. melanogaster counterparts, although there was extensive evolutionary shuffling of known TFBSs. We then built and trained a classifier using this enhancer set and identified additional related enhancers based on the presence or absence of known and putative TFBSs. Predicted FC enhancers were over-represented in proximity to known FC genes;and many of the TFBSs learned by the classifier were found to be critical for enhancer activity, including POU homeodomain, Myb, Ets, Forkhead, and T-box motifs. Empirical testing also revealed that the T-box TF encoded by org-1 is a previously uncharacterized regulator of muscle cell identity. Finally, we found extensive diversity in the composition of TFBSs within known FC enhancers, suggesting that motif combinatorics plays an essential role in the cellular specificity exhibited by such enhancers. In summary, machine learning combined with evolutionary sequence analysis is useful for recognizing novel TFBSs and for facilitating the identification of cognate TFs that coordinat

关键词： 机器学习   数据挖掘   WEKA  

摘要： 针对机器学习技术理论性强、内容抽象、实践难的特点,提出了以Weka软件为平台的机器学习实践方案,并详细阐述了实施过程。通过对结果分析,达到使用Weka实践机器学习技术解决实际问题的目的。

关键词： Transdermal enhancers   Discriminant analysis   Gaussian process regression   Percutaneous absorption  

摘要： Discriminant analysis (DA) has previously been shown to allow the proposal of simple guidelines for the classification of 73 chemical enhancers of percutaneous absorption. Pugh et al. employed DA to classify such enhancers into simple categories, based on the physicochemical properties of the enhancer molecules (Pugh et al., 2005). While this approach provided a reasonable accuracy of classification it was unable to provide a consistently reliable estimate of enhancement ratio (ER, defined as the amount of hydrocortisone transferred after 24 h, relative to control). Machine Learning methods, including Gaussian process (GP) regression, have recently been employed in the prediction of percutaneous absorption of exogenous chemicals (Moss et al., 2009;Lam et al., 2010;Sun et al., 2011). They have shown that they provide more accurate predictions of these phenomena. In this study several Machine Learning methods, including the K-nearest-neighbour (KNN) regression, single layer networks, radial basis function networks and the SVM classifier were applied to an enhancer dataset reported previously. The SMOTE sampling method was used to oversample chemical compounds with ER > 10 in each training set in order to improve estimation of GP and KNN. Results show that models using five physicochemical descriptors exhibit better performance than those with three features. The best classification result was obtained by using the SVM method without dealing with imbalanced data. Following over-sampling, GP gives the best result. It correctly assigned 8 of the 12 "good" (ER > 10) enhancers and 56 of the 59 "poor" enhancers (ER < 10). Overall success rates were similar. However, the pharmaceutical advantages of the Machine Learning methods are that they can provide more accurate classification of enhancer type with fewer false-positive results and that, unlike discriminant analysis, they are able to make predictions of enhancer ability. (C) 2011 Elsevier B.V. All rights reserved.

关键词： Streamflow   Forecast   Machine learning   Artificial neural network   Support vector regression   Gaussian process  

摘要： Weather forecast data generated by the NOAA Global Forecasting System (GFS) model, climate indices, and local meteo-hydrologic observations were used to forecast daily streamflows for a small watershed in British Columbia, Canada, at lead times of 1-7 days. Three machine learning methods - Bayesian neural network (BNN), support vector regression (SVR) and Gaussian process (GP) - were used and compared with multiple linear regression (MLR). The nonlinear models generally outperformed MLR, and BNN tended to slightly outperform the other nonlinear models. Among various combinations of predictors, local observations plus the GFS output were generally best at shorter lead times, while local observations plus climate indices were best at longer lead times. The climate indices selected include the sea surface temperature in the Nino 3.4 region, the Pacific-North American teleconnection (PNA), the Arctic Oscillation (AO) and the North Atlantic Oscillation (NAO). In the binary forecasts for extreme (high) streamflow events, the best predictors to use were the local observations plus GFS output. Interestingly, climate indices contribute to daily streamflow forecast scores during longer lead times of 5-7 days, but not to forecast scores for extreme streamflow events for all lead times studied (1-7 days). (C) 2011 Elsevier B.V. All rights reserved.

