Gene Expression Profile Classification Research Articles

Validation of the Prognostic Usefulness of the Gene Expression Profiling Test in Patients with Uveal Melanoma

To validate the prognostic usefulness of gene expression profile (GEP) testing in patients with uveal melanoma. To determine whether combining tumor size with the GEP classification provides additional prognostic value. Retrospective analysis. Patients with a diagnosis of choroidal melanoma examined at Yale New Haven Hospital; University of California, San Diego; and Memorial Sloan Kettering Cancer Center. Patients' demographic and clinical data and tumor characteristics were collected. Univariate and multivariate Cox hazard regression analysis were used to assess the association between tumor characteristics and GEP classification with metastasis as an outcome. Metastasis-free survival (MFS). Of the 337 individuals included in the study, 87 demonstrated metastases. The mean follow-up time was 37.2 (standard deviation [SD], 40.2) months for patients with metastases and 55.0 (SD, 49.3) months for those without metastases. Tumors of larger thickness and GEP class 2 (vs. class 1) were associated significantly with increased risk of metastasis. Tumor thickness showed better prognostic usefulness than GEP classification (Wald statistic, 40.7 and 24.2, respectively). Class 2 tumors with a thickness of 7.0 mm or more were associated with increased risk of metastasis than tumors with a thickness of < 7.0 mm (hazard ratio [HR], 3.23; 95% confidence interval [CI], 1.61-6.51), whereas class 1 tumors with a thickness of 9.0 mm or more were associated with increased risk of metastasis than tumors with a thickness of < 9.0 mm (HR, 2.07; 95% CI, 0.86-4.99). No difference in MFS was found between patients with class 1A tumors compared with those with class 1B tumors (P= 0.8). Patients with class 2 tumors showed an observed 5-year MFS of 47.5% (95% CI, 36.0%-62.8%). Tumor size was the most significant predictor of metastasis and provided additional prognostic value independent of GEP classification. In addition, rates of metastasis for class 2 tumors were lower than estimates reported by Castle Bioscience, and no difference in rates of metastasis were found between class 1A and 1B tumors. This indicates that tumor size should be accounted for when relying on GEP for prognostication and that patients with GEP class 1A or 1B tumors may benefit from the same metastatic surveillance protocols. The author(s) have no proprietary or commercial interest in any materials discussed in this article.

Integrative Analysis of the Genomic and Transcriptomic Landscape of Relapsed/Refractory Multiple Myeloma Patients Treated With Venetoclax in Combination With Bortezomib and Dexamethasone: Biomarker Analyses From the Phase 3 BELLINI Study

Integrative Analysis of the Genomic and Transcriptomic Landscape of Relapsed/Refractory Multiple Myeloma Patients Treated With Venetoclax in Combination With Bortezomib and Dexamethasone: Biomarker Analyses From the Phase 3 BELLINI Study

Discriminant Projection Shared Dictionary Learning for Classification of Tumors Using Gene Expression Data.

With a variety of tumor subtypes, personalized treatments need to identify the subtype of a tumor as accurately as possible. The development of DNA microarrays provides an opportunity to predict tumor classification. One strategy is to use gene expression profiling to extend current biological insights into the disease. However, overfitting problems exist in most machine learning methods when classifying tumor gene expression profile data characterized by high dimensional, small samples and nonlinearities. As a new machine learning methods, dictionary learning has become a more effective algorithm for gene expression profile classification. Here, a new method called discriminant projection shared dictionary learning (DPSDL) is proposed for classifying tumor subtypes using LINCS gene expression profile data. The method trains a shared dictionary, embeds Fisher discriminant criteria to obtain a class-specific sub-dictionary and coding coefficients. At the same time, a projection matrix is trained to widen the distance between different classes of samples. Experimental results show that our method performs better classification based on gene expression profile than the other dictionary learning methods and machine learning methods.

