TFBS/motif enrichment analysis

[sreichl]

• using Rcistarget (code in macroStim project) or SCENIC packages
• input is the same: regions & genes
• First for gene lists then region lists (=more complicated?)
• https://scenicplus.readthedocs.io/en/latest/
• pycisTarget seems promising, but only for regions? maybe mapping from genes to regions is required & easy
• pycistarget (Rcistarget) based
• with summarization to maxNES (and/or enrGene Numbers? -> config?)

[sreichl]

https://github.com/aertslab/pycistarget

[sreichl]

use HOCOMOCO in 2024 https://academic.oup.com/nar/advance-article/doi/10.1093/nar/gkad1077/7424436

[dariarom94]

Here is my solution using pycisTarget (coded for regions, but theoretically should work also for genes). That's a snippet of rules, scripts, environments and short instructions, bringing from a region set and consensus region set to a enrichment table - I didn't include the aggregarion step, but I have code for it also, let me know enrichment_pycistopic.zip . Hope it helps :)

[sreichl]

@dariarom94 amazing thanks!

[sreichl]

• [x] implement cistarget with one standard configurable database
• [x] implement multiple databases
• [x] implement plotting
• [x] implement aggregation
• [x] adapt summary plot
  • problem: direct annotation field in the results is not unique and sometimes empty (term_col is expected to be unique) -> maybe combine term_col with motif name into a description column?
  • maybe a maxNES approach is necessary per TF?
• [x] consider removing summary heatmap plots and only keep bubble heatmap -> most appealing, shows more data and is also clustered (reduce clutter and redundancy) -> factorize the script accordingly
  • No, keep. Cost nothing and provide a different perspective (e.g., adjp) and details (eg dendrogram)
• [x] remove SVG
• [x] consider removing the "significant aggregation CSV" -> not used downstream -> no potentially useful and small
• [x] implement gene-based TFBS enrichment (using Rcistarget)
• [x] consider expanding to creation of custom cistarget database based on the background regions
  • No, just point to the GitHub repo for custom database creation