Two-step Feature Selection Research Articles

A two-step feature selection procedure for relevant markers of Squamous Cell Lung Carcinoma using different survival models

There are potentially infinite gene expression markers for Lung Squamous Cell Carcinoma. This results in a high-dimensional data with a large number of features. The selection of relevant markers for analysis is thus, of utmost importance. In our study, we have aimed to select a subset of prominent and significant features from 31918 features of gene expressions. Analysis is then performed on the selected features using the Cox Proportional Hazards Model to know how each marker affects the survival estimates of a patient. We have employed a two-step selection process to select a subset of markers. The first step is done by L1 regularized Cox PH. Then the selected markers are screened a second time by running a univariate Cox PH model and checking for the p-value of each bio-marker via Wald inference (p<0.05). Once the final selection is made, we estimate the Hazard Ratio and Confidence intervals using Maximum Likelihood Estimates (MLE) and the Bayesian Approach with the Cox Proportional Hazards Model (CPH) and the Accelerated Failure Time Model (AFT) as an alternative. A forest plot has also been generated to show the graphical representation of the meta-analysis done in the study. With the proposed selection procedure we have managed to find a suitable subset out of a large number of variables available. The features selected have been analyzed and their validity has been confirmed by using survival models.

An integrative machine learning framework for classifying SEER breast cancer

Breast cancer is the commonest type of cancer in women worldwide and the leading cause of mortality for females. The aim of this research is to classify the alive and death status of breast cancer patients using the Surveillance, Epidemiology, and End Results dataset. Due to its capacity to handle enormous data sets systematically, machine learning and deep learning has been widely employed in biomedical research to answer diverse classification difficulties. Pre-processing the data enables its visualization and analysis for use in making important decisions. This research presents a feasible machine learning-based approach for categorizing SEER breast cancer dataset. Moreover, a two-step feature selection method based on Variance Threshold and Principal Component Analysis was employed to select the features from the SEER breast cancer dataset. After selecting the features, the classification of the breast cancer dataset is carried out using Supervised and Ensemble learning techniques such as Ada Boosting, XG Boosting, Gradient Boosting, Naive Bayes and Decision Tree. Utilizing the train-test split and k-fold cross-validation approaches, the performance of various machine learning algorithms is examined. The accuracy of Decision Tree for both train-test split and cross validation achieved as 98%. In this study, it is observed that the Decision Tree algorithm outperforms other supervised and ensemble learning approaches for the SEER Breast Cancer dataset.

I2OM: Toward a better prediction of 2′-O-methylation in human RNA

2′-O-methylation (2OM) is an omnipresent post-transcriptional modification in RNAs. It is important for the regulation of RNA stability, mRNA splicing and translation, as well as innate immunity. With the increase in publicly available 2OM data, several computational tools have been developed for the identification of 2OM sites in human RNA. Unfortunately, these tools suffer from the low discriminative power of redundant features, unreasonable dataset construction or overfitting. To address those issues, based on four types of 2OM (2OM-adenine (A), cytosine (C), guanine (G), and uracil (U)) data, we developed a two-step feature selection model to identify 2OM. For each type, the one-way analysis of variance (ANOVA) combined with mutual information (MI) was proposed to rank sequence features for obtaining the optimal feature subset. Subsequently, four predictors based on eXtreme Gradient Boosting (XGBoost) or support vector machine (SVM) were presented to identify the four types of 2OM sites. Finally, the proposed model could produce an overall accuracy of 84.3 % on the independent set. To provide a convenience for users, an online tool called i2OM was constructed and can be freely access at i2om.lin-group.cn. The predictor may provide a reference for the study of the 2OM.

A Two-Step Feature Selection Radiomic Approach to Predict Molecular Outcomes in Breast Cancer.

