Independent component analysis reveals new and biologically significant structures in micro array data.

An alternative to standard approaches to uncover biologically meaningful structures in micro array data is to treat the data as a blind source separation (BSS) problem. BSS attempts to separate a mixture of signals into their different sources and refers to the problem of recovering signals from several observed linear mixtures. In the context of micro array data, "sources" may correspond to specific cellular responses or to co-regulated genes. We applied independent component analysis (ICA) to three different microarray data sets; two tumor data sets and one time series experiment. To obtain reliable components we used iterated ICA to estimate component centrotypes. We found that many of the low ranking components indeed may show a strong biological coherence and hence be of biological significance. Generally ICA achieved a higher resolution when compared with results based on correlated expression and a larger number of gene clusters with significantly enriched for gene ontology (GO) categories. In addition, components characteristic for molecular subtypes and for tumors with specific chromosomal translocations were identified. ICA also identified more than one gene clusters significant for the same GO categories and hence disclosed a higher level of biological heterogeneity, even within coherent groups of genes. Although the ICA approach primarily detects hidden variables, these surfaced as highly correlated genes in time series data and in one instance in the tumor data. This further strengthens the biological relevance of latent variables detected by ICA.