关键词： 贝时斯网络   粗糙集   马尔可夫模型   机器学习方法   RNA二级结构  

摘要： 随着人类基因组计划的成功实施，充分利用这些基因组测序数据，得到对人类有用信息的需求迅速增加，随之而诞生的生物信息学充分利用信息技术、数学、生物学、计算机科学等多学科工具，有效分析了海量数据所包含的生物学意义。而机器学习方法则给生物信息学提供了可靠有力的算法支撑，使得生物信息学对生命科学的探索与研究做出了重大贡献。本文利用了生物信息学中的若干种机器学习方法，研究了RNA的分子自折叠二级结构预测及RNA分子间靶位点碱基对匹配模式与靶点预测问题。\n 针对RNA二级结构预测算法中同源多序列保守结构预测法较单序列预测法复杂度高的问题，本文提出了保守二级结构中的保守碱基对来源于每个单序列的最优和次优自由能结构的假设。有效利用单序列自由能结构中各茎子结构的相关特征属性信息，通过支持向量机算法，保守碱基对被挑选出来。基于RNA二级结构预测标准数据集和Rfam数据库中十种RNA家族的二级结构预测实验显示，本算法的预测性能不低于主流预测算法水平。单序列自由能结构的引入既保证了较高的保守结构预测性能，又在一定程度上降低了算法复杂度。另外，本文研究的粗糙集方法挑选保守碱基对以及依靠茎子结构的公共程度属性排序挑选保守茎子结构法对SVM预测算法的最终形成提供了实验基础。\n 在表示各个单序列最优及次优自由能结构的诸个茎子结构与保守二级结构的关系时，本文给出了一种有效的RNA二级结构表示方法，简约点阵图法，使得基于茎子结构粒度的预测算法原理描述更加形象与深刻。\n RNA分子间靶点预测亦主要依据RNA的一级序列信息。对于miRNA靶位点碱基对互补配对模式特征的提取，本文先后使用了粗集工具与马尔可夫链式模型工具研究了配对模式的定性与定量特征。本文首先借助于两种量度变量，然后使用粗糙集工具，分析了miRNA的22个碱基中各个部分的匹配模式。接着，对miRNA靶点的序列匹配模式进行了进一步定量分析。通过建立马氏链模型，对训练数据进行学习，再根据极大似然法得到模型参数。已作完参数估计的模型为成熟的模型。基于这个成熟模型的得分，制定出一个合适的阈值，高于阈值的匹配模式都将被视为有效的模式，用于评估候选靶点作为真正靶点的可能性。\n 随着近几年来微阵列基因芯片的广泛使用，基因表达谱数据为RNA分子间靶点预测提供了一种有效的高通量实验模式。本文通过建立描述miRNA、mRNA以及蛋白质表达量之间相互关系的贝叶斯网络模型，预测出miRNA靶点的同时指出miRNA按哪种方式发挥作用。

关键词： 机器学习   增量式SVM   MHC结合  

摘要： 随着艾滋病和抗生素抵制的加剧,人们对疫苗研究越来越关注。通过表位短肽与主要组织相容性复合物(MHC)分子结合,激发免疫应答,由此设计和研发出具有应用价值的肽疫苗,准确预测肽段与MHC结合能力是关键技术之一。采用基于机器学习的支持向量机(SVM)算法对MHC结合肽进行预测分析,构造增量式SVM训练模型,然后将算法嵌入到Web平台搭建结合肽预测系统。

