Assign sample identities based on known gene signatures in RNA-seq data

Question

[igor]

Is there an established method to assign sample identities based on known gene signatures in RNA-seq data? There are certain marker genes that are cell type specific. If those genes are present in a particular sample, that sample likely contains those cells. If you have good markers that are differentially expressed, it's easy to spot. People often make heatmaps to show how good those genes are at classifying the samples. However, is there a way to quantify that? I would like to classify many samples against many signatures in a more automated fashion. It seems like there should be an existing method to do that, but I haven't seen one. Any suggestions welcome.