Human microarray Analysis

>Independent component analysis

Samples correspond to microarray profiles obtained with Affymetrix HGU133 Plus 2.0 (GPL570) array.
After normalization and quality control dataset contained 106,462 samples and 19,635 genes.
Principal component analysis was applied to explain 90% of the variance of this dataset.
The result of PCA was then used as input for 25 runs of independent component analysis.

(consensus independent components file): Sources reproduced in 13 or more iterations were averaged into consensus sources. 


>Gene set predictions

(consensus mixing matrix file): Mixing matrix was determined algebraically by multiplying the original profiles and the inverse of the consensus sources.

(gene set definitons file) Gene sets used are provided as files. 
	Tab delimited file with three row header. One row per gene set.

(prediction sscores file) Resulting tables of prediction scores are provided as files.

Also:

(mean vector of gene set members file) Transcriptional barcodes generated usign the mixing matrix weights of member genes of every gene set are provided.
	Tab delimited file. One row header, one gene set per row.