Vector Machine Recursive Feature Elimination Research Articles

Fast Gaussian kernel support vector machine recursive feature elimination algorithm

Gaussian kernel support vector machine recursive feature elimination (GKSVM-RFE) is a method for feature ranking in a nonlinear way. However, GKSVM-RFE suffers from the issue of high computational complexity, which hinders its applications. This paper investigates the issue of computational complexity in GKSVM-RFE, and proposes two fast versions for GKSVM-RFE, called fast GKSVM-RFE (FGKSVM-RFE), to speed up the procedure of recursive feature elimination in GKSVM-RFE. For this purpose, we design two kinds of ranking scores based on the first-order and second-order approximate schemes by introducing approximate Gaussian kernels. In iterations, FGKSVM-RFE fast calculates approximate ranking scores according to approximate schemes and ranks features based on approximate ranking scores. Experimental results reveal that our proposed methods can faster perform feature ranking than GKSVM-RFE and have compared performance to GKSVM-RFE.

Machine learning algorithm for characterizing risks of hypertension, at an early stage in Bangladesh

Background and aimsHypertension has become a major public health issue as the prevalence and risk of premature death and disability among adults due to hypertension has increased globally. The main objective is to characterize the risk factors of hypertension among adults in Bangladesh using machine learning (ML) algorithms. Materials and methodsThe hypertension data was derived from Bangladesh demographic and health survey, 2017–18, which included 6965 people aged 35 and above. Two most promising risk factor identification methods, namely least absolute shrinkage operator (LASSO) and support vector machine recursive feature elimination (SVMRFE) are implemented to detect the critical risk factors of hypertension. Additionally, four well-known ML algorithms as artificial neural network, decision tree, random forest, and gradient boosting (GB) have been used to predict hypertension. Performance scores of these algorithms were evaluated by accuracy, precision, recall, F-measure, and area under the curve (AUC). ResultsThe results clarify that age, BMI, wealth index, working status, and marital status for LASSO and age, BMI, marital status, diabetes and region for SVMRFE appear to be the top-most five significant risk factors for hypertension. Our findings reveal that the combination of SVMRFE-GB gives the maximum accuracy (66.98%), recall (97.92%), F-measure (78.99%), and AUC (0.669) compared to others. ConclusionGB-based algorithm confirms the best performer for prediction of hypertension, at an early stage in Bangladesh. Therefore, this study highly suggests that the policymakers make proper judgments for controlling hypertension using SVMRFE-GB-based combination to save time and reduce cost for Bangladeshi adults.

Identification of Key Genes Involved in Pancreatic Ductal Adenocarcinoma with Diabetes Mellitus Based on Gene Expression Profiling Analysis.

This study aimed to identify key genes involved in the progression of diabetic pancreatic ductal adenocarcinoma (PDAC). Two gene expression datasets (GSE74629 and GSE15932) were obtained from Gene Expression Omnibus. Then, differentially expressed genes (DEGs) between diabetic PDAC and non-diabetic PDAC were identified, followed by a functional analysis. Subsequently, gene modules related to DM were extracted by weighed gene co-expression network analysis. The protein-protein interaction (PPI) network for genes in significant modules was constructed and functional analyses were also performed. After that, the optimal feature genes were screened by support vector machine (SVM) recursive feature elimination and SVM classification model was built. Finally, survival analysis was conducted to identify prognostic genes. The correlations between prognostic genes and other clinical factors were also analyzed. Totally, 1546 DEGs with consistent change tendencies were identified and functional analyses showed they were strongly correlated with metabolic pathways. Furthermore, there were two significant gene modules, in which RPS27A and UBA52 were key genes. Functional analysis of genes in two gene modules revealed that these genes primarily participated in oxidative phosphorylation pathway. Additionally, 21 feature genes were closely related with diabetic PDAC and the corresponding SVM classifier markedly distinguished diabetic PDAC from non-diabetic PDAC patients. Finally, decreased KIF22 and PYGL levels had good survival outcomes for PDAC. Four genes (RPS27A, UBA52, KIF22 and PYGL) might be involved in the pathogenesis of diabetic PDAC. Furthermore, KIF22 and PYGL acted as prognostic biomarkers for diabetic PDAC.

A four-lncRNA signature for predicting prognosis of recurrence patients with gastric cancer.

PurposeThis study aimed to develop a multi-long noncoding RNA (lncRNA) signature for the prediction of gastric cancer (GC) based on differential gene expression between recurrence and nonrecurrence patients.MethodsBy repurposing microarray expression profiles of RNAs from The Cancer Genome Atlas (TCGA), we performed differential expression analysis between recurrence and nonrecurrence patients. A prognostic risk prediction model was constructed based on data from TCGA database, and its reliability was validated using data from Gene Expression Omnibus database. Furthermore, the lncRNA-associated competing endogenous RNA (ceRNA) network was constructed, namely, DIANA-LncBasev2 and starBase database.ResultsWe identified 363 differentially expressed RNAs (317 mRNAs, 18 lncRNAs, and 28 microRNAs [miRNAs]). Principal component analysis showed that the seven-feature lncRNAs screened by support vector machine–recursive feature elimination algorithm was more informative for predicting recurrence of GC in comparison with the eight-feature lncRNAs screened by random forest–out-of-bag algorithm. Four of the seven-feature lncRNAs including LINC00843, SNHG3, C21orf62-AS1, and MIR99AHG were chosen to develop a four-lncRNA risk score model. This risk score model was able to distinguish patients with high and low risk of recurrence, and was tested in two independent validation sets. The ceRNA network of this four-lncRNA signature included 10 miRNAs and 178 mRNAs. The mRNAs significantly related to the Wnt-signaling pathway and relevant biological processes.ConclusionA useful four-lncRNA signature recurrence was established to distinguish GC patients with high and low risk of recurrence. Regulating the relevant miRNAs and Wnt pathway might partly affect GC metastasisby.

Radiomics model of dual-time 2-[18F]FDG PET/CT imaging to distinguish between pancreatic ductal adenocarcinoma and autoimmune pancreatitis.

Pancreatic ductal adenocarcinoma (PDAC) and autoimmune pancreatitis (AIP) are diseases with a highly analogous visual presentation that are difficult to distinguish by imaging. The purpose of this research was to create a radiomics-based prediction model using dual-time PET/CT imaging for the noninvasive classification of PDAC and AIP lesions. This retrospective study was performed on 112 patients (48 patients with AIP and 64 patients with PDAC). All cases were confirmed by imaging and clinical follow-up, and/or pathology. A total of 502 radiomics features were extracted from the dual-time PET/CT images to develop a radiomics decision model. An additional 12 maximum intensity projection (MIP) features were also calculated to further improve the radiomics model. The optimal radiomics feature set was selected by support vector machine recursive feature elimination (SVM-RFE), and the final classifier was built using a linear SVM. The performance of the proposed dual-time model was evaluated using nested cross-validation for accuracy, sensitivity, specificity, and area under the curve (AUC). The final prediction model was developed from a combination of the SVM-RFE and linear SVM with the required quantitative features. The multimodal and multidimensional features performed well for classification (average AUC: 0.9668, accuracy: 89.91%, sensitivity: 85.31%, specificity: 96.04%). The radiomics model based on 2-[18F]fluoro-2-deoxy-D-glucose (2-[18F]FDG) PET/CT dual-time images provided promising performance for discriminating between patients with benign AIP and malignant PDAC lesions, which shows its potential for use as a diagnostic tool for clinical decision-making. • The clinical symptoms and imaging visual presentations of PDAC and AIP are highly similar, and accurate differentiation of PDAC and AIP lesions is difficult. • Radiomics features provided a potential noninvasive method for differentiation of AIP from PDAC. • The diagnostic performance of the proposed radiomics model indicates its potential to assist doctors in making treatment decisions.