A Comparative Study of Different Machine Learning Algorithms on Gene Expression Profile Classification

Many machine learning algorithms have been used to classify gene expression profiles in recently years. In order to furtherly study the performances of different machine learning algorithms on gene expression profiles classification, this paper compare the classification accuracy, run time and stability the of different machine learning algorithms including SVM, Decision tree, PNN, k-Nearest Neighbor, Bayesian and ELM on benchmark gene expression profile datasets. It provides a basis for scientific use of machine learning algorithms on gene expression profiles classification.

Association between traditional clinical high-risk features and gene expression profile classification in uveal melanoma.

To evaluate the association between traditional clinical high-risk features of uveal melanoma patients and gene expression profile (GEP). This was a retrospective, single-center, case series of patients with uveal melanoma. Eighty-three patients met inclusion criteria for the study. Patients were examined for the following clinical risk factors: drusen/retinal pigment epithelium (RPE) changes, vascularity on B-scan, internal reflectivity on A-scan, subretinal fluid (SRF), orange pigment, apical tumor height/thickness, and largest basal dimensions (LBD). A novel point system was created to grade the high-risk clinical features of each tumor. Further analyses were performed to assess the degree of association between GEP and each individual risk factor, total clinical risk score, vascularity, internal reflectivity, American Joint Committee on Cancer (AJCC) tumor stage classification, apical tumor height/thickness, and LBD. Of the 83 total patients, 41 were classified as GEP class 1A, 17 as class 1B, and 25 as class 2. The presence of orange pigment, SRF, low internal reflectivity and vascularity on ultrasound, and apical tumor height/thickness ≥ 2mm were not statistically significantly associated with GEP class. Lack of drusen/RPE changes demonstrated a trend toward statistical association with GEP class 2 compared to class 1A/1B. LBD and advancing AJCC stage was statistically associated with higher GEP class. In this cohort, AJCC stage classification and LBD were the only clinical features statistically associated with GEP class. Clinicians should use caution when inferring the growth potential of melanocytic lesions solely from traditional funduscopic and ultrasonographic risk factors without GEP data.

Association between Tumor Regression Rate and Gene Expression Profile after Iodine 125 Plaque Radiotherapy for Uveal Melanoma

Gene expression profile (GEP) testing segregates uveal melanoma (UM) into 2 main prognostic classes. It is unknown if a greater tumor regression response after iodine 125 (I125) brachytherapy correlates with class 2 GEP status. The purpose of this study was to determine whether there is a significant relationship between the rate of UM height regression and GEP classification testing after I125 plaque brachytherapy. Multicenter, retrospective cohort study. Adult UM patients treated with I125 plaque brachytherapy who had concurrent tumor biopsy at the time of surgery with a GEP test result from January 1, 2010 through June 30,2014. Baseline clinical data and GEP class assignments were obtained. The ultrasonographic tumor height was recorded at baseline and at 3, 6, 9, and 12 months and at the most recent final follow-up visits. Subanalysis of paired cases based on pretreatment ultrasound height was performed. Statistical analysis was performed using Wilcoxon rank-sum tests, the Fisher exact test, and Kaplan-Meier analysis. Percentage change in tumor height from baseline. A total of 353 patients were included in the study. Median follow-up was 2.1 years (range, 0.5-5.3 years). Gene expression profile status was class 1 in 247 tumors (70%) and class 2 in 106 tumors (30%). Increased patient age, larger tumor dimensions, and greater tumor thickness were associated with class 2 GEP status (P= 0.006, P < 0.001, and P < 0.001, respectively). The percentage reduction in tumor height from baseline was significantly greater in class 1 than class 2 tumors at 3 months (17.5% vs. 11.8%; P= 0.007) and 6 months (26.8% vs. 17.1%; P= 0.007), respectively, but there was no significant difference in reduction between class 1 and 2 tumors at 9 months (P= 0.26) and 12 months (P= 0.57) after treatment. Class 1A and 1B tumors showed similar reduction compared with class 2 tumors (P < 0.05). Class 1 UM tumors tend to regress more rapidly than class 2 tumors in the first 6 months after plaque radiotherapy. Class 1A and 1B tumors regress at similar rates after plaque radiotherapy.

Driver Mutations in Uveal Melanoma: Associations With Gene Expression Profile and Patient Outcomes.

Frequent mutations have been described in the following 5 genes in uveal melanoma (UM): BAP1, EIF1AX, GNA11, GNAQ, and SF3B1. Understanding the prognostic significance of these mutations could facilitate their use in precision medicine. To determine the associations between driver mutations, gene expression profile (GEP) classification, clinicopathologic features, and patient outcomes in UM. Retrospective study of patients with UM treated by enucleation by a single ocular oncologist between November 1, 1998, and July 31, 2014. Clinicopathologic features, patient outcomes, GEP classification (class 1 or class 2), and mutation status were recorded. The study cohort comprised 81 participants. Their mean age was 61.5 years, and 37% (30 of 81) were female. The GEP classification was class 1 in 35 of 81 (43%), class 2 in 42 of 81 (52%), and unknown in 4 of 81 (5%). BAP1 mutations were identified in 29 of 64 (45%), GNAQ mutations in 36 of 81 (44%), GNA11 mutations in 36 of 81 (44%), SF3B1 mutations in 19 of 81 (24%), and EIF1AX mutations in 14 of 81 (17%). Sixteen of the mutations in BAP1 and 6 of the mutations in EIF1AX were previously unreported in UM. GNAQ and GNA11 mutations were mutually exclusive. BAP1, SF3B1, and EIF1AX mutations were almost mutually exclusive with each other. Using multiple regression analysis, BAP1 mutations were associated with class 2 GEP and older patient. EIF1AX mutations were associated with class 1 GEP and the absence of ciliary body involvement. SF3B1 mutations were associated with younger patient age. GNAQ mutations were associated with the absence of ciliary body involvement and greater largest basal diameter. GNA11 mutations were not associated with any of the analyzed features. Using Cox proportional hazards modeling, class 2 GEP was the prognostic factor most strongly associated with metastasis (relative risk, 9.4; 95% CI, 3.1-28.5) and melanoma-specific mortality (relative risk, 15.7; 95% CI, 3.6-69.1) (P < .001 for both). After excluding GEP class, the presence of BAP1 mutations was the factor most strongly associated with metastasis (relative risk, 10.6; 95% CI, 3.4-33.5) and melanoma-specific mortality (relative risk, 9.0; 95% CI, 2.8-29.2) (P < .001 for both). BAP1, SF3B1, and EIF1AX mutations occur during UM tumor progression in an almost mutually exclusive manner and are associated with different levels of metastatic risk. These mutations may have value as prognostic markers in UM.

Frequency and Implications of Discordant Gene Expression Profile Class in Posterior Uveal Melanomas Sampled by Fine Needle Aspiration Biopsy

To determine the frequency of discordant gene expression profile (GEP) classification of posterior uveal melanomas sampled at 2 tumor sites by fine-needle aspiration biopsy (FNAB). Prospective single-institution longitudinal study performed in conjunction with a multicenter validation study of the prognostic value of GEP class of posterior uveal melanoma cells for metastasis and metastatic death. FNAB aspirates of 80 clinically diagnosed primary choroidal and ciliochoroidal melanomas were obtained from 2 tumor sites prior to or at the time of initial ocular tumor treatment and submitted for independent GEP testing and classification. Frequency of discordant GEP classification of these specimens was determined. Using the support vector machine learning algorithm favored by the developer of the GEP test employed in this study, 9 of the 80 cases (11.3% [95% confidence interval: 9.0%-13.6%]) were clearly discordant. If cases with a failed classification at 1 site or a low confidence class assignment by the support vector machine algorithm at 1 or both sites are also regarded as discordant, then this frequency rises to 13 of the 80 cases (16.3% [95% confidence interval: 13.0%-19.6%]). Sampling of a clinically diagnosed posterior uveal melanoma at a single site for prognostic GEP testing is associated with a substantial probability of misclassification. Two-site sampling of such tumors with independent GEP testing of each specimen may be advisable to lessen the probability of underestimating an individual patient's prognostic risk of metastasis and metastatic death.

