Pol II pausing is an important regulatory step in gene expression. Its present in most animals, but absent in yeast. How and when did pausing evolve? . Check out our latest by @alexachivu.

RT @anshulkundaje: Great work from @missing_a_rib ! First convincing case I've seen for improved performance of sequence models (of nascent….

RT @missing_a_rib: Does training sequence-to-function models on matched functional & genome sequences improve variant effect prediction? Fo….

Latest work from our lab on improving variant interpretation for sequence to function models. Congrats to @missing_a_rib and @NathanPalamutt3 .

Our work led by the great @missing_a_rib!. See his follow up poster/ short talk on how using personalized genotypes improves SNP effect prediction. #MLCB2024.

RT @compbiologist: Concurrently-developed deep learning models to capture transcription initiation syntax:. @anshulkundaje lab's ProCapNet….

So in sum, these are interesting and valuable experiments. I look forward to talking over our separate interpretations with Chunaram and members of his lab. Hope to see them at a conference soon!.

I think this is a lovely result that is very consistent with the cog model we argue for in our paper.

Note the global increase in NVP-2 in both B and C. The effect is specific to NVP-2 (where paused Pol II changes); not to ActD (where Pol II may not change at all).

So if there's an influence of transcription on histone mods, blocking pause release and increasing paused Pol II may increase histone acetylation near that pause site. This is *precisely* the result observed in the NVP-2 experiments in Fig. 5B and C in this new paper.

Data in the Choudhary lab paper actually supports this point: .First, remember that blocking pause release increases the amount of Pol II in the pause, near the TSS of protein coding genes. See here:

That's why the correlations between transcription and marks are so strong (our paper uses ML methods to impute K27ac from Pol II with nearly the accuracy of ChIP-seq replication).

So where does transcription come into this? . Our paper argues a model where histone mods and Pol II are part of a large complex that transcribes genes. The mods are "cogs" (to borrow words from Henikoff and Shilatifard), or "gears" in that machine.

As stated above, I do 100% agree with the authors' points that histone acetylation is driven by HATs (enzymes which acetylate proteins, including histones) and that these are recruited directly by TFs.

So in sum, we're all measuring different things and the two papers do not really contradict each other directly.

Two ways K27ac could be lost from chromatin: (1) HDAC activity, and (2) removal of the +1/ -1 nucleosomes from chromatin. Our work presents evidence that both mechanisms are involved.

Our work observed a global loss in K27ac *on DNA*. We did not examine whole cell K27ac or acetylation of other targets; Choudhary did not examine chromatin bound K27ac. We're measuring different things.

Second, the authors' mass spec and IF experiments examine total histone/ protein acetylation for many acetylation targets.

Thus, these perturbations are very different from the nearly complete removal of Pol II from promoters conducted by our work and the Howe lab study.

Both small molecules may not disrupt interactions between chromatin-bound the Pol II complex and any other molecules they are interacting with.

First, the authors primarily use two small molecules to inhibit different aspects of transcription: ActD, which intercalates into DNA and blocks elongation, but does not remove Pol II, and NVP-2, which blocks pause release, leaving (or increasing) paused Pol II.