Breast Cancer (BC) is the most common cancer among women worldwide and is characterized by intra- and inter-tumor heterogeneity that strongly contributes towards its poor prognosis. The Estrogen Receptor (ER), Progesterone Receptor (PR), Human Epidermal Growth Factor Receptor 2 (HER2), and Ki67 antigen are the most examined markers depicting BC heterogeneity and have been shown to have a strong impact on BC prognosis. Radiomics can noninvasively predict BC heterogeneity through the quantitative evaluation of medical images, such as Magnetic Resonance Imaging (MRI), which has become increasingly important in the detection and characterization of BC. However, the lack of comprehensive BC datasets in terms of molecular outcomes and MRI modalities, and the absence of a general methodology to build and compare feature selection approaches and predictive models, limit the routine use of radiomics in the BC clinical practice. In this work, a new radiomic approach based on a two-step feature selection process was proposed to build predictors for ER, PR, HER2, and Ki67 markers. An in-house dataset was used, containing 92 multiparametric MRIs of patients with histologically proven BC and all four relevant biomarkers available. Thousands of radiomic features were extracted from post-contrast and subtracted Dynamic Contrast-Enanched (DCE) MRI images, Apparent Diffusion Coefficient (ADC) maps, and T2-weighted (T2) images. The two-step feature selection approach was used to identify significant radiomic features properly and then to build the final prediction models. They showed remarkable results in terms of F1-score for all the biomarkers: 84%, 63%, 90%, and 72% for ER, HER2, Ki67, and PR, respectively. When possible, the models were validated on the TCGA/TCIA Breast Cancer dataset, returning promising results (F1-score = 88% for the ER+/ER- classification task). The developed approach efficiently characterized BC heterogeneity according to the examined molecular biomarkers.

MSEs Credit Risk Assessment Model Based on Federated Learning and Feature Selection

Federated learning has been used extensively in business innovation scenarios in various industries. This research adopts the federated learning approach for the first time to address the issue of bank-enterprise information asymmetry in the credit assessment scenario. First, this research designs a credit risk assessment model based on federated learning and feature selection for micro and small enterprises (MSEs) using multi-dimensional enterprise data and multi-perspective enterprise information. The proposed model includes four main processes: namely encrypted entity alignment, hybrid feature selection, secure multi-party computation, and global model updating. Secondly, a two-step feature selection algorithm based on wrapper and filter is designed to construct the optimal feature set in multi-source heterogeneous data, which can provide excellent accuracy and interpretability. In addition, a local update screening strategy is proposed to select trustworthy model parameters for aggregation each time to ensure the quality of the global model. The results of the study show that the model error rate is reduced by 6.22% and the recall rate is improved by 11.03% compared to the algorithms commonly used in credit risk research, significantly improving the ability to identify defaulters. Finally, the business operations of commercial banks are used to confirm the potential of the proposed model for real-world implementation.

Dietary benzoic acid and supplemental enzymes alter fiber-fermenting taxa and metabolites in the cecum of weaned pigs.

Inclusion of enzymes and organic acids in pig diets is an important strategy supporting decreased antibiotic usage in pork production. However, limited knowledge exists about how these additives impact intestinal microbes and their metabolites. To examine the effects of benzoic acid and enzymes on gut microbiota and metabolome, 160 pigs were assigned to one of four diets 7 days after weaning: a control diet or the addition of 0.5% benzoic acid, 0.045% dietary enzymes (phytase, β-glucanase, xylanase, and α-amylase), or both and fed ad libitum for 21 to 22 d. Individual growth performance and group diarrhea incidence data were collected throughout the experimental period. A decrease of 20% in pen-level diarrhea incidence from days 8 to 14 in pigs-fed both benzoic acid and enzymes compared to the control diet (P = 0.047). Cecal digesta samples were collected at the end of the experimental period from 40 piglets (n = 10 per group) and evaluated for differences using 16S rRNA sequencing and two-dimensional gas chromatography and time-of-flight mass spectrometry (GCxGC-TOFMS). Analysis of cecal microbiota diversity revealed that benzoic acid altered microbiota composition (Unweighted Unifrac, P = 0.047, r2 = 0.07) and decreased α-diversity (Shannon, P = 0.041; Faith's Phylogenetic Diversity, P = 0.041). Dietary enzymes increased fiber-fermenting bacterial taxa such as Prevotellaceae. Two-step feature selection identified 17 cecal metabolites that differed among diets, including increased microbial cross-feeding product 1,2-propanediol in pigs-fed benzoic acid-containing diets. In conclusion, dietary benzoic acid and enzymes affected the gut microbiota and metabolome of weaned pigs and may support the health and resolution of postweaning diarrhea.

A Hybrid Feature Reduction Approach for Medical Decision Support System

Feature reduction is essential at the preprocessing stage of designing any reliable and fast disease diagnosis model. Addressing the limitations like disease specificity, information loss, and operating NP problem in polynomial time, this paper introduces a two-step hybrid feature selection approach to identify a subset of most relevant and contributing features of each medical dataset for constructing diagnostic model. The concept of information gain is used in Step I to select the informative features, whereas a correlation coefficient-based approach is employed in Step II to retain the informative features possessing much dependency with class attribute but less dependency among the non-class attributes. In particular, both the approaches are sequentially fused to select approximately optimal features in order to construct better classification model in terms of performance and time. The optimal threshold criteria are decided to choose the most appropriate features from the datasets. The effectiveness of the proposed approach is assessed using six individual competent learners and one ensemble learner over seventeen disease datasets of smaller to larger dimensions. The empirical results indicate that the proposed approach improves the performance over the datasets after feature selection, reducing considerable amount of irrelevant and redundant data.

A real-time IoT-based botnet detection method using a novel two-step feature selection technique and the support vector machine classifier

A botnet, which is a collection of devices polluted by malicious software programs, is among the top security challenges in the Internet of Things (IoT) environments. Therefore, to deal with such an abnormality in these environments, different machine learning-based studies, which have resulted in outstanding findings, have been carried out and applied to predict the botnets. However, the existing techniques may still suffer from three main limitations. First, some of them are not suitable for real-time applications because they spend a lot of time to determine the normal/abnormal traffic. Second, the functionality of some of the approaches is not satisfactory because they ignore/do not utilize the efficient feature selection methods. Third, these studies have usually focused on generating a binary botnet prediction model without taking the attack types into consideration. To reduce the botnet prediction time and address the second and third restrictions, the present study suggested a two-step machine learning method designed based on our previously developed optimization algorithm (WCC) and the support vector machine classifier. The outcomes indicated that the proposed method outperforms the existing approaches since it can precisely classify the data streams into their related groups and make a trade-off between the total number of the selected features and the performance of the prediction model. The results also showed that IP addresses, source ports as well as destination hosts-related features, and the total number of the transferred data streams and their statistical measurements are possible key factors in identifying botnet traffics.

Rumour identification on Twitter as a function of novel textual and language-context features.

Social microblogs are one of the popular platforms for information spreading. However, with several advantages, these platforms are being used for spreading rumours. At present, the majority of existing approaches identify rumours at the topic level instead of at the tweet/post level. Moreover, prior studies used the sentiment and linguistic features for rumours identification without considering discrete positive and negative emotions and effective part-of-speech features in content-based approaches. Similarly, the majority of prior studies used content-based approaches for feature generation, and recent context-based approaches were not explored. To cope with these challenges, a robust framework for rumour detection at the tweet level is designed in this paper. The model used word2vec embeddings and bidirectional encoder representations from transformers method (BERT) from context-based and discrete emotions, linguistic, and metadata characteristics from content-based approaches. According to our knowledge, we are the first ones who used these features for rumour identification at the tweet/post level. The framework is tested on four real-life twitter microblog datasets. The results show that the detection model is capable of detecting 97%, 86%, 85%, and 80% of rumours on four datasets respectively. In addition, the proposed framework outperformed the three latest state-of-the-art baselines. BERT model presented the best performance among context-based approaches, and linguistic features are best performing among content-based approaches as a stand-alone model. Moreover, the utilization of two-step feature selection further improves the detection model performance.

Interpretable machine learning identification of arginine methylation sites

Protein methylation is one of the most prominent posttranslation modifications that essentially regulates several biological processes in eukaryotes. Therefore, identification of the arginine methylation site is crucial in deciphering its characteristics and functions in cell biology, disease mechanisms, and guided drug development. The computation methods address the long-term bottleneck together with the cost, time, and labor required in experimental methods for large-scale identification of protein arginine methylation sites. In this study, we proposed a robust machine learning-based computational tool known as iIRMethyl, employing the primary sequence and physicochemical properties of protein along with a two-step feature selection method for optimal selection of feature descriptors. Moreover, the performance of iIRMethyl was comprehensively evaluated via k-fold cross-validation on a benchmark dataset and independent test dataset. iIRMethyl demonstrated a remarkably greater performance than the state-of-the-art method and achieved an average area under the curve value of 0.99 for both k-fold cross-validation and an independent test set in the identification of protein arginine methylation sites. Furthermore, the outcomes reveal that iIRMethyl is a robust and accurate computational tool for large-scale identification of arginine methylation sites and would facilitate the understanding of their functional mechanisms and accelerating their application in drug development and clinical therapy. Additionally, the prediction mechanism of the proposed model iIRMethyl is interpreted using the SHapley Additive exPlanation algorithm.

Sequence-Based Prediction with Feature Representation Learning and Biological Function Analysis of Channel Proteins.

Channel proteins are proteins that can transport molecules past the plasma membrane through free diffusion movement. Due to the cost of labor and experimental methods, developing a tool to identify channel proteins is necessary for biological research on channel proteins. 17 feature coding methods and four machine learning classifiers to generate 68-dimensional data probability features. Then, the two-step feature selection strategy was used to optimize the features, and the final prediction Model M16-LGBM (light gradient boosting machine) was obtained on the 16-dimensional optimal feature vector. A new predictor, CAPs-LGBM, was proposed to identify the channel proteins effectively. CAPs-LGBM is the first channel protein machine learning predictor was used to construct the final prediction model based on protein primary sequences. The classifier performed well in the training and test sets.

ATGPred-FL: sequence-based prediction of autophagy proteins with feature representation learning.

Autophagy plays an important role in biological evolution and is regulated by many autophagy proteins. Accurate identification of autophagy proteins is crucially important to reveal their biological functions. Due to the expense and labor cost of experimental methods, it is urgent to develop automated, accurate and reliable sequence-based computational tools to enable the identification of novel autophagy proteins among numerous proteins and peptides. For this purpose, a new predictor named ATGPred-FL was proposed for the efficient identification of autophagy proteins. We investigated various sequence-based feature descriptors and adopted the feature learning method to generate corresponding, more informative probability features. Then, a two-step feature selection strategy based on accuracy was utilized to remove irrelevant and redundant features, leading to the most discriminative 14-dimensional feature set. The final predictor was built using a support vector machine classifier, which performed favorably on both the training and testing sets with accuracy values of 94.40% and 90.50%, respectively. ATGPred-FL is the first ATG machine learning predictor based on protein primary sequences. We envision that ATGPred-FL will be an effective and useful tool for autophagy protein identification, and it is available for free at http://lab.malab.cn/~acy/ATGPred-FL , the source code and datasets are accessible at https://github.com/jiaoshihu/ATGPred .

