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

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