Dataset Information

Decentralized Learning Framework of Meta-Survival Analysis for Developing Robust Prognostic Signatures.

ABSTRACT: PURPOSE:A significant hurdle in developing reliable gene expression-based prognostic models has been the limited sample size, which can cause overfitting and false discovery. Combining data from multiple studies can enhance statistical power and reduce spurious findings, but how to address the biologic heterogeneity across different datasets remains a major challenge. Better meta-survival analysis approaches are needed. MATERIAL AND METHODS:We presented a decentralized learning framework for meta-survival analysis without the need for data aggregation. Our method consisted of a series of proposals that together alleviated the influence of data heterogeneity and improved the performance of survival prediction. First, we transformed the gene expression profile of every sample into normalized percentile ranks to obtain platform-agnostic features. Second, we used Stouffer's meta-z approach in combination with Harrell's concordance index to prioritize and select genes to be included in the model. Third, we used survival discordance as a scale-independent model loss function. Instead of generating a merged dataset and training the model therein, we avoided comparing patients across datasets and individually evaluated the loss function on each dataset. Finally, we optimized the model by minimizing the joint loss function. RESULTS:Through comprehensive evaluation on 31 public microarray datasets containing 6,724 samples of several cancer types, we demonstrated that the proposed method has outperformed (1) single prognostic genes identified using conventional meta-analysis, (2) multigene signatures trained on single datasets, (3) multigene signatures trained on merged datasets as well as by other existing meta-analysis methods, and (4) clinically applicable, established multigene signatures. CONCLUSION:The decentralized learning approach can be used to effectively perform meta-analysis of gene expression data and to develop robust multigene prognostic signatures.

SUBMITTER: Cui Y

PROVIDER: S-EPMC6873986 | biostudies-literature | 2017 Nov

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Decentralized Learning Framework of Meta-Survival Analysis for Developing Robust Prognostic Signatures.

Cui Yi Y Li Bailiang B Li Ruijiang R

JCO clinical cancer informatics 20171101

<h4>Purpose</h4>A significant hurdle in developing reliable gene expression-based prognostic models has been the limited sample size, which can cause overfitting and false discovery. Combining data from multiple studies can enhance statistical power and reduce spurious findings, but how to address the biologic heterogeneity across different datasets remains a major challenge. Better meta-survival analysis approaches are needed.<h4>Material and methods</h4>We presented a decentralized learning fr ...[more]

PMID: 30657395

Similar Datasets

Project description:Advances in transcriptomics have impacted the way of looking atimproved our understanding of leukemic development and helped enhancing the stratification of patients. Analyses of individual cohortsTranscriptomics studies often combined AML samples regardless of cytogenetic abnormalities. , which would lead to bias in Hence, deregulated genesdifferential gene expression analysis that may result from the differential representation of in less frequent AML subgroups have low weight, contrarily to those in abundant subgroups. We, thusHence, we performed a horizontal meta-analysis that integrated transcriptomic data of AML from multicentric multiple studies to enrich the less frequent cytogenetic subgroups and to uncover common genes involved in AML development and response to therapy. A total of 28 Affymetrix microarray datasets harboring 3940 AML samples were downloaded from the GEO database. After stringent quality control, transcriptomics data of 1523 samples from 11 datasets, covering 10 AML cytogenetic subgroups, were retained and merged with the transcriptomic data of 198 healthy bone marrow samples. Differentially expressed genes between each cytogenetic subgroup and normal samples were extracted, allowing the unbiased identification of 330 commonly deregulated genes (CODEGs)., which CODEGs showed enriched expression profiles in myeloid differentiation, leukemic stem cell status and relapse. Most of these genes were downregulated, in accordance with DNA hypermethylation. CODEGS were then used to create a prognostic score based on the weighted sum of expression of 22 core genes (CODEG22). The score was validated in on microarray data of five independent cohorts, and by qRT-PCR in a cohort of 142 samples, by qRT-PCR. CODEG22 stratified patients globally, as well as in subpopulations of cytologically normal and elderly patients. , and hence, couldSo, it may complement the European LeukemiaNet classification for a more accurate prediction of AML outcomes.

Dataset Information

Decentralized Learning Framework of Meta-Survival Analysis for Developing Robust Prognostic Signatures.

Publications

Decentralized Learning Framework of Meta-Survival Analysis for Developing Robust Prognostic Signatures.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets