GitLab

reorganised sem sim stuff into 3 different scripts. 1 for data prep, 1 for sem sim (java) and 1 for the network reconstruction (readding the tf to module and module to phenotype layers)

small variation from the original results in DOROTHEA - however this difference dissapears if we tighten the p value cut off to .05. semantic similarity is used to re assign edge-weights primarily for heat diffusion

added DOROTHEA analysis - very slight difference to the one presented in the draft manuscript - dont know why will investigate shortly. Also, both my and Eirini's method of using the transcription factor databases relies on using the enrichr api, which updates continuously for the relevant databases and i cant find any legacy files showing the previous version that she used back in 2017. Hence, our enrichments here are slightly different (only one pathway different). To account for this i have included the legacy pathways extracted from pathway commons in AUGUST 2020 so that the results presented in the manuscript can be reproduced. Obviously for the purpose of using these scripts on another database, one should use the updated TF databases available on enrichR

added TRRUST, which is reproducable with eirini's previous work. runpython.py must be run in python2 and also you need to convert the module ENS ids into gene names before you run this using the conversion scripts coded by eirini in /scripts. Also NOTE that in this case we find the trrust regulon enrichments for ALL modules so you need to filter at some point to only include ones that are correlated

BIOPAX pathways derived from TF enrichment and subsequence reactome 2016 enrichment of TFs. I have enountered a reproducability problem in that the transcription factor libraries have been updates since 2017, and i cant seem to find a way to access legacy version. This has no affected my results TOO much, but means that the pathway Degradation of BCatenin is not in our reproduced result