Really excited to see this work, in collaboration with Thomas Graf from @CRGenomica, finally out. This is the capstone of @aklonizakis PhD and if it doesn't change the way you think about enhancers forever, it will certainly challenge the view you may have had up to now. 🧵👇

RT @Raws_LFC: Let’s take a look at what rival fans were saying about Wirtz BEFORE he chose Liverpool vs. What they are saying AFTER he chos….

Join millions who have switched to Grok.

RT @TheBenShepherd: I take it back. THIS is the best video so far.

A working-class hero is something to be.

RT @ManUtd: Never forgotten.

RT @oilysailor: timely reminder than Alan Shearer is one day younger than Paris Saint-Germain.

RT @aklonizakis: The last part of my PhD thesis is out! If you are interested in enhancer synergy and enhancer clusters have a look at the….

@threadreaderapp please unroll.

This work was driven at both computational and experimental levels by the brilliant @aklonizakis. A great collaboration between our lab at @BSRC_Fleming and the very special (also in my heart) @CRGenomica.

End of thread. Please check out the paper here: and read more about it.

Our discovery that individual enhancers within PTCs may not only act synergistically but also antagonistically and in a differentiation-stage-specific manner, could represent as-yet-overlooked new principles of enhancer interactions.

We don't believe these are singular cases but rather that they represent examples of complex enhancer "grammars" that can give rise to equally sophisticated gene expression programs.

Thus, in both cases we were able to untangle intricate enhancer interactions that can be explained down at the molecular level.

In the case of FOS, on the other hand, we were able to observe a clear binding site of the transcriptional repressor ETV6. Two tandem ETV6 sites had a strong contribution to the chromatin accessibility patterns at both early and late stages of transdifferentiation.

For IRF8, we found binding sites for two positive regulators (PU.1 and CEBPA) being particularly important at the early stage of cell-fate transition.

In this way, we could go deeper in the particular enhancer elements that we had characterized and look at them through a "magnifying glass" in order to identify the DNA sequences that were likely responsible for the binding of transcriptional regulators.

ChromBPNet ( not only identifies potential binding sites of transcriptional regulators but can also predict the degree to which their binding contributes to chromatin accessibility for the given conditions.

What could explain such behaviour? As a last step we deployed ChromBPNet, a deep learning model that can predict the specific contribution of DNA elements in regions of increased chromatin accessibility.

We now had two examples of radically different enhancer behaviour. A synergistic enhancer with elements functioning redundantly and probably compensatorily (IRF8) and -strikingly- an antagonistic one, with two elements functioning in opposite ways.

Modeling also supported the unequal contribution of the two distinct FOS enhancer elements but this time the contributions of the two elements were not only quantitatively but also qualitatively different. The upstream element having an antagonistic, repressive function.

In plain words, by removing an *enhancer* element we observed a significant and robust increase in the gene expression of the downstream gene.