Congrats!! Will read it ASAP

Looking at the STARR + ML my guess would be a paper from the Stark lab @IMP Vienna. But I might be wrong

www.imp.ac.at/groups/alexa...

Sounds like someone opened a for loop for downloading stuff and never closed it.

like E-box, E2Fs motifs, with larger structures like CpGi, and it turns out it's way harder that i thought. That's the reason in the post i'm referring to "not as clear" as a single motif instance for a single TF, as here it's most likely the cumulative sum of different TFs that shapes recruitment

Yes, that's exactly what i was thinking. It's incredibly unclear and therefore I reckon it's worth the shot. I don't know how much i can fit the Drosophila into the lab's interests (we love mammalian cells), but i'll try. One thing though, is i need to balance known small local features

Yes, and this is definitely an option. I was thinking some like this as a starting point, or either Hox gene clusters, which are very known in the field.

Though local attributions aren't always clear, that's why i wanted to try first with a "simpler" but probably more robust model.

That's the assumption that has strong foundations (at least to be tested). I'm still gathering metrics from cross-val and other tests, but the performance is good. I can also say that with more complex models (i.e. borzoi), i could get the same predictive performance.

Transcription is there. I’m modeling one cell type for now, mouse embryonic stem cells, as this is the starting point for mechanistic studies. The idea is to figure out what sequence structure models different signals of polycomb proteins.

Thanks for asking and sorry it wasn't clear enough. Sequence —> Polycomb ChIPseq for now. There are some proteins in these complexes that have putative binding sequences but it’s super unclear, I want to figure out what’s going on.
I’ll might add CAGE/RNAseq at some point as the interplay with

I wanna learn promoter/target gene’s structure. Any ideas? Previous experience? Advice? People I should talk to? Please RT if you in the DL for genomics space.