Identification of TYR, TYRP1, DCT and LARP7 as related biomarkers and immune infiltration characteristics of vitiligo via comprehensive strategies

ABSTRACT This study aims to explore biomarkers associated with vitiligo and analyze the pathological role of immune cell infiltration in the disease. We used the robust rank aggregation (RRA) method to integrate three vitiligo data sets downloaded from gene expression omnibus database, identify the differentially expressed genes (DEGs) and analyze the functional correlation. Then, the comprehensive strategy of combined weighted gene coexpression network analysis (WGCNA) and logical regression of the selection operator (LASSO), support vector machine recursive feature elimination (SVM-RFE), and random forest (RF) machine learning algorithm are employed to screen and biomarkers associated with vitiligo. Finally, the immune cell infiltration of vitiligo was evaluated by CIBERSORT, and the correlation between biomarkers and infiltrating immune cells was analyzed. Herein, we identified 131 robust DEGs, and enrichment analysis results showed that robust DEGs and melanogenesis were closely associated with vitiligo development and progression. TYR, TYRP1, DCT and LARP7 were identified as vitiligo-related biomarkers. Immune infiltration analysis demonstrated that CD4 T Cell, CD8 T Cell, Tregs, NK cells, dendritic cells, and macrophages were involved in vitiligo’s pathogenesis. In summary, we adopted a comprehensive strategy to screen biomarkers related to vitiligo and explore the critical role of immune cell infiltration in vitiligo. Abbreviations: TYR, Tyrosinase; TYRP1, Tyrosinase-related protein-1; DCT, dopachrome tautomerase; LARP7, La ribonucleoprotein domain family, member-7; RRA, robust rank aggregation; DEGs, differentially expressed genes; WGCNA, weighted gene coexpression network analysis; LASSO, logical regression of the selection operator; SVM-RFE, support vector machine recursive feature elimination; RF, random forest; GWAS, Genome-wide association study; FasL, Fas-Fas ligand; Tregs, T-regulatory cells; NK, natural killer; GEPCs, gene expression profiling chips; GO, gene ontology; GSEA, gene set enrichment analysis; FDR, false discovery rate; AUC, area under the curve; ROC, receiver-operating characteristic; BP, biological process; CC, cellular component; MF, molecular function.

Estimating Soil Moisture Over Winter Wheat Fields During Growing Season Using Machine-Learning Methods

Soil moisture is vital for the crop growth and directly affects the crop yield. The conventional synthetic aperture radar (SAR) based soil moisture monitoring is often influenced by vegetation cover and surface roughness. The machine-learning methods are not constrained by physical parameters and have high nonlinear fitting capabilities. In this study, machine-learning methods were applied to estimate soil moisture over winter wheat fields during its growing season. RADARSAT-2 data with quad polarizations and 240 sample plots in the study area were acquired and collected, respectively. In addition to the four linear polarization channels, polarimetric decomposition parameters were extracted to expand the SAR feature space. Three advanced machine-learning models were selected and compared, which were support vector regression, random forests (RF), and gradient boosting regression tree. To improve the performances of the models, three feature-selection methods were compared, which were based on Pearson correlation, support vector machine recursive feature elimination, and RF, respectively. The coefficient of determination ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) and root-mean-square error (RMSE) were used to compare and assess the performances of those models. The results revealed that polarimetric decomposition parameters were effective in estimating soil moisture, and RF model obtained the highest prediction accuracy (training set: RMSE = 2.44 vol.% and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.94; and validation set: RMSE = 4.03 vol.%, and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.79). This study finally concluded that using polarimetric decomposition parameters combined with machine-learning and feature-selection methods could effectively estimate soil moisture at a high accuracy, which helps monitor soil moisture across the agricultural field during its growing season.

Potential role of a three-gene signature in predicting diagnosis in patients with myocardial infarction

ABSTRACT In this study, we evaluated the diagnostic value of key genes in myocardial infarction (MI) based on data from the Gene Expression Omnibus (GEO) database. We used data from GSE66360 to identify a set of significant differentially expressed genes (DEGs) between MI and healthy controls. Logistic regression, least absolute shrinkage and selection operator (LASSO) regression, support vector machine recursive feature elimination (SVM-RFE), and SignalP 3.0 server were used to identify the potential role of genes in predicting diagnosis in patients with MI. Principal component analysis (PCA), receiver operating characteristic (ROC) curve analyses, area under the curve (AUC) analyses, and C-index were used to estimate the diagnostic value of genes in patients with MI. The association was validated using six other independent data sets. Subsequently, bioinformatics analysis was conducted based on the aforementioned potential genes. A meta-analysis was performed to evaluate the diagnostic value of the genes in MI. Forty-four DEGs were selected from the GSE66360 dataset. A three-gene signature consisting of CCL20, IL1R2, and ITLN1 could effectively distinguish patients with MI. The three-gene signature was validated in seven independent cohorts. Functional enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were performed to reveal the involvement of the three-gene signature in inflammation-related biological processes and pathways. Moreover, diagnostic meta-analysis results of the three-gene signature showed that the pooled sensitivity, specificity, and AUC for MI were 0.80, 0.90, and 0.93, respectively. These results suggest that the three-gene signature is a novel candidate biomarker for distinguishing MI from healthy controls.

Binary biogeography-based optimization based SVM-RFE for feature selection

Rapid data growth presents many challenges for Machine Learning (ML) tasks as it can include lots of irrelevant, noisy, and redundant features. Thus, it is vital to select the most relevant features to the classification task, known as Feature Selection (FS). The main goal of FS techniques is to maximize the performance of a classification task while keeping the number of features to a minimum. In this study, a hybrid metaheuristic model is designed to solve FS problems based on Binary Biogeography Optimization (BBO) followed by the application of Support Vector Machine Recursive Feature Elimination (SVM-RFE), known as BBO-SVM-RFE. The SVM-RFE is embedded into the BBO to improve the quality of the obtained solutions in the mutation operator in order to enhance the exploitation capability as well as striking an adequate balance between exploitation and exploration of the original BBO. The proposed BBO-SVM-RFE method for solving FS problems was assessed on eighteen benchmark datasets. Comparative results showed that the BBO-SVM-RFE method outperforms the BBO method and other existing wrapper and filter methods in terms of accuracy and number of selected features. The obtained results reveal the high potentiality of BBO-SVM-RFE in reliably searching the feature space to obtain the optimal combination of features.

Single-Beat Measurement of Left Ventricular Contractility in Normothermic Ex Situ Perfused Porcine Hearts.

For heart transplantation, donor heart status needs to be evaluated during normothermic ex situ perfusion (ESHP). Left ventricular end-systolic elastance (E es) measures the left ventricular contractile function, but its estimation requires the occlusion of the left atrium line in the ESHP, which may cause unnecessary damage to the donor heart. We present a novel method to quantify E es based on hemodynamic parameters obtained from only one steady-state PV loop in ESHP. E es was obtained by the end-systolic point (P es, V es) and the volume axis intercept point of E es (V 0). V 0 was estimated through the support vector machine regression (SVR) method using parameters derived from the measured steady-state PV loop. To achieve high V 0 estimation accuracy, a filter-based support vector machine recursive feature elimination method (SVM-RFE) algorithm selected the parameters for V 0 estimation. Hemodynamic parameter samples (n = 101) obtained from ESHP experiments with pig's hearts were used to train the E es calculation model. Early post-transplantation outcomes in six heart transplantation experiments were then estimated from the trained E es calculation model. E es calculated by the proposed method agreed well with conventional multi-beat estimates obtained by the occlusion process (r = 0.88, p <0.001, n = 101) and was capable of predicting the early post-transplant cardiac index (r = 0.84, p <0.05, n = 6). This method effectively assesses left ventricular contractility during ESHP and predicts early post-transplant outcomes in the porcine model. Our approach is the first to quantify E es by estimating V 0 from steady-state beats in ESHP for accurately predicting early post-transplantation outcomes.

