10-fold Cross-validation Accuracy Research Articles

StackDPP: a stacking ensemble based DNA-binding protein prediction model.

DNA-binding proteins (DNA-BPs) are the proteins that bind and interact with DNA. DNA-BPs regulate and affect numerous biological processes, such as, transcription and DNA replication, repair, and organization of the chromosomal DNA. Very few proteins, however, are DNA-binding in nature. Therefore, it is necessary to develop an efficient predictor for identifying DNA-BPs. In this work, we have proposed new benchmark datasets for the DNA-binding protein prediction problem. We discovered several quality concerns with the widely used benchmark datasets, PDB1075 (for training) and PDB186 (for independent testing), which necessitated the preparation of new benchmark datasets. Our proposed datasets UNIPROT1424 and UNIPROT356 can be used for model training and independent testing respectively. We have retrained selected state-of-the-art DNA-BP predictors in the new dataset and reported their performance results. We also trained a novel predictor using the new benchmark dataset. We extracted features from various feature categories, then used a Random Forest classifier and Recursive Feature Elimination with Cross-validation (RFECV) to select the optimal set of 452 features. We then proposed a stacking ensemble architecture as our final prediction model. Named Stacking Ensemble Model for DNA-binding Protein Prediction, or StackDPP in short, our model achieved 0.92, 0.92 and 0.93 accuracy in 10-fold cross-validation, jackknife and independent testing respectively. StackDPP has performed very well in cross-validation testing and has outperformed all the state-of-the-art prediction models in independent testing. Its performance scores in cross-validation testing generalized very well in the independent test set. The source code of the model is publicly available at https://github.com/HasibAhmed1624/StackDPP . Therefore, we expect this generalized model can be adopted by researchers and practitioners to identify novel DNA-binding proteins.

Nomogram for predicting early complications after distal gastrectomy.

Reducing or preventing postoperative morbidity in patients with gastric cancer (GC) is particularly important in perioperative treatment plans. To identify risk factors for early postoperative complications of GC post-distal gastrectomy and to establish a nomogram prediction model. This retrospective study included 131 patients with GC who underwent distal gastrectomy at the Second Hospital of Shandong University between January 2019 and February 2023. The factors influencing the development of complications after distal gastrectomy in these patients were evaluated using univariate and multivariate logistic regression analysis. Based on the results obtained, a predictive nomogram was established. The nomogram was validated using internal and external (n = 45) datasets. Its sensitivity and specificity were established by receiver operating characteristic curve analysis. Decision curve (DCA) analysis was used to determine its clinical benefit and ten-fold overfitting was used to establish its accuracy and stability. Multivariate logistic regression analysis showed that hypertension, diabetes, history of abdominal surgery, and perioperative blood transfusion were independent predictors of postoperative complications of distal gastrectomy. The modeling and validation sets showed that the area under the curve was 0.843 [95% confidence interval (CI): 0.746-0.940] and 0.877 (95%CI: 0.719-1.000), the sensitivity was 0.762 and 0.778, respectively, and the specificity was 0.809 and 0.944, respectively, indicating that the model had good sensitivity and specificity. The C-indexes of the modeling and validation datasets were 0.843 (95%CI: 0.746-0.940) and 0.877 (95%CI: 0.719-1.000), respectively. The calibration curve (Hosmer Lemeshow test: χ2 = 7.33) showed that the model had good consistency. The results of the DCA analysis indicated that this model offered good clinical benefits. The accuracy of 10-fold cross-validation was 0.878, indicating that the model had good accuracy and stability. The nomogram prediction model based on independent risk factors related to postoperative complications of distal gastrectomy can facilitate perioperative intervention for high-risk populations and reduce the incidence of postoperative complications.

