Search

Predicting the accurate prognosis of breast cancer from high throughput microarray data is often a challenging task. Although many statistical methods and machine learning techniques were applied to diagnose the prognosis outcome of breast cancer, they are suffered from the low prediction accuracy (usually lower than 70%). In this paper, we propose a better method (genetic algorithm-support vector machine, we called GASVM) to significant improve the prediction accuracy of breast cancer from gene expression profiles. To further improve the classification performance, we also apply GASVM model using combined clinical and microarray data. In this paper, we evaluate the performance of the GASVM model based on data provided by 97 breast cancer patients. Four kinds of gene selection methods are used: all genes (All), 70 correlation-selected genes (C70), 15 medical literature-selected genes (R15), and 50 T-test-selected genes (T50). With optimized parameter values identified from GASVM model, the average predictive accuracy of our model approaches 95% for T50 and 90% for C70 or R15 in all four kernel functions using integrated clinical and microarray data. Our model produces results more accurately than the average 70% predictive accuracy of other machine learning methods. The results indicate that the GASVM model has the potential to better assist physicians in the prognosis of breast cancer through the use of both clinical and microarray data.

In the past decade, DNA microarray technologies have had a great impact on cancer genome research; this technology has been viewed as a promising approach in predicting cancer classes and prognosis outcomes. In this paper, we proposed two systematic methods which can predict cancer classification. By applying the genetic algorithm gene selection (GAGS) method in order to find an optimal information gene subset, we avoid the over-fitting problem caused by attempting to apply a large number of genes to a small number of samples. By extracting significant samples (which we refer to as support vector samples because they are located only on support vectors), we allow the back propagation neural network (BPNN) to learn more tasks. We call this approach the multi-task support vector sample learning (MTSVSL) technique. We demonstrate experimentally that the GAGS and MTSVSL methods yield superior classification performance with application to leukemia and prostate cancer gene expression datasets. Our proposed GAGS and MTSVSL methods are novel approaches which are expedient and perform exceptionally well in cancer diagnosis and clinical outcome prediction.

By extracting significant samples (which we refer to as support vector samples as they are located only on support vectors), we can identify principal genes and then use these genes to classify cancers either by support vector machines (SVM) or back-propagation neural networking (BPNN). We call this approach the support vector sampling technique (SVST). No matter the number of genes selected, our SVST method shows a significant improvement of classification performance. Our SVST method has averages 2-3% better performance when applied to leukemia and 6-7% better performance when applied to prostate cancer.

Developments in manufacturing techniques have beenheavily dependent on the ability to characterize the process through measuring and analyzing the produced part. This is especially important for new manufacturing processes that are still in the early stages of development, such as MEMS fabrication. Metrology techniques (that is, tomography, interferometry) are used on these parts to gather 3-D point cloud data. The data are then analyzed and compared with the original design. Many engineered components can be analyzed by fitting primitives to the point cloud using nonlinear least-squares minimization. The minimizer requires a starting-point (initial guess) that does not lead to a local minimum (an incorrect solution). This paper will present numerical methods that use bounding box strategies for obtaining initial estimates for the iterative least-squares fitting of analytic primitives (plane, torus, and the specific quadrics of sphere, right circular cone, and right circular cylinder) to point cloud data.

Microarray-based gene expression profiling has been a promising approach in predicting cancer classification and prognosis outcomes over the past few years. In this paper, we have implemented a systematic method that can improve cancer classification. By extracting significant samples (which we refer to as support vector samples because they are located only on support vectors), we allow the back propagation neural networking (BPNN) to learn more tasks. The proposed method named as the multi-task support vector sample learning (MTSVSL) technique. We demonstrate experimentally that the genes selected by our MTSVSL method yield superior classification performance with application to leukemia and prostate cancer gene expression datasets. Our proposed MTSVSL method is a novel approach which is expedient and perform exceptionally well in cancer diagnosis and clinical outcome prediction. Our method has been successfully applied to cancer type-based classifications on microarray gene expression datasets, and furthermore, MTSVSL improves the accuracy of traditional BPNN technique.