关键词： 蛋白质组学   医学成像   机器学习   上海同步辐射光源   医学图像数据  

摘要： 上海同步辐射光源(Shanghai Synchrotron Radiation Facility,SSRF)是一台高性能的第三代同步辐射光源,是我国迄今为止最大的科学研究装置和公共实验平台。利用上海同步辐射装置的高亮度、短波长的同步辐射光在空间分辨上的优势,将可以进行许多前沿学科的探索,尤其是生命科学领域。利用BL17U—生物大分子晶体学光束线站,生物学家能获得生物大分子的三维结构,进而研究其结构与功能之间的关系;而利用BLl3W1一X射线成像及生物医学应用光束线站,医学家则可以揭示活体肿瘤和脑血管病的发生和发展机制,为发展重大疾病的早期诊断与治疗提供关键理论基础和技术支撑。\n 目前利用上海光源已经产生了大量的蛋白质和医学图像数据,单纯地利用传统的生物和医学实验方法,或者根据专家经验将很难快速且全面的处理如此多的数据,从而必然制约了上海光源在生命科学领域应用的快速发展。在这种情况下,利用计算机自动的理解这些数据的生物信息学和医学图像处理应运而生。\n 本文中讨论利用机器学习的方法,对这些蛋白质结构和医学图像数据进行分类、检索以及相关特性的预测。旨在运用机器学习的理论和方法来研究蛋白质结构与功能特性,以及医学影像和疾病特点之间的内在联系,在已有数据的基础之上发现相应的规律和知识并进而用来进一步指导与解释新产生的蛋白质和医学影像,加速对生命本质特征的认识。本论文在机器学习及生物信息学、医学图像理解理论与方法上进行了深入的研究与探索。分别提出了新颖的医学图像检索,基于医学图像的组织分类,蛋白质数据库检索,以及蛋白质折叠类心预测算法。相关成果发表于顶级期刊Pattern Recognition,BMCBioinformatics和IIEEE Transactions on Medical Imaging等期刊上。论文的主要内容和创新点如下:\n 1.基于内容的医学图像检索指的是,给出一幅医学图像,系统根据图像的内容相似度,从数据库中检索出若干相似的医学图像。这对医学研究人员和医生判断医学图像所隐含的疾病信息很有帮助。基于“特征袋”这一图像表达模型,对医学图像的局部特征量化分配问题进行了创新研究,并应用于医学图像检索。建立了局部特征到量化的视觉词典的贡献函数,并基于二次规划方法,提出了一种新的分配侧率QP分类。进一步地,对视觉词的加权进行了研究,提出了为每个视觉词构建弱分类器,再通过AdaBoost算法学习每个视觉词权值的算法。\n 2.医学图像中的组织进行自动分类,对医疗诊断和研究有着重要的意义。我们提出了利用基于“特征袋”方法的组织分类方法。“特征袋”方法最重要的两个环节就是视觉词典的学习,以及视觉词的加权。虽然目前针对这两个问题有很多研究,但是都是相互独立的。而实际上他们存在着密切的联系。我们提出了视觉词典及其加权矢量的联合学习算法,在一个目标函数中统一这两个参数,并且在算法中迭代地优化。\n 3.蛋白质数据库检索,或者蛋白质的最近邻分类,指的是给出一个查询蛋白质,从数据库中找出若干性质(如折叠类型)相同的蛋白质。这以过程中关键的步骤就是如何定义“相似度”的问题。传统的相似测度只考虑了两个蛋白质,而最近提出的上下文敏感的相似测度则可以同时考虑数据库中的其他样本。我们通过定义“参考集合”来考察到底数据库中的哪些样本会起到影响的作用。进而提出了最短路径传播算法,指出两个蛋白质之间的最短路径上的蛋白质起到的作用最大。\n 4.目前为止,所有的上下文相关的相似度学习都未用到数据库中蛋白质的类别标号(如折叠类型的标号)。这阻碍了这一领域的发展。我们创造性的提出了利用蛋白质类别标号的监督式上下文敏感的相似度学习方法。利用数据库中的同类蛋白质对(相关的),以及异类蛋白质对(无关的),我们计算他们的上下文相似度矢量,来训练一个SVM模型来对一对未知的蛋白质对的相似度进行规整,得到新的相似度。而基于新的相似度,我们又重新确定他们的上下文。如此,我们提出了新的上下文敏感相似度计算方法。\n 5.最后,我们利用我们提出的机器学习的方法,对上海光源产生的医学图像和蛋白质结构进行了检索和分析的示例。对BLl3W1一X射线成像及生物医学应用光束线站城乡形成的肺癌图像,进行了数据库检索试验。对BL17U-生物大分子晶体学光束线站产生的蛋白质结构数据3EOP,进行了数据库检索和折叠类型预测的实验。实验结果证明我们提出的算法能有效应用于SSFR数据的分析。

