Closed rmovva closed 6 years ago
Good idea and great figures so far.
On Wed, Jul 26, 2017 at 11:14 AM, Rajiv Movva notifications@github.com wrote:
Fig. 2: Enhancer-promoter interactions can be disrupted by noncoding variants in transcription factor motifs, affecting downstream gene expression.
[image: enhancerpromoterlandscape_rev1_26jul] https://user-images.githubusercontent.com/17256372/28636624-96b3d7c6-71f3-11e7-9978-472e70f83d7b.png
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kundajelab/mpra/issues/3#issuecomment-318137945, or mute the thread https://github.com/notifications/unsubscribe-auth/AAI7EYCQVNMQ4EkrUCLrCEDcoerWT3ayks5sR4IQgaJpZM4OcZ-E .
Here I will be posting some drafts of figures + captions that I am making along the way. This will hopefully balance out some of my work when I begin writing a paper at the end of summer, and it's also easier to edit figures while I'm working on the relevant tasks. Of course, I'll polish these off in Illustrator when the time comes, but it would be great to hear any thoughts/suggestions (however small) before that.
Fig. 1: Distributions of experimental and predicted SuRE fragment activity scores, separated by predicted chromatin state (cite Ernst K562 chromHMM paper). Numbers of fragments for each chromatin state and their experimental vs. predicted Spearman correlations are shown below. Generally, predicted distributions match shape with experimental distributions, with highest concordance in the Pol2 and TssF states.
(I plan to add some names of groups of states like "Transcription Associated", "Promoter", "Enhancer", "Repressed", as well as row names for the chromHMM states, numbers of fragments, and Spearman correlations. I'll also touch-up the violin borders in Illustrator as some of them look a little cut off.)
With this figure in particular I noticed that the predictions for the TssF and PromP (debatably Tss as well) chromatin states were bimodal in nature. I wonder if performing differential motif discovery (with MEME, probably) on the two ends of the distributions would turn up anything interesting. Something interpretable, like TATA etc. If so, I could add a figure based on that as well.
Next I am working on scatter plots of predictions vs. experiments for the SuRE-seq data, with points colored by replicate variance. Should be done by tomorrow.