Metagene projection for cross-platform, cross-species characterization of global transcriptional states

Abstract

The high dimensionality of global transcription profiles, the expression level of 20,000 genes in a much small number of samples, presents challenges that affect the sensitivity and general applicability of analysis results. In principle, it would be better to describe the data in terms of a small number of metagenes, positive linear combinations of genes, which could reduce noise while still capturing the invariant biological features of the data. Here, we describe how to accomplish such a reduction in dimension by a metagene projection methodology, which can greatly reduce the number of features used to characterize microarray data. We show, in applications to the analysis of leukemia and lung cancer data sets, how this approach can help assess and interpret similarities and differences between independent data sets, enable cross-platform and cross-species analysis, improve clustering and class prediction, and provide a computational means to detect and remove sample contamination.

Fig. 1.

Schematic of the metagene projection methodology.

Fig. 2.

Heat maps of metagene projection of leukemia samples. These heat maps of the H_M and H_T matrices show the metagene expression levels for each sample. Each factor clearly corresponds to same leukemia subtype in both model (Left) and test (Right) sets.

Fig. 3.

Hierarchical clustering of the leukemia model and test samples. (A) Clustering of the merged test and model data sets after metagene projection, i.e., columns of the merged H_M and H_T matrices. (B) Clustering of merged model and test sets normalized but without projection. For clarity, some dendrogram vertical lines have been truncated in A; for full dendrograms see SI Fig. 7.

Fig. 4.

Leukemia subclasses metagene projection. Heat maps of the model (A) and test (B) sets after metagene projection show consistent representation of subtype structure across technology platform and laboratory group. SI Text contains a detailed description of the different leukemia subtypes shown here.

Fig. 5.

Metagene projection of the lung cancer data set. Heat maps showing projection of model and test data sets into four-metagene space F1-F4 (A) and three-metagene space F1-F3 after numerical removal of normal component F4 and reconstruction of model (B). AD, adenocarcinoma; SQ, squamous; C, cell lines; NL, normal lung.