SCORPION is a stacking-based ensemble learning framework for accurate prediction of phage virion proteins

Fast and accurate identification of phage virion proteins (PVPs) would greatly aid facilitation of antibacterial drug discovery and development. Although, several research efforts based on machine learning (ML) methods have been made for in silico identification of PVPs, these methods have certain limitations. Therefore, in this study, we propose a new computational approach, termed SCORPION, (StaCking-based Predictior fOR Phage VIrion PrOteiNs), to accurately identify PVPs using only protein primary sequences. Specifically, we explored comprehensive 13 different feature descriptors from different aspects (i.e., compositional information, composition-transition-distribution information, position-specific information and physicochemical properties) with 10 popular ML algorithms to construct a pool of optimal baseline models. These optimal baseline models were then used to generate probabilistic features (PFs) and considered as a new feature vector. Finally, we utilized a two-step feature selection strategy to determine the optimal PF feature vector and used this feature vector to develop a stacked model (SCORPION). Both tenfold cross-validation and independent test results indicate that SCORPION achieves superior predictive performance than its constitute baseline models and existing methods. We anticipate SCORPION will serve as a useful tool for the cost-effective and large-scale screening of new PVPs. The source codes and datasets for this work are available for downloading in the GitHub repository (https://github.com/saeed344/SCORPION).

Multi-view learning for lymph node metastasis prediction using tumor and nodal radiomics in gastric cancer

Purpose. This study aims to develop and validate a multi-view learning method by the combination of primary tumor radiomics and lymph node (LN) radiomics for the preoperative prediction of LN status in gastric cancer (GC). Methods. A total of 170 contrast-enhanced abdominal CT images from GC patients were enrolled in this retrospective study. After data preprocessing, two-step feature selection approach including Pearson correlation analysis and supervised feature selection method based on test-time budget (FSBudget) was performed to remove redundance of tumor and LN radiomics features respectively. Two types of discriminative features were then learned by an unsupervised multi-view partial least squares (UMvPLS) for a latent common space on which a logistic regression classifier is trained. Five repeated random hold-out experiments were employed. Results. On 20-dimensional latent common space, area under receiver operating characteristic curve (AUC), precision, accuracy, recall and F1-score are 0.9531 ± 0.0183, 0.9260 ± 0.0184, 0.9136 ± 0.0174, 0.9468 ± 0.0106 and 0.9362 ± 0.0125 for the training cohort respectively, and 0.8984 ± 0.0536, 0.8671 ± 0.0489, 0.8500 ± 0.0599, 0.9118 ± 0.0550 and 0.8882 ± 0.0440 for the validation cohort respectively (reported as mean ± standard deviation). It shows a better discrimination capability than single-view methods, our previous method, and eight baseline methods. When the dimension was reduced to 2, the model not only has effective prediction performance, but also is convenient for data visualization. Conclusions. Our proposed method by integrating radiomics features of primary tumor and LN can be helpful in predicting lymph node metastasis in patients of GC. It shows multi-view learning has great potential for guiding the prognosis and treatment decision-making in GC.

Machine Learning in Prediction of Bladder Cancer on Clinical Laboratory Data.

Bladder cancer has been increasing globally. Urinary cytology is considered a major screening method for bladder cancer, but it has poor sensitivity. This study aimed to utilize clinical laboratory data and machine learning methods to build predictive models of bladder cancer. A total of 1336 patients with cystitis, bladder cancer, kidney cancer, uterus cancer, and prostate cancer were enrolled in this study. Two-step feature selection combined with WEKA and forward selection was performed. Furthermore, five machine learning models, including decision tree, random forest, support vector machine, extreme gradient boosting (XGBoost), and light gradient boosting machine (GBM) were applied. Features, including calcium, alkaline phosphatase (ALP), albumin, urine ketone, urine occult blood, creatinine, alanine aminotransferase (ALT), and diabetes were selected. The lightGBM model obtained an accuracy of 84.8% to 86.9%, a sensitivity 84% to 87.8%, a specificity of 82.9% to 86.7%, and an area under the curve (AUC) of 0.88 to 0.92 in discriminating bladder cancer from cystitis and other cancers. Our study provides a demonstration of utilizing clinical laboratory data to predict bladder cancer.

