I was wondering if anybody could point me to applications for deep learning in omics technologies, and even useful papers if present. I have come across applications in the field of the technologies themselves - mapping, aligning etc but I am more curious as to the applications for deep learning in aiding in, for example, target validation.
An example is once RNA-seq reads have been trimmed, mapped, aligned, counted etc and you have your list of differentially expressed RNAs, are there any frameworks that can help identify or predict causal relationships or key drivers within that DE gene list (down to a very few). This is because alot of work seems to arbitrarily discount a certain amount of genes based on higher or lower counts or FDR figures, yet this could just be an artefact of exponential signalling cascades which can mask the driver. This would be much more helpful than running knockout/in/down studies, faster and cheaper. This can be for proteomics, miRNA-seq, ChIP-seq or any other omics technologies (apart from metabolomics as I am not very familiar with that! :) )
And secondly, in drug discovery as to predicting drugs or molecules that can bind a certain protein. For example, I find protein X is the key driver in my experimental condition, can I run that protein against a database of existing drugs to find out the best inhibitor for that protein? Or vice versa, I have a publicly available molecule (drug) and want to predict its possible targets in a database of all known proteins, or a subset of them?
I have seen some things similar, but not quite hit the nail on the head as to the question I am trying to ask.