关键词： 稀疏中心核   图矩阵   半监督分类   核低秩编码   子空间分割  

摘要： 随着信息技术的飞速发展，机器学习已经发展成为当前人工智能领域的研究热点之一。稀疏性理论已被成功应用于许多机器学习方法中。核学习和谱图理论易于理解且具有较强推广能力，是当今机器学习领域的热点问题。本文将稀疏性引入到核函数的构造和图矩阵的构造问题中，构造了几种稀疏学习算法，采用标准测试问题验证了其性能。研究内容主要包括以下几方面：\n (1)在现有的稀疏编码分类算法的基础上，提出一种稀疏中心核编码分类算法。该方法首先通过k近邻思想在字典中选择中心样本，然后利用所有中心样本构造稀疏核函数，将字典和测试样本通过稀疏核函数映射至特征空间，最后在特征空间内对映射后的测试样本进行稀疏编码，利用得到的稀疏系数和误差判别函数进行分类。由于核映射时使用的是稀疏核函数，既增强了数据的可分性又减少了计算核函数的次数，使得该算法既能达到较高的识别精度，又能减少识别消耗时间。在常用的标准人脸数据库和手写体数字数据库上进行的仿真实验验证了该算法的有效性。\n (2)提出了一种基于核l1图的半监督分类算法。在现有的l1图方法的基础上，通过引入核的思想，使得线性不可分的样本在特征空间内变得线性可分。将核l1图方法构造的图矩阵与半监督分类方法结合，构造了基于核l1图的半监督分类。由于获得的图矩阵增强了同类样本间的相似性以及不同类样本之间的差异性，因此在后续分类中可获得更高的分类正确率。在标准人脸数据库和手写体数字数据库上进行实验仿真，结果显示：与同类算法相比，分类正确率有明显提高。\n (3)提出了一种基于核低秩编码的子空间分割算法。在该算法中，先通过核函数将样本集合映射至特征空间，再对新的样本集合联合求解得到低秩编码系数矩阵，从而构造无向图进行谱聚类。由于核映射加强了数据的可分性，使得后续子空间分割争取率有所提高。在人工数据集、标准的人脸数据库和手写体数据的实验仿真表明：该算法与现有方法相比，分割正确率有所提高且对噪声具有鲁棒性。

关键词： Conditional Random Fields   Deep Belief Networks   Machine Learning   Ultrasound Imaging   Linguistics   Articulatory Speech Data   Automatic Speech Recognition  

摘要： Humans make use of more than just the audio signal to perceive speech. Behavioral and neurological research has shown that a person's knowledge of how speech is produced influences what is perceived. With methods for collecting articulatory data becoming more ubiquitous, methods for extracting useful information are needed to make this data useful to speech scientists, and for speech technology applications. This dissertation presents feature extraction methods for ultrasound images of the tongue and for data collected with an Electro-Magnetic Articulograph (EMA). The usefulness of these features is tested in several phoneme classification tasks. Feature extraction methods for ultrasound tongue images presented here consist of automatically tracing the tongue surface contour using a modified Deep Belief Network (DBN) (Hinton et al. 2006), and methods inspired by research in face recognition which use the entire image. The tongue tracing method consists of training a DBN as an autoencoder on concatenated images and traces, and then retraining the first two layers to accept only the image at runtime. This 'translational' DBN (tDBN) method is shown to produce traces comparable to those made by human experts. An iterative bootstrapping procedure is presented for using the tDBN to assist a human expert in labeling a new data set. Tongue contour traces are compared with the Eigentongues method of (Hueber et al. 2007), and a Gabor Jet representation in a 6-class phoneme classification task using Support Vector Classifiers (SVC), with Gabor Jets performing the best. These SVC methods are compared to a tDBN classifier, which extracts features from raw images and classifies them with accuracy only slightly lower than the Gabor Jet SVC method. For EMA data, supervised binary SVC feature detectors are trained for each feature in three versions of Distinctive Feature Theory (DFT): Preliminaries (Jakobson et al. 1954), The Sound Pattern of English (Chomsky and Halle 1968), and Un