Improved Classification of Histopathological images using the feature fusion of Thepade sorted block truncation code and Niblack thresholding

Histopathology is the study of disease-affected tissues, and it is particularly helpful in diagnosis and figuring out how severe and rapidly a disease is spreading. It also demonstrates how to recognize a variety of human tissues and analyze the alterations brought on by sickness. Only through histopathological pictures can a specific collection of disease characteristics, such as lymphocytic infiltration of malignancy, be determined. The "gold standard" for diagnosing practically all cancer forms is a histopathological picture. Diagnosis and prognosis of cancer at an early stage are essential for treatment, which has become a requirement in cancer research. The importance and advantages of classification of cancer patients into more-risk or less-risk divisions have motivated many researchers to study and improve the application of machine learning (ML) methods. It would be interesting to explore the performance of multiple ML algorithms in classifying these histopathological images. Something crucial in this field of ML for differentiating images is feature extraction. Features are the distinctive identifiers of an image that provide a brief about it. Features are drawn out for discrimination between the images using a variety of handcrafted algorithms. This paper presents a fusion of features extracted with Thepade sorted block truncation code (TSBTC) and Niblack thresholding algorithm for the classification of histopathological images. The experimental validation is done using 960 images present in the Kimiapath-960 dataset of histopathological images with the help of performance metrics like sensitivity, specificity and accuracy. Better performance is observed by an ensemble of TSBTC N-ary and Niblack's thresholding features as 97.92% of accuracy in 10-fold cross-validation.

Using Pearson Correlation and Mutual Information (PC-MI) to Select Features for Accurate Breast Cancer Diagnosis Based on a Soft Voting Classifier

Breast cancer is one of the most critical diseases suffered by many people around the world, making it the most common medical risk they will face. This disease is considered the leading cause of death around the world, and early detection is difficult. In the field of healthcare, where early diagnosis based on machine learning (ML) helps save patients’ lives from the risks of diseases, better-performing diagnostic procedures are crucial. ML models have been used to improve the effectiveness of early diagnosis. In this paper, we proposed a new feature selection method that combines two filter methods, Pearson correlation and mutual information (PC-MI), to analyse the correlation amongst features and then select important features before passing them to a classification model. Our method is capable of early breast cancer prediction and depends on a soft voting classifier that combines a certain set of ML models (decision tree, logistic regression, and support vector machine) to produce one model that carries the strengths of the models that have been combined, yielding the best prediction accuracy. Our work is evaluated by using the Wisconsin Diagnostic Breast Cancer datasets. The proposed methodology outperforms previous work, achieving 99.3% accuracy, an F1 score of 0.9922, a recall of 0.9846, a precision of 1, and an AUC of 0.9923. Furthermore, the accuracy of 10-fold cross-validation is 98.2%.

Prediction of wart treatment response using a hybrid GA-ensemble learning approach

Warts are a common type of skin disease that is caused mainly by human papillomavirus (HPV). Several approaches are used to treat warts, which include the popular cryotherapy and immunotherapy methods. Various machine learning-based approaches have been introduced to help physicians select a proper treatment method for each patient. The main drawback of those approaches is the limitation of samples used for each of those models (90 samples for cryotherapy model and 90 samples for immunotherapy model). This study develops a reliable wart treatment prediction system using a hybrid genetic algorithm (GA)-ensemble learning approach with a larger dataset. The immunotherapy and cryotherapy datasets used in previous studies are separately balanced and then combined into one wart dataset, and the treatment method is modeled as one of the input features. GA is combined with learning algorithms, namely backpropagation neural networks (BPNN), support vector machine (SVM), classification and regression tree (CART), and K-nearest neighbors (KNN), to determine the optimal features for the prediction model. The four base classifiers, GA-BPNN, GA-SVM, GA-CART, and GA-KNN, are ensembled using bagging, boosting, and stacking techniques. The results show that ensemble models yield better classification results on the combined balanced wart dataset than individual classifiers. In particular, the stacking model with GA-SVM, GA-CART, and GA-KNN as layer-1 classifiers performs the best (accuracies of 100 % and 98.3 % on 10-fold cross validation and testing samples, respectively). Furthermore, the stacking model outperforms the model introduced in the literature on the same combined dataset with 180 samples (97.2 % average accuracy of 10-fold cross validation compared to 89.3 %). The proposed prediction system can assist physicians to reliably select appropriate treatment methods for patients who have warts with high accuracy, which in turn will have a positive impact on clinical resources.

Detection of autism spectrum disorder using graph representation learning algorithms and deep neural network, based on fMRI signals.

Can we apply graph representation learning algorithms to identify autism spectrum disorder (ASD) patients within a large brain imaging dataset? ASD is mainly identified by brain functional connectivity patterns. Attempts to unveil the common neural patterns emerged in ASD are the essence of ASD classification. We claim that graph representation learning methods can appropriately extract the connectivity patterns of the brain, in such a way that the method can be generalized to every recording condition, and phenotypical information of subjects. These methods can capture the whole structure of the brain, both local and global properties. The investigation is done for the worldwide brain imaging multi-site database known as ABIDE I and II (Autism Brain Imaging Data Exchange). Among different graph representation techniques, we used AWE, Node2vec, Struct2vec, multi node2vec, and Graph2Img. The best approach was Graph2Img, in which after extracting the feature vectors representative of the brain nodes, the PCA algorithm is applied to the matrix of feature vectors. The classifier adapted to the features embedded in graphs is an LeNet deep neural network. Although we could not outperform the previous accuracy of 10-fold cross-validation in the identification of ASD versus control patients in this dataset, for leave-one-site-out cross-validation, we could obtain better results (our accuracy: 80%). The result is that graph embedding methods can prepare the connectivity matrix more suitable for applying to a deep network.

Environmental toxicity risk evaluation of nitroaromatic compounds: Machine learning driven binary/multiple classification and design of safe alternatives.

Nitroaromatic compounds (NACs) represent a significant source of organic pollutants in the environment. In this study, a well-rounded dataset containing 371 NACs with rat oral median lethal doses (LD50s) was developed. Based on the dataset, binary and multiple classification models were established. Seven machine learning algorithms were used to establish the prediction models in combination with six fingerprints. In the binary classification models, the overall predictive accuracy of 10-fold cross-validation for training set in the top ten models ranged from 0.823 to 0.874. In the multiple classification models, the combination of graph fingerprint and random forest (Graph-RF) yielded the best predictive effects with AUC values of 0.929 and 0.956 for the training set and the test set, respectively. Model prediction performance was further evaluated using the true external set comprising 1366 NACs, including 96.6% belonging to the applicability domain. Further, we determined the structural features influencing the acute oral toxicity based on information gain and substructure frequency analysis. Finally, we identified highly toxic compounds based on the structural alerts and successfully transformed a representative highly toxic compound into low-toxic alternatives via structural modification. Overall, the models constructed facilitate environmental risk assessment and the design of green and safe chemicals.

Prediction of Hypertension Outcomes Based on Gain Sequence Forward Tabu Search Feature Selection and XGBoost.

For patients with hypertension, serious complications, such as myocardial infarction, a common cause of heart failure, occurs in the late stage of hypertension. Hypertension outcomes can lead to complications, including death. Hypertension outcomes threaten patients’ lives and need to be predicted. In our research, we reviewed the hypertension medical data from a tertiary-grade A class hospital in Beijing, and established a hypertension outcome prediction model with the machine learning theory. We first proposed a gain sequence forward tabu search feature selection (GSFTS-FS) method, which can search the optimal combination of medical variables that affect hypertension outcomes. Based on this, the XGBoost algorithm established a prediction model because of its good stability. We verified the proposed method by comparing other commonly used models in similar works. The proposed GSFTS-FS improved the performance by about 10%. The proposed prediction method has the best performance and its AUC value, accuracy, F1 value, and recall of 10-fold cross-validation were 0.96. 0.95, 0.88, and 0.82, respectively. It also performed well on test datasets with 0.92, 0.94, 0.87, and 0.80 for AUC, accuracy, F1, and recall, respectively. Therefore, the XGBoost with GSFTS-FS can accurately and effectively predict the occurrence of outcomes for patients with hypertension, and can provide guidance for doctors in clinical diagnoses and medical decision-making.

A New Subject-Specific Discriminative and Multi-Scale Filter Bank Tangent Space Mapping Method for Recognition of Multiclass Motor Imagery.

Objective: Tangent Space Mapping (TSM) using the geometric structure of the covariance matrices is an effective method to recognize multiclass motor imagery (MI). Compared with the traditional CSP method, the Riemann geometric method based on TSM takes into account the nonlinear information contained in the covariance matrix, and can extract more abundant and effective features. Moreover, the method is an unsupervised operation, which can reduce the time of feature extraction. However, EEG features induced by MI mental activities of different subjects are not the same, so selection of subject-specific discriminative EEG frequency components play a vital role in the recognition of multiclass MI. In order to solve the problem, a discriminative and multi-scale filter bank tangent space mapping (DMFBTSM) algorithm is proposed in this article to design the subject-specific Filter Bank (FB) so as to effectively recognize multiclass MI tasks.Methods: On the 4-class BCI competition IV-2a dataset, first, a non-parametric method of multivariate analysis of variance (MANOVA) based on the sum of squared distances is used to select discriminative frequency bands for a subject; next, a multi-scale FB is generated according to the range of these frequency bands, and then decompose multi-channel EEG of the subject into multiple sub-bands combined with several time windows. Then TSM algorithm is used to estimate Riemannian tangent space features in each sub-band and finally a liner Support Vector Machines (SVM) is used for classification.Main Results: The analysis results show that the proposed discriminative FB enhances the multi-scale TSM algorithm, improves the classification accuracy and reduces the execution time during training and testing. On the 4-class BCI competition IV-2a dataset, the average session to session classification accuracy of nine subjects reached 77.33 ± 12.3%. When the training time and the test time are similar, the average classification accuracy is 2.56% higher than the latest TSM method based on multi-scale filter bank analysis technology. When the classification accuracy is similar, the training speed is increased by more than three times, and the test speed is increased two times more. Compared with Supervised Fisher Geodesic Minimum Distance to the Mean (Supervised FGMDRM), another new variant based on Riemann geometry classifier, the average accuracy is 3.36% higher, we also compared with the latest Deep Learning method, and the average accuracy of 10-fold cross validation improved by 2.58%.Conclusion: Research shows that the proposed DMFBTSM algorithm can improve the classification accuracy of MI tasks.Significance: Compared with the MFBTSM algorithm, the algorithm proposed in this article is expected to select frequency bands with good separability for specific subject to improve the classification accuracy of multiclass MI tasks and reduce the feature dimension to reduce training time and testing time.

The functional brain network based on the combination of shortest path tree and its application in fatigue driving state recognition and analysis of the neural mechanism of fatigue driving

Aimed at studying the method of constructing a functional brain network (FBN) that can effectively recognize the state of fatigue driving based on electroencephalogram (EEG), and analyzing which regions of the brain (electrode) are closely related to the occurrence of fatigue driving. A method based on the combination of shortest path tree (CSPT) for constructing a functional brain network (denoted as CSP-FBN) is proposed, which is applied to fatigue driving state recognition and neural mechanism analysis of fatigue driving. Through the comparison experiment of the classification accuracy in the same frequency band (beta band), the results show that the functional brain network constructed by the combined shortest path tree in fatigue state recognition is better than the functional brain network constructed by other methods, the accuracy of 10-fold cross validation reaches 99.17%. At the same time, we also find that Fz, F4, Fc3, Fcz, Fc4, C3, Cz4, Cp3, Cpz, Cp4, P3, Pz and P4 are important electrodes for fatigue driving state recognition, which reflects that the right central region and the central parietal region of the brain have a close relationship with the occurrence of fatigue driving.

Overlap matrix completion for predicting drug-associated indications.

Identification of potential drug-associated indications is critical for either approved or novel drugs in drug repositioning. Current computational methods based on drug similarity and disease similarity have been developed to predict drug-disease associations. When more reliable drug- or disease-related information becomes available and is integrated, the prediction precision can be continuously improved. However, it is a challenging problem to effectively incorporate multiple types of prior information, representing different characteristics of drugs and diseases, to identify promising drug-disease associations. In this study, we propose an overlap matrix completion (OMC) for bilayer networks (OMC2) and tri-layer networks (OMC3) to predict potential drug-associated indications, respectively. OMC is able to efficiently exploit the underlying low-rank structures of the drug-disease association matrices. In OMC2, first of all, we construct one bilayer network from drug-side aspect and one from disease-side aspect, and then obtain their corresponding block adjacency matrices. We then propose the OMC2 algorithm to fill out the values of the missing entries in these two adjacency matrices, and predict the scores of unknown drug-disease pairs. Moreover, we further extend OMC2 to OMC3 to handle tri-layer networks. Computational experiments on various datasets indicate that our OMC methods can effectively predict the potential drug-disease associations. Compared with the other state-of-the-art approaches, our methods yield higher prediction accuracy in 10-fold cross-validation and de novo experiments. In addition, case studies also confirm the effectiveness of our methods in identifying promising indications for existing drugs in practical applications.

Celiac disease diagnosis from videocapsule endoscopy images with residual learning and deep feature extraction

Background and ObjectiveVideocapsule endoscopy (VCE) is a relatively new technique for evaluating the presence of villous atrophy in celiac disease patients. The diagnostic analysis of video frames is currently time-consuming and tedious. Recently, computer-aided diagnosis (CAD) systems have become an attractive research area for diagnosing celiac disease. However, the images captured from VCE are susceptible to alterations in light illumination, rotation direction, and intestinal secretions. Moreover, textural features of the mucosal villi obtained by VCE are difficult to characterize and extract. This work aims to find a novel deep learning feature learning module to assist in the diagnosis of celiac disease. MethodsIn this manuscript, we propose a novel deep learning recalibration module which shows significant gain in diagnosing celiac disease. In this recalibration module, the block-wise recalibration component is newly employed to capture the most salient feature in the local channel feature map. This learning module was embedded into ResNet50, Inception-v3 to diagnose celiac disease using a 10-time 10-fold cross-validation based upon analysis of VCE images. In addition, we employed model weights to extract feature points from training and test samples before the last fully connected layer, and then input to a support vector machine (SVM), k-nearest neighbor (KNN), and linear discriminant analysis (LDA) for differentiating celiac disease images from heathy controls. ResultsOverall, the accuracy, sensitivity and specificity of the 10-time 10-fold cross-validation were 95.94%, 97.20% and 95.63%, respectively. ConclusionsA novel deep learning recalibration module, with global response and local salient factors is proposed, and it has a high potential for utilizing deep learning networks to diagnose celiac disease using VCE images.

Image quantification technology of the heterochromatin and euchromatin region for differential diagnosis in the lobular endocervical glandular hyperplasia.

Lobular endocervical glandular hyperplasia (LEGH) was first described by Nucci et al. in 1999 and is believed to be a precancerous lesion of minimal deviation adenocarcinoma and gastric-type adenocarcinoma in the uterine cervix. LEGH lesions do not always exhibit apparent cellular and structural atypia, so are difficult to distinguish from normal endocervical cells (EC cells) with cytological examination. Therefore, we often struggle to make a definite diagnosis of LEGH. We used microscopy images of cytological specimens that were diagnosed as EC cells and LEGH cells. Signal intensity in whole nuclear area and in heterochromatin and euchromatin regions, euchromatin area ratio, and nuclear morphological features were quantified in each cell nucleus of the cases. Statistical analyses were conducted to determine statistical significance. Finally, we performed linear support vector machine (LSVM) modeling as a discriminant analysis using the quantified features. Signal intensity in whole nuclear area, and heterochromatin and euchromatin regions of EC cell nuclei were higher than that of the LEGH cell nuclei. Morphologically, EC cell nuclei were larger than LEGH cell nuclei, and nuclei of LEGH cells had irregular nuclear respectively membrane structure and an elongated shape. The LSVM accuracy of 10-fold cross validation and leave-one-case-out cross-validation (LOCOCV) using all measured features were 84.7% to 89.3% and 78.6% to 86.0%, respectively. The LVSM analysis using features extracted from signal intensity and morphological analysis was useful for discrimination of EC cells vs LEGH cells. We therefore believe that this image analysis method could be used for early detection of LEGH.

Machine-learning-assisted prediction of surgical outcomes in patients undergoing gastrectomy.

ObjectivePostoperative complications adversely affected the prognosis in patients with gastric cancer. This study intends to investigate the feasibility of using machine-learning model to predict surgical outcomes in patients undergoing gastrectomy.MethodsIn this study, cancer patients who underwent gastrectomy at Shanghai Rui Jin Hospital in 2017 were randomly assigned to a development or validation cohort in a 9:1 ratio. A support vector classification (SVC) model to predict surgical outcomes in patients undergoing gastrectomy was developed and further validated.ResultsA total of 321 patients with 32 features were collected. The positive and negative outcomes of postoperative complication after gastrectomy appeared in 100 (31.2%) and 221 (68.8%) patients, respectively. The SVC model was constructed to predict surgical outcomes in patients undergoing gastrectomy. The accuracy of 10-fold cross validation and external verification was 78.17% and 78.12%, respectively. Further, an online web server has been developed to share the SVC model for machine-learning-assisted prediction of surgical outcomes in patients undergoing gastrectomy in the future procedures, which is accessible at the web address: http://47.100.47.97:5005/r_model_prediction.ConclusionsThe SVC model was a useful predictor for measuring the risk of postoperative complications after gastrectomy, which may help stratify patients with different overall status for choice of surgical procedure or other treatments. It can be expected that machine-learning models in cancer informatics research are possibly shareable and accessible via web address all over the world.

A novel seizure diagnostic model based on kernel density estimation and least squares support vector machine

Abstract The automated system can be an effective tool for assisting neurologists in seizure detection. However, most of the existing methods are failed to trade off the effectivity and computation cost, which is not appropriate for on-line application. In this research, we propose a novel method for dealing with 3-class electroencephalogram (EEG) problem, based upon kernel density estimation (KDE) and least squares support vector machine (LS-SVM). The filtered EEG is decomposed into several sub-bands by wavelet packet transform (WPT), then KDE is explored to calculate the corresponding probability density. Five parameters are employed for EEG representation: the maximum (Max), the skewness (Ske), the kurtosis (Kur), the energy (En), and the central moment (CM). And significant features selected by Analysis of Variance (ANOVA) are fed to LS-SVM for pattern recognition. Furthermore, eight types of wavelet bases and four well-known functions are considered for feature extraction. Experimental results show that our approach has achieved satisfactory and comparable results for all validation methods when configured with coiflet of order 1 and uniform kernel. The highest accuracy of 10-fold cross-validation and standard 50-50 methodology is 99.40% and 99.60% with 27 and 26 features, respectively. As compared to previous literature, our proposed scheme is more suitable for diagnosis of epilepsy with higher accuracy and less number of feature that can be extracted with less computational cost. Overall, the advantages of high accuracy, easy implementation and low computational consumption have made this technique a suitable candidate for extensive clinical deployment.

Abstract 429: Establishment of the classifier for a response to vascular endothelial growth factor (VEGF)-tyrosine kinase inhibitor (TKI) in metastatic renal cell carcinoma

