Unknown,Transcriptomics,Genomics,Proteomics

Dataset Information

An Investigation of Biomarkers Derived from Legacy Microarray Data for Their Utility in the RNA-Seq Era

ABSTRACT: Gene expression microarray has been the primary biomarker platform ubiquitously applied in biomedical research, resulting in enormous data, predictive models and biomarkers accrued. Recently, RNA-seq has looked likely to replace microarrays, but there will be a period where both technologies coexist. This raises two important questions: can microarray-based models and biomarkers be directly applied to RNA-Seq data? Can future RNA-Seq-based predictive models and biomarkers be applied to microarray data to leverage past investment? We systematically evaluated the transferability of predictive models and signature genes between microarray and RNA-seq using two large clinical data sets. The complexity of cross-platform sequence correspondence was considered in the analysis and examined using three human and two rat data sets, and three levels of mapping complexity were revealed. Three algorithms representing different modeling complexity were applied to the three levels of mappings for each of the eight binary endpoints and Cox regression was used to model survival times with expression data. In total, 240,096 predictive models were examined. Signature genes of predictive models are reciprocally transferable between microarray and RNA-seq data for model development, and microarray-based models can accurately predict RNA-seq-profiled samples; while RNA-seq-based models are less accurate in predicting microarray-profiled samples and are affected both by the choice of modeling algorithm and the gene mapping complexity. The results suggest continued usefulness of legacy microarray data and established microarray biomarkers and predictive models in the forthcoming RNA-seq era. Definitions of characteristics: EFS day: number of days for event free survival EFS bin: binary classification of event free survival OS day: number of days for overall survival OS bin: binary classification of overall survival High Risk: Indicating whether a sample belongs to high risk group or not A_EFS_All: binary class label for event free survival for all samples B_OS_All: binary class label for overall survival for all samples C_SEX_All: binary class label for sex D_FAV_All: binary class label for favorable and unfavorable samples E_EFS_HR: binary class label for event free survival of High Risk group F_OS_HR: binary class label for overall survival of High Risk group. The same set of Samples is submitted under GEO accession GSE49711. This Series is a reanalysis of the data. The same set of RNA samples were profiled with microarray and RNA-Seq platforms. We explore the transferability of predictive models and signature genes between microarray and RNA-Seq data

ORGANISM(S): Homo sapiens

SUBMITTER: Leming Shi

PROVIDER: E-GEOD-62564 | biostudies-arrayexpress |

REPOSITORIES: biostudies-arrayexpress

ACCESS DATA

Similar Datasets

Project description:Gene expression microarray has been the primary biomarker platform ubiquitously applied in biomedical research, resulting in enormous data, predictive models and biomarkers accrued. Recently, RNA-seq has looked likely to replace microarrays, but there will be a period where both technologies coexist. This raises two important questions: can microarray-based models and biomarkers be directly applied to RNA-Seq data? Can future RNA-Seq-based predictive models and biomarkers be applied to microarray data to leverage past investment? We systematically evaluated the transferability of predictive models and signature genes between microarray and RNA-seq using two large clinical data sets. The complexity of cross-platform sequence correspondence was considered in the analysis and examined using three human and two rat data sets, and three levels of mapping complexity were revealed. Three algorithms representing different modeling complexity were applied to the three levels of mappings for each of the eight binary endpoints and Cox regression was used to model survival times with expression data. In total, 240,096 predictive models were examined. Signature genes of predictive models are reciprocally transferable between microarray and RNA-seq data for model development, and microarray-based models can accurately predict RNA-seq-profiled samples; while RNA-seq-based models are less accurate in predicting microarray-profiled samples and are affected both by the choice of modeling algorithm and the gene mapping complexity. The results suggest continued usefulness of legacy microarray data and established microarray biomarkers and predictive models in the forthcoming RNA-seq era. Definitions of characteristics: EFS day: number of days for event free survival EFS bin: binary classification of event free survival OS day: number of days for overall survival OS bin: binary classification of overall survival High Risk: Indicating whether a sample belongs to high risk group or not A_EFS_All: binary class label for event free survival for all samples B_OS_All: binary class label for overall survival for all samples C_SEX_All: binary class label for sex D_FAV_All: binary class label for favorable and unfavorable samples E_EFS_HR: binary class label for event free survival of High Risk group F_OS_HR: binary class label for overall survival of High Risk group. The same set of Samples is submitted under GEO accession GSE49711. This Series is a reanalysis of the data.

