Identification of expression patterns in the progression of disease stages by integration of transcriptomic data

BMC Bioinformatics

Research

Sara Aibar¹  Javier De Las Rivas¹  Francisco Jose Campos-Laborie¹  Jose Manuel Sánchez-Santos²  Maria Abaigar³  Jesus M. Hernandez-Rivas³ 

[1] Bioinformatics and Functional Genomics research group, Cancer Research Center (IMBCC, CSIC/USAL) and Instituto de Investigación Biomédica de Salamanca (IBSAL), Salamanca, Spain;Bioinformatics and Functional Genomics research group, Cancer Research Center (IMBCC, CSIC/USAL) and Instituto de Investigación Biomédica de Salamanca (IBSAL), Salamanca, Spain;Department of Statistics, University of Salamanca (USAL), Salamanca, Spain;Unidad de Diagnóstico Molecular y Celular del Cáncer, Cancer Research Center (IMBCC, CSIC/USAL) and Instituto de Investigación Biomédica de Salamanca (IBSAL), Salamanca, Spain;Servicio de Hematología, Hospital Universitario de Salamanca (HUS/IBSAL/USAL), Salamanca, Spain;

关键词: Disease stage;    Disease subtype;    Disease progression;    Cancer;    Leukemia;    Transcriptomics;    Gene expression;    Gene signature;    Expression profiling;    Expression pattern;    Bioinformatics;    Pattern recognition;    Data integration;   

DOI  :  10.1186/s12859-016-1290-4

来源: Springer

【 摘 要 】

BackgroundIn the study of complex diseases using genome-wide expression data from clinical samples, a difficult case is the identification and mapping of the gene signatures associated to the stages that occur in the progression of a disease. The stages usually correspond to different subtypes or classes of the disease, and the difficulty to identify them often comes from patient heterogeneity and sample variability that can hide the biomedical relevant changes that characterize each stage, making standard differential analysis inadequate or inefficient.ResultsWe propose a methodology to study diseases or disease stages ordered in a sequential manner (e.g. from early stages with good prognosis to more acute or serious stages associated to poor prognosis). The methodology is applied to diseases that have been studied obtaining genome-wide expression profiling of cohorts of patients at different stages. The approach allows searching for consistent expression patterns along the progression of the disease through two major steps: (i) identifying genes with increasing or decreasing trends in the progression of the disease; (ii) clustering the increasing/decreasing gene expression patterns using an unsupervised approach to reveal whether there are consistent patterns and find genes altered at specific disease stages. The first step is carried out using Gamma rank correlation to identify genes whose expression correlates with a categorical variable that represents the stages of the disease. The second step is done using a Self Organizing Map (SOM) to cluster the genes according to their progressive profiles and identify specific patterns. Both steps are done after normalization of the genomic data to allow the integration of multiple independent datasets. In order to validate the results and evaluate their consistency and biological relevance, the methodology is applied to datasets of three different diseases: myelodysplastic syndrome, colorectal cancer and Alzheimer’s disease. A software script written in R, named genediseasePatterns, is provided to allow the use and application of the methodology.ConclusionThe method presented allows the analysis of the progression of complex and heterogeneous diseases that can be divided in pathological stages. It identifies gene groups whose expression patterns change along the advance of the disease, and it can be applied to different types of genomic data studying cohorts of patients in different states.

【 授权许可】

CC BY   
© The Author(s). 2016