Abstract Vascular endothelial growth factor (VEGF)-targeted therapy has been demonstrated to improve the outcome of metastatic renal cell carcinoma (mRCC) patients. However, validated predictors that predict response and prognosis to VEGF-tyrosine kinase inhibitors (VEGF-TKIs) remain to be elucidated. We aimed to define a classifier for VEGF-TKI response in mRCC patients. Among 101 mRCC patients treated with VEGF-TKIs, 73 patients were responder defined as patients showing complete or partial response, or ≥24 weeks stable disease to VEGF-TKI; and 28 patients were nonresponder defined as <24 weeks stable or progressive disease. Clinical and laboratory data were obtained from the medical records. Histologic features of all tumors were reviewed. Twenty-one protein expressions such as VEGF, pAKT, PD-L1, PD-L2, HIF-1α and HIF-2α were evaluated on immunohistochemical staining. Mutation of various cancer-related genes was investigated on OncoMap ver. 4.4 core. Microarray and quantitative real-time reverse transcriptase-polymerase chain reaction (qRT-PCR) were also performed to compare the expression patterns of miRNAs between responders and nonresponders in tissues samples. Patient's age, time from diagnosis to TKI initiation and thrombodytosis were different between responder and nonresponder (all, p <0.05). Histologically, tumor size, T stage, ISUP grade, sarcomatous change, necrosis and lymph node metastasis of responder were significantly lower or smaller than nonresponder (all, p <0.05). pAKT, PD-L1, PD-L2, FGFR2, pS6 and PDGFRβ showed more expression in nonresponder than in responder; on the other hands, HIF-1α, IL-8 and CA9 showed more expression in responder than in nonresponder (all, p <0.05). Expression of miR-421 was higher in responder than nonresponder (p = 0.008). A classifier to VEGF-TKI response was developed with various clinicopathological features, protein expression and miRNA expression. Tumor size, T stage, ISUP grade, necrosis, sarcomatoid change, pAKT, PD-L1, CA9, pS6, HIF-1α and miR-421 were selected by chi-square test with p <0.01. Using 10-fold cross validation (CV) by support vector machine, 3 features, i.e., necrosis, sarcomatoid change and HIF-1α, were finally selected as features for classifier and accuracy of 10-fold CV was 0.87. When the classifier was checked with all patients, apparent accuracy was 0.875 (95% CI, 0.782-0.938). In addition, the classifier could be presented by a simple decision tree for clinical use. In conclusion, we built a VEGF-TKI response classifier by comprehensive inclusion of clinical, laboratorical, histological and immunihistochemical features, mutation of cancer-related genes and miRNA expression using machine learning method and it may be helpful to receive a proper treatment in mRCC patients. Citation Format: Heounjeong Go, Mun Jung Kang, Pil-Jong Kim, Ja-Min Park, Jae-Lyun Lee, Ji Young Park, Yong Mee Cho. Establishment of the classifier for a response to vascular endothelial growth factor (VEGF)-tyrosine kinase inhibitor (TKI) in metastatic renal cell carcinoma. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 429.

Interactions with Fibroblasts Are Distinct in Basal-Like and Luminal Breast Cancers

Basal-like breast cancers have several well-characterized distinguishing molecular features, but most of these are features of the cancer cells themselves. The unique stromal-epithelial interactions, and more generally, microenvironmental features of basal-like breast cancers have not been well characterized. To identify characteristic microenvironment features of basal-like breast cancer, we performed cocultures of several basal-like breast cancer cell lines with fibroblasts and compared these with cocultures of luminal breast cancer cell lines with fibroblasts. Interactions between basal-like cancer cells and fibroblasts induced expression of numerous interleukins and chemokines, including IL-6, IL-8, CXCL1, CXCL3, and TGFβ. Under the influence of fibroblasts, basal-like breast cancer cell lines also showed increased migration in vitro. Migration was less pronounced for luminal lines; but, these lines were more likely to have altered proliferation. These differences were relevant to tumor biology in vivo, as the gene set that distinguished luminal and basal-like stromal interactions in coculture also distinguishes basal-like from luminal tumors with 98% accuracy in 10-fold cross-validation and 100% accuracy in an independent test set. However, comparisons between cocultures where cells were in direct contact and cocultures where interaction was solely through soluble factors suggest that there is an important impact of direct cell-to-cell contact. The phenotypes and gene expression changes invoked by cancer cell interactions with fibroblasts support the microenvironment and cell-cell interactions as intrinsic features of breast cancer subtypes.