Project description:We generated gene expression profiles from 498 primary neuroblastomas using RNA-Seq and microarrays. We sought to systematically evaluate the capability of RNA deep-sequencing (RNA-Seq)-based classification for clinical endpoint prediction in comparison to microarray-based ones. The neuroblastoma cohort was randomly divided into training and validation sets, and 360 predictive models on six clinical endpoints were generated and evaluated. While prediction performances did not differ considerably between the two technical platforms, the RNA-Seq data processing pipelines, or feature levels (i.e., gene, transcript, and exon junction levels), RNA-Seq models based on the AceView database performed best on most endpoints. Collectively, our study reveals an unprecedented complexity of the neuroblastoma transcriptome, and provides guidelines for the development of gene expression-based predictive classifiers using high-throughput technologies. Sample clinical characteristics definitions: dataset: Expression data set used for classification training (1) or validation (2) Sex: M = male; F= female age at diagnosis: the age in days at diagnosis mycn status: Amplification status of the MYCN proto-oncogene (amplified = 1, no amplification = 0; no information = N/A) high risk: Clinically considered as high-risk neuroblastoma (yes=1, no= 0) INSS stage: disease stage according to International Neuroblastoma Staging System (INSS) (1, 2, 3, 4 and 4S) class label: Maximally divergent disease courses - unfavorable (= 1): patient died despite intensive chemotherapy, favorable (=0): patient survived without chemotharapy for at least 1000 days post diagnosis; not applicable (N/A) progression: Occurrence of a tumor progression event (yes=1; no=0) death from disease: Occurrence of death from the disease (yes=1; no=0) Gene expression of 498 neuroblastoma samples was quantified by RNA sequencing as well as by microarray analyses in order to understand the neuroblastoma transcriptome and predict clinical endpoints.

Project description:Purpose: There is a need for point-of-care diagnostics for future mass casualty events involving radiation exposure. The development of radiation exposure and dose prediction algorithms for biodosimetry is needed for screening of large populations during these scenarios, and exploration of the potential effects which sex, age, genetic heterogeneity, and physiological comorbidities may have on the utility of biodosimetry diagnostics is needed. In the current study, proteomic profiling was used to examine sex specific differences in age matched C57BL6 mice on the blood proteome following radiation exposure and the usefulness of development and application of biodosimetry algorithms using both male and female samples. Methods: C57BL6 male and female mice between 9-11 weeks of age received a single total body radiation exposure of either 2, 4 or 8 Gy with plasma collection at days 1, 3 and 7 post-irradiation. Plasma was then screened using the SomaScan v4.1 assay for ~7000 protein analytes. A subset panel of protein biomarkers demonstrated significant (FDR<0.05 and |logFC|>0.2) changes in expression following radiation exposure. All proteins were used for feature selection to build predictive models of radiation exposure using different sample and sex specific cohorts. Both binary (prediction of any radiation exposure) and multidose (prediction of specific radiation dose) model series were developed using either female and male samples combined or only female or only male samples. The binary series (Models 1, 2 and 3) and multidose series (Models 4, 5 and 6) series included female/male combined, female only and male only respectively. Results: Detectable values were obtained for all ~7000 proteins included in the SomaScan assay for all samples. Each model algorithm built using a unique sample cohort was validated with a training set of samples and tested with a separate new sample series. Overall predictive accuracies in the binary model series was ~100% at the model training level and when tested with fresh samples 97.9% for Model 1(Female and Male) and 100% for Model 2 (Female only) and Model 3 (Male only). When sex specific Models’ 2 and 3 were tested with the opposite sex, the overall predictive accuracy rate dropped to 62.5% for Model 2 and remained 100% for Model 3. The overall predictive accuracy rate in the multidose model series was 100% for all models at the model training level and when tested with fresh samples 83.3%, 75% and 83.3% for Multidose Models 4-6 respectively. When sex specific Models’ 5 (Female only) and Model 6 (Male only) were tested with the opposite sex the overall predictive accuracy rate dropped to 52.1% and 68.8% respectively. Conclusion: These models represent novel predictive panels of radiation responsive proteomic biomarkers and illustrate the utility and necessity of considering sex specific differences in development of radiation biodosimetry prediction algorithms. As sex specific differences were observed in this study, and as use of point-of-care radiation diagnostics in future mass casualty settings will necessarily include persons of both sexes, consideration of sex specific variation is essential to ensure these diagnostic tools have practical utility in the field.

Dataset Information

An Investigation of Biomarkers Derived from Legacy Microarray Data for Their Utility in the RNA-Seq Era

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets