Skip to main content


Fig. 1 | BMC Genomics

Fig. 1

From: Determining the optimal number of independent components for reproducible transcriptomic data analysis

Fig. 1

Defining Maximally Stable Transcriptomic Dimension (MSTD) value in 37 transcriptomics cancer datasets (13,027 samples in total). In a-c) an example of the analysis is presented for the largest breast cancer dataset METABRIC. a stability profiles for ICA decompositions in various dimensions (from 2 to 100) shown by grey lines. Two-line clustering result is shown by blue and red dashed lines, with MSTD determined as the point of their intersection (vertical dashed line). b average stability profile SM total (blue line) and the average stability of 10 most stable components SM(10) (red line). c visualizing the results of computing ICA 100 times with MSTD = 29 components in the METABRIC dataset and component clustering (icasso package, Canonical Correlation Analysis (CCA) plot). Each black point represents a component, red lines show significant correlations between them, polygons show the convex hull area of the clusters. d Dependence of MSTD on the number of samples for all datasets. e Dependence of the fraction of explained variance on MSTD for all datasets

Back to article page