A Model Using Support Vector Machines Recursive Feature Elimination (SVM-RFE) Algorithm to Classify Whether COPD Patients Have Been Continuously Managed According to GOLD Guidelines.

PurposePatients with chronic obstructive pulmonary disease (COPD) would have a poor prognosis if they were not continuously managed according to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines. We aim to develop a model to classify whether COPD patients have been continuously managed according to GOLD in the previous year.MethodsThe Managed group were COPD patients from a prospective cohort from November 2017 to November 2019, who have been continuously managed according to GOLD for 1 year. The Control group were COPD patients who were not continuously managed according to GOLD. They were from a retrospective cohort from October 2016 to October 2017 in the same hospitals as the Managed group. A synthetic minority over-sampling technique (SMOTE) algorithm was used to up-sample the Managed group in a training dataset. Features for classification were selected using a support vector machine recursive feature elimination (SVM-RFE) algorithm. The classification model was developed using LibSVM, and its performance was assessed on the testing dataset.ResultsThe final analysis included 15 subjects in the Managed group and 191 in the Control group. SVM-RFE selects nine features including smoking history, post-bronchodilator (post-)FVC before management, and those after 1-year follow-up (BMI, moderate and severe AECOPD frequency in previous 12 months, mMRC score, post-FEV1, post-FEV1%pred, post-FVC, and post-FEV1/FVC). For our model, positive predictive value is 66.7%, F1 score is 0.978, and AUC is 0.987.ConclusionSVM classifier combined with SVM-REF feature selection algorithm could achieve good classification between COPD patients who are or are not continuously managed. This model could be applied in clinical practice to help doctors make decisions and enhance COPD patients’ compliance with standard treatment.

Predicting Diagnostic Gene Biomarkers Associated With Immune Infiltration in Patients With Acute Myocardial Infarction.

Objective: The present study was designed to identify potential diagnostic markers for acute myocardial infarction (AMI) and determine the significance of immune cell infiltration in this pathology.Methods: Two publicly available gene expression profiles (GSE66360 and GSE48060 datasets) from human AMI and control samples were downloaded from the GEO database. Differentially expressed genes (DEGs) were screened between 80 AMI and 71 control samples. The LASSO regression model and support vector machine recursive feature elimination (SVM-RFE) analysis were performed to identify candidate biomarkers. The area under the receiver operating characteristic curve (AUC) value was obtained and used to evaluate discriminatory ability. The expression level and diagnostic value of the biomarkers in AMI were further validated in the GSE60993 dataset (17 AMI patients and 7 controls). The compositional patterns of the 22 types of immune cell fraction in AMI were estimated based on the merged cohorts using CIBERSORT.Results: A total of 27 genes were identified. The identified DEGs were mainly involved in carbohydrate binding, Kawasaki disease, atherosclerosis, and arteriosclerotic cardiovascular disease. Gene sets related to atherosclerosis signaling, primary immunodeficiency, IL-17, and TNF signaling pathways were differentially activated in AMI compared with the control. IL1R2, IRAK3, and THBD were identified as diagnostic markers of AMI (AUC = 0.877) and validated in the GSE60993 dataset (AUC = 0.941). Immune cell infiltration analysis revealed that IL1R2, IRAK3, and THBD were correlated with M2 macrophages, neutrophils, monocytes, CD4+ resting memory T cells, activated natural killer (NK) cells, and gamma delta T cells.Conclusion: IL1R2, IRAK3, and THBD can be used as diagnostic markers of AMI, and can provide new insights for future studies on the occurrence and the molecular mechanisms of AMI.

Hybrid adapted fast correlation FCBF-support vector machine recursive feature elimination for feature selection

This paper investigates feature selection methods based on hybrid architecture using feature selection algorithm called Adapted Fast Correlation Based Feature selection and Support Vector Machine Recursive Feature Elimination (AFCBF-SVMRFE). The AFCBF-SVMRFE has three stages and composed of SVMRFE embedded method with Correlation based Features Selection. The first stage is the relevance analysis, the second one is a redundancy analysis, and the third stage is a performance evaluation and features restoration stage. Experiments show that the proposed method tested on different classifiers: Support Vector Machine SVM and K nearest neighbors KNN provide a best accuracy on various dataset. The SVM classifier outperforms KNN classifier on these data. The AFCBF-SVMRFE outperforms FCBF multivariate filter, SVMRFE, Particle swarm optimization PSO and Artificial bees colony ABC.

Combined Metabolomic Analysis of Plasma and Tissue Reveals a Prognostic Risk Score System and Metabolic Dysregulation in Esophageal Squamous Cell Carcinoma

Background: Esophageal squamous cell carcinoma (ESCC) is a gastrointestinal malignancy with a poor prognosis. Although studies have shown metabolic reprogramming to be linked to ESCC development, no prognostic metabolic biomarkers or potential therapeutic metabolic targets have been identified.Method: The present study investigated some circulating metabolites associated with overall survival in 276 curatively resected ESCC patients using liquid chromatography/mass spectrometry metabolomics and Kaplan-Meier analysis. Tissue metabolomic analysis of 23-paired ESCC tissue samples was performed to discover metabolic dysregulation in ESCC cancerous tissue. A method consisting of support vector machine recursive feature elimination and LIMMA differential expression analysis was utilized to select promising feature genes within transcriptomic data from 179-paired ESCC tissue samples. Joint pathway analysis with genes and metabolites identified relevant metabolic pathways and targets for ESCC.Results: Four metabolites, kynurenine, 1-myristoyl-glycero-3-phosphocholine (LPC(14:0)sn-1), 2-piperidinone, and hippuric acid, were identified as prognostic factors in the preoperative plasma from ESCC patients. A risk score consisting of kynurenine and LPC(14:0)sn-1 significantly improved the prognostic performance of the tumor-node-metastasis staging system and was able to stratify risk for ESCC. Combined tissue metabolomic analysis and support vector machine recursive feature elimination gene selection revealed dysregulated kynurenine pathway as an important metabolic feature of ESCC, including accumulation of tryptophan, formylkynurenine, and kynurenine, as well as up-regulated indoleamine 2,3-dioxygenase 1 in ESCC cancerous tissue.Conclusions: This work identified for the first time four potential prognostic circulating metabolites. In addition, kynurenine pathway metabolism was shown to be up-regulated tryptophan-kynurenine metabolism in ESCC. Results not only provide a metabolite-based risk score system for prognosis, but also improve the understanding of the molecular basis of ESCC onset and progression, and as well as novel potential therapeutic targets for ESCC.

Statistical feature and channel selection for upper limb classification using sEMG signal processing

Feature and channel selection stage plays an important role in identifying movement intent through surface electromyography (sEMG), once the best feature or features set and channel ensemble is still an open topic in this research field. In this work, we present three different strategies to feature and channel selection used in movement recognition through sEMG processing. This paper analyzes the feature selection stage, considering the relevance of each feature-channel pair to movement characterization. Three feature selection methods based on support vector machine recursive feature elimination (SVM-RFE), Monte Carlo feature selection, and singular value decomposition (SVD) entropy that are benchmark techniques for DNA microarray data analysis formed a ranking of feature/channel pairs. The ranking, organized according to their importance to movement classification, forms a heuristic for a wrapper method based feature selection algorithm, which uses the regularized extreme learning machine (RELM) classifier for the movement recognition. To test the proposed methods, we used the NINAPro Databases 2 and 3 (40 intact and 10 amputee volunteers, respectively). The results derived from the technique reached average accuracies of 84.9%, 84.0%, and 83.9% for the SVM-RFE, Monte Carlo, and SVD entropy ranking methods, respectively, for database 2. At the same time, the same methods achieved 74.8%, 74.5%, and 74.1% of accuracy for database 3, respectively. In terms of most selected features, zero crossings (TZC), mean frequency (FMN), signal slope changes (TSLPCH2), and power (TPWR) were the most frequently selected. The results obtained showed no statistically significant difference among the three feature selection algorithms implemented regarding the classification accuracy. The introduction of the pre-processing steps in the classification process provided promising classification accuracy levels, outperforming the results presented by related works with the same NINAPro database.

A CpG Methylation Classifier to Predict Relapse in Adults with T-Cell Lymphoblastic Lymphoma.

Adults with T-cell lymphoblastic lymphoma (T-LBL) generally benefit from treatment with acute lymphoblastic leukemia (ALL)-like regimens, but approximately 40% will relapse after such treatment. We evaluated the value of CpG methylation in predicting relapse for adults with T-LBL treated with ALL-like regimens. A total of 549 adults with T-LBL from 27 medical centers were included in the analysis. Using the Illumina Methylation 850K Beadchip, 44 relapse-related CpGs were identified from 49 T-LBL samples by two algorithms: least absolute shrinkage and selector operation (LASSO) and support vector machine-recursive feature elimination (SVM-RFE). We built a four-CpG classifier using LASSO Cox regression based on association between the methylation level of CpGs and relapse-free survival in the training cohort (n = 160). The four-CpG classifier was validated in the internal testing cohort (n = 68) and independent validation cohort (n = 321). The four-CpG-based classifier discriminated patients with T-LBL at high risk of relapse in the training cohort from those at low risk (P < 0.001). This classifier also showed good predictive value in the internal testing cohort (P < 0.001) and the independent validation cohort (P < 0.001). A nomogram incorporating five independent prognostic factors including the CpG-based classifier, lactate dehydrogenase levels, Eastern Cooperative Oncology Group performance status, central nervous system involvement, and NOTCH1/FBXW7 status showed a significantly higher predictive accuracy than each single variable. Stratification into different subgroups by the nomogram helped identify the subset of patients who most benefited from more intensive chemotherapy and/or sequential hematopoietic stem cell transplantation. Our four-CpG-based classifier could predict disease relapse in patients with T-LBL, and could be used to guide treatment decision.

Evaluation of global climate models for precipitation projection in sub-Himalaya region of Pakistan

The selection of global climate models (GCMs) for a region remained a difficult step in climate change studies. A state-of-the-art Support Vector Machine Recursive Feature Elimination (SVM-RFE) algorithm is proposed in this paper for GCM selection. The ranking of GCMs obtained using SVM-RFE was compared to that obtained using entropy-based similarity assessment index known as Symmetrical Uncertainty (SU). The study was conducted in the sub-Himalayan region of Pakistan where a reliable projection of climate is highly significant for water resources management in the entire western part of South Asia. The RF-based regression model was employed to generate a multi-model ensemble (MME) mean of the top-ranked GCMs. The MME mean projection was utilized to estimate the spatiotemporal changes in annual precipitation in comparison with precipitation of 1961‐–2000 for various representative concentration pathway (RCP) scenarios. The SVM-RF selected five GCMs (MIROC5, EC-EARTH, CNRM-CM5, BCC-CSM1.1(m) and BCC-CSM1.1) as most suitable for climate change projections in the study area. Obtained results were found to collaborate well with the results of multiple conventional statistical metrics. The MME mean projections revealed precipitation alteration between −1% and 18% during 2020‐–2059, and 0 and 24% during 2060–2099 for different RCPs. Precipitation was projected to increase up to 20% in the north whereas a decrease up-to −16% in the south.

