A gold standard for transcription factor targets were obtained from the study by Chen et al. (2008 (link)). The article defined an association score between each gene and a transcription factor that varied from 0 to 1. As scores showed a clear bimodal distribution, we considered all genes with a score >0.6 to be targets of the corresponding transcription factor.
In the MEM webtool (http://biit.cs.ut.ee/mem ), we selected 12 datasets on mouse Affymetrix platform 430 2.0. Since the ChIP-seq study was performed on mouse embryonic stem (ES) cells, we selected only the datasets that mentioned ES cell in their description. The list of datasets is available in the Supplementary Material .
We queried each of the transcription factors, for which we had binding information, separately. The MEM webtool performed a similarity search on each dataset using correlation distance between the transcription factor and other genes. The resulting lists of correlated genes were used in aggregation and assessing the AUC.
To combine the data from ChIP-seq with co-expression queries, we translated all the results into Ensembl gene identifiers using g:Convert (Reimand et al., 2007 (link)).
In the MEM webtool (
We queried each of the transcription factors, for which we had binding information, separately. The MEM webtool performed a similarity search on each dataset using correlation distance between the transcription factor and other genes. The resulting lists of correlated genes were used in aggregation and assessing the AUC.
To combine the data from ChIP-seq with co-expression queries, we translated all the results into Ensembl gene identifiers using g:Convert (Reimand et al., 2007 (link)).