Classification of magnetic ground states and prediction of magnetic moments of inorganic magnetic materials based on machine learning

Magnetic materials are important basic materials in the information age. Different magnetic ground states are the prerequisite for the wide application of magnetic materials, among which the ferromagnetic ground state is a key requirement for future high-performance magnetic materials. In this paper, machine learning is used to study the classification of ferromagnetic, antiferromagnetic, ferrimagnetic and paramagnetic ground states of inorganic magnetic materials and the prediction of magnetic moments of inorganic ferromagnetic materials. We obtain 98888 inorganic magnetic materials data from the Materials Project database, containing material ids, chemical formulae, CIF files, magnetic ground states and magnetic moments, and extract 582 elemental and structural features for the inorganic magnetic materials by using Matminer. We design a two-step feature selection method. In the first step, RFECV is used to evaluate material features one by one to remove redundant features without degrading the model accuracy. In the second step, we rank the material features to further refine and select the most important material features for the model, and 20 material features are selected for the classification of magnetic ground states and the prediction of magnetic moments, respectively. Among the selected material features, it is found that the electronegativity, the atomic own magnetic moment and the number of unfilled electrons in the atomic peripheral orbitals all make important contributions to the classification of magnetic ground states and the prediction of magnetic moments. We build a magnetic ground state classification model and a magnetic moment prediction model by using the random forest, and quantitatively evaluate the machine learning models by using the 10-fold cross-validation approach, and the results show that the constructed machine learning models has sufficient accuracy and generalization capability. In the test set, the magnetic ground state classification model has an accuracy of 85.23%, a precision of 85.18%, a recall of 85.04%, and an F1 score of 85.24%; the magnetic moment prediction model has a goodness-of-fit of 91.58% and an average absolute error of 0.098 μ&lt;sub&gt;B&lt;/sub&gt; per atom. This study provides a new method and choice for high-throughput classification and screening of magnetic ground states of inorganic magnetic materials and predicting the magnetic moment of ferromagnetic materials.

IAIPs: Identifying Anti-Inflammatory Peptides Using Random Forest.

Recently, several anti-inflammatory peptides (AIPs) have been found in the process of the inflammatory response, and these peptides have been used to treat some inflammatory and autoimmune diseases. Therefore, identifying AIPs accurately from a given amino acid sequences is critical for the discovery of novel and efficient anti-inflammatory peptide-based therapeutics and the acceleration of their application in therapy. In this paper, a random forest-based model called iAIPs for identifying AIPs is proposed. First, the original samples were encoded with three feature extraction methods, including g-gap dipeptide composition (GDC), dipeptide deviation from the expected mean (DDE), and amino acid composition (AAC). Second, the optimal feature subset is generated by a two-step feature selection method, in which the feature is ranked by the analysis of variance (ANOVA) method, and the optimal feature subset is generated by the incremental feature selection strategy. Finally, the optimal feature subset is inputted into the random forest classifier, and the identification model is constructed. Experiment results showed that iAIPs achieved an AUC value of 0.822 on an independent test dataset, which indicated that our proposed model has better performance than the existing methods. Furthermore, the extraction of features for peptide sequences provides the basis for evolutionary analysis. The study of peptide identification is helpful to understand the diversity of species and analyze the evolutionary history of species.

Quantitative detection of azodicarbonamide in wheat flour by near-infrared spectroscopy based on two-step feature selection

