I have two different stem cell types and their respective gene expressions from RNAseq.
I noticed that there is differential expression in some of the genes coding for histones as well as transcription factors?
What kind of interesting downstream analysis might be available?
Is it feasible to survey ENCODE for downstream TF binding sites? Can anything interesting be investigated with regard to the histone gene data?
Thanks in advance.
Expanding my comment.
TF usually bind in the upper region of the genes it regulates, so you could see which genes are in the region below the binding site. Perhaps there are some databases for TF. I found a website base on the ENCODE project with the genes regulated by each TF.You can use them for pathway or gene set enrichment analysis, but I am not aware of any specific analysis for genes regulated by a single TF.
The histone data is harder if it was methylation I would say to look if different states of the histones are in different spatial state, but with gene expression is harder to conclude anything. Nevertheless you can check methylation studies to see if gene expression is associated with methylation changes.
It sounds like you’ve already done differential expression analysis (with genes or transcripts) and a pathway enrichment analysis to find these. Differential expression of these functional groups is an interesting finding in its own right but there are several further investigations that could follow, depending on your biological research interests.
Clustering the expression patterns of the differentially expressed genes would be able to show if there are relationships between them (such as inter-gene correlations) or relationships to other phenotype data. This is commonly performed with unsupervised hierarchical clustering and displayed by a heatmap. You could also look into whether these differentially expressed genes are expressed in a tissue-specific manner. There is plenty of reference material available from ENCODE, FANTOM4/5, or GTEx that could be used for comparison (as well as cancer datasets) and the Human Cell Atlas Project (currently underway).
Another approach is pathway structure. There are many hierarchical annotation of gene functions such as gene ontology (GO terms) that overlap and are ranked: identifying smaller more specific pathways for future investigations. No pathway annotation is perfect so it’s best to replicate important findings across pathway databases (ideally different expression datasets as well). These functional groups of genes aren’t unordered: the structural relationships between pathways are well characterised in some cases. This includes many gene regulatory networks involving transcription factors. These graph structures are supported on databases such as WikiPathways and Reactome. Network analysis is relatively unexplored compared to differential expression.
While this is a bioinformatics question, there’s always a need to consider experimental validation as a future direction. If you already have a hypothesis based on your analysis, it may be more appropriate to design an experiment or contact experimental collaborators. If you have access to further genomics data, you could also explore how these differences are regulated: CAGE-Seq for enhancers and lncRNA, ATAC-Seq, and 3C/4C for chromatin structure, or Bisulfite-Seq and ChIP-Seq for epigenetic marks. An integrative analysis of expression and epigenomics data would be especially compelling and would likely account for different activity of transcription factor binding sites. Unfortunately, this can’t be done without more ‘omics data available but it would greatly benefit investigations of histones if it is feasible.
