Hidden variable analysis of transcription factor cooperativity from microarray time courses

Gene expression is regulated by transcription factor activity, which can be extremely difficult to measure directly. Previous work has established a method to extract the `hidden` transcription factor activity profile from microarray data and use it to effectively identify genes that are targets of a single transcription factor. However, most genes are regulated by two or more transcription factors, and so may not be recognised by this method. Here, the authors present a model-based analysis technique which is able to extract two separate `hidden` transcription factor profiles using microarray data from wild-type and gene knock-down samples. The algorithm can predict targets of each of the transcription factors as well as the amount of cooperative regulation of genes which occurs because of the interaction between the two transcription factors. The authors evaluate this method using simulated data, and show that it is highly effective at classifying genes into categories based on their relative regulation by each of the transcription factors. The authors also show that our method can accurately measure the effectiveness of a gene knock-down when including of a reasonable amount of measurement error.