A novel risk score model based on eight genes and a nomogram for predicting overall survival of patients with osteosarcoma

BackgroundThis study aims to identify a predictive model to predict survival outcomes of osteosarcoma (OS) patients.MethodsA RNA sequencing dataset (the training set) and a microarray dataset (the validation set) were obtained from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) database, respectively. Differentially expressed genes (DEGs) between metastatic and non-metastatic OS samples were identified in training set. Prognosis-related DEGs were screened and optimized by support vector machine (SVM) recursive feature elimination. A SVM classifier was built to classify metastatic and non-metastatic OS samples. Independent prognosic genes were extracted by multivariate regression analysis to build a risk score model followed by performance evaluation in two datasets by Kaplan-Meier (KM) analysis. Independent clinical prognostic indicators were identified followed by nomogram analysis. Finally, functional analyses of survival-related genes were conducted.ResultTotally, 345 DEGs and 45 prognosis-related genes were screened. A SVM classifier could distinguish metastatic and non-metastatic OS samples. An eight-gene signature was an independent prognostic marker and used for constructing a risk score model. The risk score model could separate OS samples into high and low risk groups in two datasets (training set: log-rank p < 0.01, C-index = 0.805; validation set: log-rank p < 0.01, C-index = 0.797). Tumor metastasis and RS model status were independent prognostic factors and nomogram model exhibited accurate survival prediction for OS. Additionally, functional analyses of survival-related genes indicated they were closely associated with immune responses and cytokine-cytokine receptor interaction pathway.ConclusionAn eight-gene predictive model and nomogram were developed to predict OS prognosis.

SigFeature: Novel Significant Feature Selection Method for Classification of Gene Expression Data Using Support Vector Machine and t Statistic.

Biological data are accumulating at a faster rate, but interpreting them still remains a problem. Classifying biological data into distinct groups is the first step in understanding them. Data classification in response to a certain treatment is an extremely important aspect for differentially expressed genes in making present/absent calls. Many feature selection algorithms have been developed including the support vector machine recursive feature elimination procedure (SVM-RFE) and its variants. Support vector machine RFEs are greedy methods that attempt to find superlative possible combinations leading to binary classification, which may not be biologically significant. To overcome this limitation of SVM-RFE, we propose a novel feature selection algorithm, termed as “sigFeature” (https://bioconductor.org/packages/sigFeature/), based on SVM and t statistic to discover the differentially significant features along with good performance in classification. The “sigFeature” R package is centered around a function called “sigFeature,” which provides automatic selection of features for the binary classification. Using six publicly available microarray data sets (downloaded from Gene Expression Omnibus) with different biological attributes, we further compared the performance of “sigFeature” to three other feature selection algorithms. A small number of selected features (by “sigFeature”) also show higher classification accuracy. For further downstream evaluation of its biological signature, we conducted gene set enrichment analysis with the selected features (genes) from “sigFeature” and compared it with the outputs of other algorithms. We observed that “sigFeature” is able to predict the signature of four out of six microarray data sets accurately, whereas the other algorithms predict less data set signatures. Thus, “sigFeature” is considerably better than related algorithms in discovering differentially significant features from microarray data sets.

Multiclass classification of acute lymphoblastic leukemia microarrays data using support vector machine algorithms

Acute lymphoblastic leukemia (ALL) is a form of leukemia, or cancer of the white blood cells characterized by excess lymphoblast. Classification of acute lymphoblastic leukemia subtypes based on fusion genes that have a translocation. The fusion genes are BCR-ABL, E2A-PBX1, Hyperdiploid > 50 chromosomes, MLL, T-ALL, and TEL-AML1. The classification of acute lymphoblastic leukemia subtypes has an important role for the type of treatment that will be received, duration of treatment, medication needed during treatment, and other treatments that may be needed. In this paper, the method used is Multiclass Support Vector Machine Recursive Feature Elimination (MSVM-RFE) as the feature selection and One-Against-One Multiclass Support Vector Machine (OAO-MSVM) with RBF-Kernel with σ = 0.01 and Polynomial-Kernel with d = 4 as the classification methods. For the multiclass classification of acute lymphoblastic leukemia microarrays data, the best method to use is the MSVM Polynomial-Kernel with d = 4 that produces overall accuracy about 94%, precision about 96%, recall about 95%, F1 score about 95%, and the running time is 0.66 seconds.