Excessive azodicarbonamide (ADA) in flour would do harm to the health of consumers. How to quickly and accurately detect the content of ADA in flour was of great significance. Based on the rapid, efficient and non-destructive advantages of near-infrared spectroscopy (NIRS) technology in material detection, the NIRS technology was used to quantitatively detect ADA in 101 wheat flour samples in this study. Firstly, after eliminating 1 abnormal sample with isolation forest (IF) method, the remaining 100 sample sets were divided into the training set and the prediction set using sample set partitioning based on joint x-y distances (SPXY) method. Then, the prediction performance of the three models under various spectral preprocessing methods and combined methods were compared. Among them, the performance of random forest (RF) model combined with second derivative (2D) was proved to be the best. Then, RReliefF and maximal information coefficient (MIC) were used for the first-step feature selection, respectively. On this basis, the second-step feature selection was performed based on the elastic net (EN). Among them, the performance of MIC ​+ ​EN was proved to be the best, and 40 effective features were selected. Finally, the 2D ​+ ​MIC ​+ ​EN ​+ ​RF model was established to predict the content of ADA in wheat flour. The coefficient of determination (R2), root mean square error of prediction (RMSEP) and relative percent difference (RPD) of the model on the prediction set reached 0.99814, 2.91345, and 23.54332, respectively. The results showed that NIRS technology could accurately detect the content of ADA in flour, and the two-step feature selection method could be effectively used for feature selection.

ITTCA-RF: a random forest predictor for tumor T cell antigens

BackgroundCancer is one of the most serious diseases threatening human health. Cancer immunotherapy represents the most promising treatment strategy due to its high efficacy and selectivity and lower side effects compared with traditional treatment. The identification of tumor T cell antigens is one of the most important tasks for antitumor vaccines development and molecular function investigation. Although several machine learning predictors have been developed to identify tumor T cell antigen, more accurate tumor T cell antigen identification by existing methodology is still challenging.MethodsIn this study, we used a non-redundant dataset of 592 tumor T cell antigens (positive samples) and 393 tumor T cell antigens (negative samples). Four types feature encoding methods have been studied to build an efficient predictor, including amino acid composition, global protein sequence descriptors and grouped amino acid and peptide composition. To improve the feature representation ability of the hybrid features, we further employed a two-step feature selection technique to search for the optimal feature subset. The final prediction model was constructed using random forest algorithm.ResultsFinally, the top 263 informative features were selected to train the random forest classifier for detecting tumor T cell antigen peptides. iTTCA-RF provides satisfactory performance, with balanced accuracy, specificity and sensitivity values of 83.71%, 78.73% and 88.69% over tenfold cross-validation as well as 73.14%, 62.67% and 83.61% over independent tests, respectively. The online prediction server was freely accessible at http://lab.malab.cn/~acy/iTTCA.ConclusionsWe have proven that the proposed predictor iTTCA-RF is superior to the other latest models, and will hopefully become an effective and useful tool for identifying tumor T cell antigens presented in the context of major histocompatibility complex class I.

Generalized methodology for radiomic feature selection and modeling in predicting clinical outcomes

Background. Quantitative radiomic features of medical images could provide clinical significance in assisting decision-making, but the existing feature selection and modeling methods are usually parameter-dependent. We aim to develop and validate a generalized radiomic method applicable to a variety of clinical outcomes. Methods and materials. A generalized methodology for radiomic feature selection and modeling (‘GRFM’ for short), including two-step feature selection and logistic regression, was proposed for studying clinical outcomes correlations. The two-step feature selection consists of Pearson correlation analysis followed by a sequential forward floating selection algorithm to identify robust feature subsets. We also applied an adaptive searching strategy to systematically determine globally optimal parameters, rather than relying on preset parameters. The area under the receiver operating characteristic curve (AUC) was used to evaluate the performance of three outcomes: lymph node metastasis of gastric cancer (GC), the five-year survival status of high-grade osteosarcoma (HOS), and the pathological grade of pancreatic neuroendocrine tumors (pNETs). Results. The optimal Pearson thresholds were 0.85, 0.80 and 0.75, and the optimal feature numbers were 11, 14 and 8 in GC, HOS and pNETs, respectively. The AUC values of the three predictive models combined with the corresponding parameters were 0.9017 versus 0.9026, 0.7652 versus 0.7113, and 0.8438 versus 0.8212 for the training and validation cohorts, showing promissing generality and classifier performance . Conclusion. The proposed method was helpful in predicting different clinical outcomes, and has potential application as a general and noninvasive prediction tool to guide clinical decision-making in various cancer sites.