High-throughput techniques, such as microarray experiments, have given biologists the opportunity to measure the expression levels of a huge amount of genes at the same time. How to utilize these huge amounts of data, however, has become a major challenge in the post-genomic research era. One approach utilizes a Bayesian network, a graphical model that has been applied toward inferring genetic regulatory networks from microarray experiments. However, a user-friendly system that can display and analyze various gene networks from microarray experimental datasets is now needed. In this paper, we developed a novel system for constructing and analyzing gene networks. Firstly, we developed five Bayesian network algorithms to construct gene networks of the yeast cell cycle from four different microarray datasets. Secondly, we implemented a user-friendly gene network analyzing system. GNAnalyzer is capable of generating gene networks of the yeast cell cycle from experimental microarray data but also analyzing the performance of gene networks for every algorithm. Thirdly, our system utilizes both the powerful processing abilities of MatLab and the dynamic interface of LabVIEW in a single platform.

Background: The diagnosis of cancer in most cases depends on a complex combination of clinical and histopathological data. Because of this complexity, there exists a significant amount of interest among clinical professionals and researchers regarding the efficient and accurate prediction of breast cancers. Results: In this paper, we develop a breast cancer prognosis predict system that can assist medical professionals in predicting breast cancer prognosis status based on the clinical data of patients. Our approaches include three steps. Firstly, we select genes based on statistics methodologies. Secondly, we develop three artificial neural network algorithms and four kernel functions of support vector machine for classifying breast cancers based on either clinical features or microarray gene expression data. The results are extremely good; both ANN and SVM have near perfect performance (99 – 100%) for either clinical or microarray data. Finally, we develop a user-friendly breast cancer prognosis predict (BCPP) system that generates prediction results using either support vector machine (SVM) or artificial neural network (ANN) techniques. Conclusions: Our approaches are effective in predicting the prognosis of a patient because of the very high accuracy of the results. The BCPP system developed in this study is a novel approach that can be used in the classification of breast cancer.

Background: Since the development of biotechnologies such as array-based hybridization, massive amounts of gene expression profiles are quickly accumulating. How to utilize these huge amounts of data has become a major challenge in the post-genomic research era. One approach utilizes a Bayesian network, a graphical model that has been applied toward inferring genetic regulatory networks from microarray experiments. However, a user-friendly system that can display and analyze various gene networks from microarray experimental datasets is now needed. Results: In this paper, we have developed a novel system to construct and analyze various gene networks from microarray datasets. Three aspects characterize the major contributions of this paper. (1) Five Bayesian network algorithm codes were developed and written to construct gene networks of the yeast cell cycle using the information from four different microarray datasets. (2) A gene network analyzing system, GNAnalyzer, consisting of several user-friendly interfaces was implemented. GNAnalyzer is capable of running Bayesian algorithms, constructing gene networks, and analyzing the performance of each network algorithm simultaneously. (3) The system utilizes both the powerful processing ability of MatLab and the dynamic interfaces of LabVIEW in a single platform. Conclusions: This is the first time of this kind of design to be applied in bioinformatics. The system is designed to be extendible. Our next goal is to apply this technique to other real biomedical applications, such as human cancer classification and prognostic prediction.

In this paper, we have described a novel method to approach the study of gene networks. Firstly, we have developed and written five Bayesian network algorithms to construct gene networks of the yeast cell cycle based on four different microarray datasets. Secondly, we have implemented a gene network generating system that is more user-friendly. GNGS is capable of generating gene networks of the yeast cell cycle from experimental microarray data and comparing the performance of gene networks using five different Bayesian network algorithms. Our system utilizes both the powerful processing abilities of MatLab and the dynamic interface of LabVIEW in a single platform. Thirdly, we have compared the performance of each algorithm through measures such as execution time, sensitivity, and specificity for all five algorithms based on four different datasets. In the near future, we intend to further improve performance by utilizing dynamic Bayesian network algorithms that more accurately reflect living cells' dynamic behavior. Our approach will then be used to explore the gene networks of human cells based on the microarray datasets of human cancers.