Search

Clinical trajectory is a clinically relevant sequence of ordered patient phenotypes representing consecutive states of a developing disease and leading to some final state. Extracting trajectories from large scale medical data is of great interest for dynamical phenotyping of various diseases but remains a challenge for machine learning methods, especially in the case of synchronic (with short follow up) observations. Here we describe an approach for trajectory-based analysis of cancer data using elastic principal trees and test it on a large collection of molecular tumoral profiles for breast cancer. We show that the disease progress quantified with pseudotime (the geodesic distance from the root) along a particular trajectory can serve as a significant prognostic factor, not redundant with gene expression-based predictors. We conclude that application of the elastic principal trees to transcriptomic data can be of interest for clinical applications.

State of the art genomic methods produce an abundance of data which ultimately increases the quantity within and of data repositories. Thereby, open data is of much importance to boost scientific studies for combating disease. More and more often, there are several data sources available, offering diverse sample data. Particularly, interpretation of genomic data remains a challenge due to data size and differing quality. We present an approach for visualizing uncertainty of heterogeneous data sources on mutation rates in genomic data for pediatric cancer analysis. Visualization as method for knowledge discovery will be of great importance in order to put inhomogeneity of sample data into perspective.

We propose an effect size based approach to compute initial dissimilarities for Ensemble Algorithm of Clustering Cancer Data (EACCD). The proposed method is applied to the colon cancer data from the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute and compared with the log-rank approach where initial dissimilarities are computed from the log-rank test statistic. The experimental results show that under the proportional hazards assumption, the effect size approach generates robust results and has a better performance than the log-rank approach.

Gene Expression data comprising more number genes and samples are used for cancer classification. Gene Expression information provides the gene expression level contributing to specific action. Cancer classification is done using the gene expression data. In our work, feature selection methods such as mutual information and genetic algorithm are used for examining cancer microarray Gene expression Data. By using feature selection methods most probable cancer associated genes are selected from large microarray Gene Expression data. The objective of the present work to attain enhanced classification accuracy. We have used the colon cancer dataset contains 2000 gene has 62 types, in which 40 are colon cancer tissue's types and the other 20 are normal types.

Boolean implications (if-then rules) provide a conceptually simple, uniform and highly scalable way to find associations between pairs of random variables. In this paper, we describe how Boolean implications can be derived from large, heterogeneous cancer data sets. We demonstrate two applications of Boolean implications to discover new actionable insights in cancer biology.

Bioengineering, bioinformatics, and other advanced technologies provide crosscutting tools that facilitate research in many disciplines. This paper gives Analysis of Cancer data taken from Microarray profiles.

Cancer classification from microarray gene expression data is a challenging task in computational biology and bioinformatics as the sufficient number of labeled samples (required to train the traditional classifiers) are very expensive and difficult to collect. Therefore, the predication accuracies of the classifiers trained with limited training samples are often very low. Although, the unlabeled samples are relatively inexpensive and readily available, traditional classifiers not generally utilize the distribution of those unlabeled samples. In this context, this article presents a novel `self-training' based semi-supervised classification method using fuzzy K-Nearest Neighbour algorithm which utilizes the unlabeled samples along with the labeled samples to improve the prediction accuracy of the cancer classification. The proposed method is evaluated with a number of microarray gene expression cancer data sets. Experimental results justify the potentiality of the proposed semi-supervised method for cancer classification using microarray gene expression data in comparison to its other supervised counterparts.

DNA microarray technologies are leading to an explosion in available gene expression data which simultaneously monitor the expression pattern of thousands of genes. All the genes may not be biologically significant in diagnosing the disease. In this paper, a novel approach has been proposed to select significant genes of leukemia cancer using K-Means clustering algorithm. It is an unsupervised machine learning approach, which is being used to identify the unknown patterns from the huge amount of data. The proposed K-Means algorithm has been experimented to cluster the genes for K=5,10 and 15. The significant genes have been identified through the best accuracy obtained from the clusters generated. The accuracy of the clusters are determined again by using K-Means algorithm compared with ground truth values.

In this paper, we present an analysis of the ensemble algorithm of clustering of cancer data (EACCD) by Chen et al. Using a breast cancer dataset, we demonstrate the effectiveness of EACCD in capturing differences among survival curves. We also investigate the impact of different settings in EACCD and compare EACCD with several other clustering approaches.

Cancer is one of the most the gravest problems facing mankind. In 2008, it is estimated that over 7.6 million lives have been claimed by cancer. Early and precise detection plays a key role in treating the disease and improve survivability of patient. Among data classification algorithms, discrete particle swarm optimization (DPSO), a technique based on standard PSO has proved to be competitive in predicting breast cancer, and in this paper, we implement a classifier using DPSO with new rule pruning procedure for detecting lung cancer and breast cancer, which are the most common cancer for men and women. Experiment shows the new pruning method further improves the classification accuracy, and the new approach is effective in making cancer prediction.

In this paper, a novel method for improving flexible neural tree is proposed to classify the leukemia cancer data. The hybrid flexible neural tree with pre-defined instruction sets can be created and evolved. The structure and parameter of hybrid flexible neural tree are optimized using probabilistic incremental program evolution (PIPE) and particle swarm optimization (PSO) algorithm. The experimental results indicate that the proposed method illustrates feasible and efficient for the classifications of microarray data.

In order to perform successful diagnosis and treatment of cancer, discovering and classifying cancer types correctly is essential. One of the challenges in cancer class discovery is to estimate the number of classes given a set of unknown microarray data. In the paper, we propose a new cluster validity criterion called Penalty-based Disagreement Index (PDI) based on the perturbation technique to estimate the number of classes in microarray data, PDI not only considers the disagreement between the partition results obtained from the original data and those obtained from the perturbed data, but also includes a penalty measure which is a function of the number of classes. Our experiments show that PDI successfully estimates the true number of classes in a number of challenging real cancer datasets.

The cluster ensemble approach is gaining more and more attention in recent years due to its useful applications in bioinformatics and pattern recognition. In this paper, we present a new cluster ensemble approach named as the neural gas based cluster ensemble algorithm (NGCEA) for class discovery from biological meaningful data, NGCEA first adopts the perturbed function to generate a set of new datasets. Then, it proposes to adopt the neural gas algorithm to obtain the clustering solutions from the perturbed datasets, In the following, NGCEA views the row of each clustering solution as the new features, and forms a new dataset. Finally, it adopts the neural gas algorithm as consensus function to perform clustering again on the new dataset and obtains the final result. The experiments in cancer datasets show that (i)NGCEA works well on most of cancer datasets (ii) NGCEA outperforms most of the state-of-the-art cluster ensemble algorithms when applied to gene expression data

This paper describes an intelligent classification system for cancer data. The system employs a hybrid radial basis function (HRBF) network in order to classify cancer data into several classes. The HRBF network is trained using the moving k-means clustering algorithm to position the network's centre and the Given least square (GLS) algorithm to estimate the network's weights. Two cancer data, i.e. cervical cancer and breast cancer, are used as case studies. For cervical cancer, the system classifies the data into three classes, i.e. normal, low grade squamous intraepithclial lesion (LSIL) and high grade squamous intraepithclial lesion (HSIL). The system produces 98.00% accuracy. While for breast cancer, the system classifies the data into benign and malignant data. The system produces 98.57% accuracy. The result illustrates the promising capabilities of the system for assisting cervical and breast cancer detection.