A Critical Evaluation of Analytic Aspects of Gene Expression Profiling in Lymphoid Leukemias with Broad Applications to Cancer Genomics

AIMS Medical Science

article

Giuliano Crispatzu¹  Alexandra Schrader¹  Michael Nothnagel³  Marco Herling¹  Carmen Diana Herling¹ 

[1] Department of Internal Medicine I, Center for Integrated Oncology (CIO) Köln-Bonn, University of Cologne (UoC);Excellence Cluster for Cellular Stress Response and Aging-Associated Diseases (CECAD);Cologne Center for Genomics (CCG), Department of Statistical Genetics and Bioinformatics

关键词: Cancer genomics;    gene expression profiling;    microarray;    RNA-Seq;    survival analysis;    CLL;    T-PLL;    leukemia;    lymphoma;    TCL1;    contamination;    SVM;    random forest;    ;   

DOI  :  10.3934/medsci.2016.3.248

来源: American Institute of Mathematical Sciences

【 摘 要 】

In cancer research, transcriptional aberrations are often deduced from mRNA-based gene expression profiling (GEP). Although transcriptome sequencing (RNA-seq) has gained ground in the recent past, mRNA-based microarrays remain a useful asset for high-throughput experiments in many laboratories. Possible reasons are the lower per-sample costs and the opportunity to analyze obtained GEP data in association with published data sets. There are established and widely used methods for the analysis of microarray data, which increase the comparability of different GEP data sets and facilitate data-mining approaches. However, analytic pitfalls, such as batch effects and issues of sample purity, e.g. by complex tissue composition, are often not properly addressed by these standard approaches. Moreover, most of these tools do not capitalize on the full range of public data sources or do not take advantage of the analytic possibilities for functional interpretation or of comprehensive meta-analyses. We present an overview of the most critical steps in the analysis of microarray-based GEP data. We discuss software and database query solutions that may be useful for each step and for generally overcoming analytic challenges. Aside from machine-learning applications to classify and cluster samples, we describe clinical applications of GEP, including a novel exploratory algorithm to identify potential biomarkers of prognosis in small sample cohorts as demonstrated by exemplary data from lymphatic leukemias. Overall, this review and the attached source code provide guidance to both molecular biologists and bioinformaticians / biostatisticians to properly conduct GEP analyses as well as to evaluate the clinical / biological relevance of obtained results.

【 授权许可】

CC BY