A Systematic Evaluation of Feature Selection and Classification Algorithms Using Simulated and Real miRNA Sequencing Data.
Ontology highlight
ABSTRACT:
SUBMITTER: Yang S
PROVIDER: S-EPMC4609795 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:BACKGROUND:With modern methods in biotechnology, the search for biomarkers has advanced to a challenging statistical task exploring high dimensional data sets. Feature selection is a widely researched preprocessing step to handle huge numbers of biomarker candidates and has special importance for the analysis of biomedical data. Such data sets often include many input features not related to the diagnostic or therapeutic target variable. A less researched, but also relevant aspect for medical applications are costs of different biomarker candidates. These costs are often financial costs, but can also refer to other aspects, for example the decision between a painful biopsy marker and a simple urine test. In this paper, we propose extensions to two feature selection methods to control the total amount of such costs: greedy forward selection and genetic algorithms. In comprehensive simulation studies of binary classification tasks, we compare the predictive performance, the run-time and the detection rate of relevant features for the new proposed methods and five baseline alternatives to handle budget constraints. RESULTS:In simulations with a predefined budget constraint, our proposed methods outperform the baseline alternatives, with just minor differences between them. Only in the scenario without an actual budget constraint, our adapted greedy forward selection approach showed a clear drop in performance compared to the other methods. However, introducing a hyperparameter to adapt the benefit-cost trade-off in this method could overcome this weakness. CONCLUSIONS:In feature cost scenarios, where a total budget has to be met, common feature selection algorithms are often not suitable to identify well performing subsets for a modelling task. Adaptations of these algorithms such as the ones proposed in this paper can help to tackle this problem.
| S-EPMC6986087 | biostudies-literature
Project description:Feature selection and classification are the main topics in microarray data analysis. Although many feature selection methods have been proposed and developed in this field, SVM-RFE (Support Vector Machine based on Recursive Feature Elimination) is proved as one of the best feature selection methods, which ranks the features (genes) by training support vector machine classification model and selects key genes combining with recursive feature elimination strategy. The principal drawback of SVM-RFE is the huge time consumption. To overcome this limitation, we introduce a more efficient implementation of linear support vector machines and improve the recursive feature elimination strategy and then combine them together to select informative genes. Besides, we propose a simple resampling method to preprocess the datasets, which makes the information distribution of different kinds of samples balanced and the classification results more credible. Moreover, the applicability of four common classifiers is also studied in this paper. Extensive experiments are conducted on six most frequently used microarray datasets in this field, and the results show that the proposed methods have not only reduced the time consumption greatly but also obtained comparable classification performance.
| S-EPMC6101392 | biostudies-literature
Project description:BACKGROUND:MicroRNAs (miRNAs) are noncoding RNA molecules heavily involved in human tumors, in which few of them circulating the human body. Finding a tumor-associated signature of miRNA, that is, the minimum miRNA entities to be measured for discriminating both different types of cancer and normal tissues, is of utmost importance. Feature selection techniques applied in machine learning can help however they often provide naive or biased results. RESULTS:An ensemble feature selection strategy for miRNA signatures is proposed. miRNAs are chosen based on consensus on feature relevance from high-accuracy classifiers of different typologies. This methodology aims to identify signatures that are considerably more robust and reliable when used in clinically relevant prediction tasks. Using the proposed method, a 100-miRNA signature is identified in a dataset of 8023 samples, extracted from TCGA. When running eight-state-of-the-art classifiers along with the 100-miRNA signature against the original 1046 features, it could be detected that global accuracy differs only by 1.4%. Importantly, this 100-miRNA signature is sufficient to distinguish between tumor and normal tissues. The approach is then compared against other feature selection methods, such as UFS, RFE, EN, LASSO, Genetic Algorithms, and EFS-CLA. The proposed approach provides better accuracy when tested on a 10-fold cross-validation with different classifiers and it is applied to several GEO datasets across different platforms with some classifiers showing more than 90% classification accuracy, which proves its cross-platform applicability. CONCLUSIONS:The 100-miRNA signature is sufficiently stable to provide almost the same classification accuracy as the complete TCGA dataset, and it is further validated on several GEO datasets, across different types of cancer and platforms. Furthermore, a bibliographic analysis confirms that 77 out of the 100 miRNAs in the signature appear in lists of circulating miRNAs used in cancer studies, in stem-loop or mature-sequence form. The remaining 23 miRNAs offer potentially promising avenues for future research.
| S-EPMC6751684 | biostudies-literature
Project description:Despite great increase of the amount of data from genome-wide association studies (GWAS) and whole-genome sequencing (WGS), the genetic background of a partially heritable Alzheimer's disease (AD) is not fully understood yet. Machine learning methods are expected to help researchers in the analysis of the large number of SNPs possibly associated with the disease onset. To date, a number of such approaches were applied to genotype-based classification of AD patients and healthy controls using GWAS data and reported accuracy of 0.65-0.975. However, since the estimated influence of genotype on sporadic AD occurrence is lower than that, these very high classification accuracies may potentially be a result of overfitting. We have explored the possibilities of applying feature selection and classification using random forests to WGS and GWAS data from two datasets. Our results suggest that this approach is prone to overfitting if feature selection is performed before division of data into the training and testing set. Therefore, we recommend avoiding selection of features used to build the model based on data included in the testing set. We suggest that for currently available dataset sizes the expected classifier performance is between 0.55 and 0.7 (AUC) and higher accuracies reported in literature are likely a result of overfitting.
| S-EPMC8315124 | biostudies-literature
Project description:This paper focuses on the application of machine learning algorithms for predicting spinal abnormalities. As a data preprocessing step, univariate feature selection as a filter based feature selection, and principal component analysis (PCA) as a feature extraction algorithm are considered. A number of machine learning approaches namely support vector machine (SVM), logistic regression (LR), bagging ensemble methods are considered for the diagnosis of spinal abnormality. The SVM, LR, bagging SVM and bagging LR models are applied on a dataset of 310 samples publicly available in Kaggle repository. The performance of classification of abnormal and normal spinal patients is evaluated in terms of a number of factors including training and testing accuracy, recall, and miss rate. The classifier models are also evaluated by optimizing the kernel parameters, and by using the results of receiver operating characteristic (ROC) and precision-recall curves. Results indicate that when 78% data are used for training, the observed training accuracies for SVM, LR, bagging SVM and bagging LR are 86.30%, 85.47%, 86.72% and 85.06%, respectively. On the other hand, the accuracies for the test dataset for SVM, LR, bagging SVM and bagging LR are the same being 86.96%. However, bagging SVM is the most attractive as it has a higher recall value and a lower miss rate compared to others. Hence, bagging SVM is suitable for the classification of spinal patients when applied on the most five important features of spinal samples.
| S-EPMC7004343 | biostudies-literature
Project description:High dimensionality and small sample sizes, and their inherent risk of overfitting, pose great challenges for constructing efficient classifiers in microarray data classification. Therefore a feature selection technique should be conducted prior to data classification to enhance prediction performance. In general, filter methods can be considered as principal or auxiliary selection mechanism because of their simplicity, scalability, and low computational complexity. However, a series of trivial examples show that filter methods result in less accurate performance because they ignore the dependencies of features. Although few publications have devoted their attention to reveal the relationship of features by multivariate-based methods, these methods describe relationships among features only by linear methods. While simple linear combination relationship restrict the improvement in performance. In this paper, we used kernel method to discover inherent nonlinear correlations among features as well as between feature and target. Moreover, the number of orthogonal components was determined by kernel Fishers linear discriminant analysis (FLDA) in a self-adaptive manner rather than by manual parameter settings. In order to reveal the effectiveness of our method we performed several experiments and compared the results between our method and other competitive multivariate-based features selectors. In our comparison, we used two classifiers (support vector machine, [Formula: see text]-nearest neighbor) on two group datasets, namely two-class and multi-class datasets. Experimental results demonstrate that the performance of our method is better than others, especially on three hard-classify datasets, namely Wang's Breast Cancer, Gordon's Lung Adenocarcinoma and Pomeroy's Medulloblastoma.
| S-EPMC4105478 | biostudies-literature
Project description:MotivationEpistasis or gene-gene interaction has gained increasing attention in studies of complex diseases. Its presence as an ubiquitous component of genetic architecture of common human diseases has been contemplated. However, the detection of gene-gene interaction is difficult due to combinatorial explosion.ResultsWe present a novel feature selection method incorporating variable interaction. Three gene expression datasets are analyzed to illustrate our method, although it can also be applied to other types of high-dimensional data. The quality of variables selected is evaluated in two ways: first by classification error rates, then by functional relevance assessed using biological knowledge. We show that the classification error rates can be significantly reduced by considering interactions. Secondly, a sizable portion of genes identified by our method for breast cancer metastasis overlaps with those reported in gene-to-system breast cancer (G2SBC) database as disease associated and some of them have interesting biological implication. In summary, interaction-based methods may lead to substantial gain in biological insights as well as more accurate prediction.
| S-EPMC3577111 | biostudies-literature
Project description:The new technology of single-cell RNA sequencing (scRNA-seq) can yield valuable insights into gene expression and give critical information about the cellular compositions of complex tissues. In recent years, vast numbers of scRNA-seq datasets have been generated and made publicly available, and this has enabled researchers to train supervised machine learning models for predicting or classifying various cell-level phenotypes. This has led to the development of many new methods for analyzing scRNA-seq data. Despite the popularity of such applications, there has as yet been no systematic investigation of the performance of these supervised algorithms using predictors from various sizes of scRNA-seq datasets. In this study, 13 popular supervised machine learning algorithms for cell phenotype classification were evaluated using published real and simulated datasets with diverse cell sizes. This benchmark comprises two parts. In the first, real datasets were used to assess the computing speed and cell phenotype classification performance of popular supervised algorithms. The classification performances were evaluated using the area under the receiver operating characteristic curve, F1-score, Precision, Recall, and false-positive rate. In the second part, we evaluated gene-selection performance using published simulated datasets with a known list of real genes. The results showed that ElasticNet with interactions performed the best for small and medium-sized datasets. The NaiveBayes classifier was found to be another appropriate method for medium-sized datasets. With large datasets, the performance of the XGBoost algorithm was found to be excellent. Ensemble algorithms were not found to be significantly superior to individual machine learning methods. Including interactions in the ElasticNet algorithm caused a significant performance improvement for small datasets. The linear discriminant analysis algorithm was found to be the best choice when speed is critical; it is the fastest method, it can scale to handle large sample sizes, and its performance is not much worse than the top performers.
| S-EPMC8905542 | biostudies-literature
Project description:The rapid emergence of novel psychoactive substances (NPS) poses new challenges and requirements for forensic testing/analysis techniques. This paper aims to explore the application of unsupervised clustering of NPS compounds' infrared spectra. Two statistical measures, Pearson and Spearman, were used to quantify the spectral similarity and to generate similarity matrices for hierarchical clustering. The correspondence of spectral similarity clustering trees to the commonly used structural/pharmacological categorization was evaluated and compared to the clustering generated using 2D/3D molecular fingerprints. Hybrid model feature selections were applied using different filter-based feature ranking algorithms developed for unsupervised clustering tasks. Since Spearman tends to overestimate the spectral similarity based on the overall pattern of the full spectrum, the clustering result shows the highest degree of improvement from having the nondiscriminative features removed. The loading plots of the first two principal components of the optimal feature subsets confirmed that the most important vibrational bands contributing to the clustering of NPS compounds were selected using non-negative discriminative feature selection (NDFS) algorithms.
| S-EPMC8638022 | biostudies-literature
Project description:Classification models built on biological expression data are increasingly used to predict distinct disease subtypes. Selected features that separate sample groups can be the candidates of biomarkers, helping us to discover biological functions/pathways. However, three challenges are associated with building a robust classification and feature selection model: 1) the number of significant biomarkers is much smaller than that of measured features for which the search will be exhaustive; 2) current biological expression data are big in both sample size and feature size which will worsen the scalability of any search algorithms; and 3) expression profiles of certain features are typically highly correlated which may prevent to distinguish the predominant features. Unfortunately, most of the existing algorithms are partially addressing part of these challenges but not as a whole. In this paper, we propose a unified framework to address the above challenges. The classification and feature selection problem is first formulated as a nonconvex optimisation problem. Then the problem is relaxed and solved iteratively by a sequence of convex optimisation procedures which can be distributed computed and therefore allows the efficient implementation on advanced infrastructures. To illustrate the competence of our method over others, we first analyse a randomly generated simulation dataset under various conditions. We then analyse a real gene expression dataset on embryonal tumour. Further downstream analysis, such as functional annotation and pathway analysis, are performed on the selected features which elucidate several biological findings.
| S-EPMC5744982 | biostudies-literature