A simulation to analyze feature selection methods utilizing gene ontology for gene expression classification

Gene expression profile classification is a pivotal research domain assisting in the transformation from traditional to personalized medicine. A major challenge associated with gene expression data classification is the small number of samples relative to the large number of genes. To address this problem, researchers have devised various feature selection algorithms to reduce the number of genes. Recent studies have been experimenting with the use of semantic similarity between genes in Gene Ontology (GO) as a method to improve feature selection. While there are few studies that discuss how to use GO for feature selection, there is no simulation study that addresses when to use GO-based feature selection. To investigate this, we developed a novel simulation, which generates binary class datasets, where the differentially expressed genes between two classes have some underlying relationship in GO. This allows us to investigate the effects of various factors such as the relative connectedness of the underlying genes in GO, the mean magnitude of separation between differentially expressed genes denoted by δ, and the number of training samples. Our simulation results suggest that the connectedness in GO of the differentially expressed genes for a biological condition is the primary factor for determining the efficacy of GO-based feature selection. In particular, as the connectedness of differentially expressed genes increases, the classification accuracy improvement increases. To quantify this notion of connectedness, we defined a measure called Biological Condition Annotation Level BCAL(G), where G is a graph of differentially expressed genes. Our main conclusions with respect to GO-based feature selection are the following: (1) it increases classification accuracy when BCAL(G)⩾0.696; (2) it decreases classification accuracy when BCAL(G)⩽0.389; (3) it provides marginal accuracy improvement when 0.389<BCAL(G)<0.696 and δ<1; (4) as the number of genes in a biological condition increases beyond 50 and δ⩾0.7, the improvement from GO-based feature selection decreases; and (5) we recommend not using GO-based feature selection when a biological condition has less than ten genes. Our results are derived from datasets preprocessed using RMA (Robust Multi-array Average), cases where δ is between 0.3 and 2.5, and training sample sizes between 20 and 200, therefore our conclusions are limited to these specifications. Overall, this simulation is innovative and addresses the question of when SoFoCles-style feature selection should be used for classification instead of statistical-based ranking measures.

Application of committee kNN classifiers for gene expression profile classification

In this study, we develop a two-class classification system based on a committee of k-Nearest Neighbour (kNN) classifiers. The system includes a sequence of simple data preprocessing steps. Each committee consists of 5 kNN classifiers of different architectures. Each classifier on the committee takes in a different set of features. The classification system is then applied to a set of microarray gene expression profiles from leukaemia patients. We show that the system can be effectively used for classifying microarray gene expression data. The results demonstrate the committee approach consistently outperforms individual kNN classifiers in terms of both classification accuracy and stability.

Differential Expression in Clear Cell Renal Cell Carcinoma Identified by Gene Expression Profiling

Gene expression profiling has been shown to provide prognostic information on patients with solitary sporadic renal cell carcinoma. To our knowledge there is no reliable way to differentiate synchronous renal metastasis from bilateral primary tumors in patients with bilateral renal cell carcinoma. We present data using a custom kidney cancer cDNA array that can predict the outcome in patients with unilateral and bilateral renal cell carcinoma. Fresh frozen tissue from 38 clear cell renal cell carcinomas was analyzed using a cancer cDNA array containing 3,966 genes relevant to cancer or kidney development. Median followup was 5.3 years. Cancer recurred in 12 patients (43%) and 11 (39%) had died by the last followup. Using a training data set of 8 tumors a 44 gene expression profile distinguishing aggressive and indolent clear cell renal cell carcinoma was identified. Of 29 single clear cell renal cell carcinomas 16 and 13 were predicted to be indolent and aggressive, respectively, by the gene expression profile. Recurrence-free survival at 5 years was 68% and 42% in these 2 groups, respectively (p = 0.032). Clear cell renal cell carcinoma classified as indolent or aggressive according to SSIGN (stage, size, grade and necrosis) score showed a 5-year recurrence-free survival rate of 78% and 42%, respectively (p = 0.021). On Cox proportional hazards analysis the gene expression profile was not an independent predictor of recurrence-free survival after accounting for SSIGN score. Gene expression profile classification correlated with cancer specific survival at 5 years in 4 of 4 patients with metachronous clear cell renal cell carcinoma but in only 2 of 4 with bilateral synchronous clear cell renal cell carcinoma. Gene expression profiling using a kidney cancer relevant cDNA array can differentiate between aggressive and indolent clear cell renal cell carcinomas. Gene expression profile results may be most useful for unilateral clear cell renal cell carcinoma when results are discordant with predictions of tumor behavior based on standard clinicopathological features. In addition, gene expression profiling can provide prognostic information that may help characterize tumors of unknown clinical stage, such as bilateral metachronous clear cell renal cell carcinoma.

Gene expression profile class prediction using linear Bayesian classifiers

Due to recent advances in DNA microarray technology, using gene expression profiles, diagnostic category of tissue samples can be predicted with high accuracy. In this study, we discuss shortcomings of some existing gene expression profile classification methods and propose a new approach based on linear Bayesian classifiers. In our approach, we first construct gene-level linear classifiers to identify genes that provide high class-prediction accuracies, i.e., low error rates. After this screening phase, starting with the gene that offers the lowest error rate, we construct a multi-dimensional linear classifier by incorporating next best-performing genes, until the prediction error becomes minimum or 0, if possible. When we compared classification performance of our approach against prediction analysis of microarrays (PAM) and support vector machines (SVM) based approaches, we found that our method outperforms PAM and produces comparable results with SVM. In addition, we observed that the gene selection scheme of PAM could be misleading. Albeit SVM achieves relatively higher prediction performance, it has two major disadvantages: Complexity and lack of insight about important genes. Our intuitive approach offers competing performance and also an efficient means for finding important genes.

Feature Selection for Predicting Tumor Metastases in Microarray Experiments using Paired Design

Among the major issues in gene expression profile classification, feature selection is an important and necessary step in achieving and creating good classification rules given the high dimensionality of microarray data. Although different feature selection methods have been reported, there has been no method specifically proposed for paired microarray experiments. In this paper, we introduce a simple procedure based on a modified t-statistic for feature selection to microarray experiments using the popular matched case-control design and apply to our recent study on tumor metastasis in a low-malignant group of breast cancer patients for selecting genes that best predict metastases. Gene or feature selection is optimized by thresholding in a leaving one-pair out cross-validation. Model comparison through empirical application has shown that our method manifests improved efficiency with high sensitivity and specificity.

Gene Expression Profile Classification: A Review

In this review, we have discussed the class-prediction and discovery methods that are applied to gene expression data, along with the implications of the findings. We attempted to present a unified approach that considers both class-prediction and class-discovery. We devoted a substantial part of this review to an overview of pattern classification/recognition methods and discussed important issues such as preprocessing of gene expression data, curse of dimensionality, feature extraction/selection, and measuring or estimating classifier performance. We discussed and summarized important properties such as generalizability (sensitivity to overtraining), built-in feature selection, ability to report prediction strength, and transparency (ease of understanding of the operation) of different class-predictor design approaches to provide a quick and concise reference. We have also covered the topic of biclustering, which is an emerging clustering method that processes the entries of the gene expression data matrix in both gene and sample directions simultaneously, in detail.