Zeitlinger Lab

GitHub - mmtrebuchet/bpreveal: A suite of tools for machine learning with genomics data A suite of tools for machine learning with genomics data - mmtrebuchet/bpreveal

(10/10) PISA is, at its core, a way to ask how one stretch of DNA affects a biological signal in its surrounding region. If you want to try it out, our complete software suite is available here: github.com/mmtrebuchet/...

(9/10) Our BPReveal package provides tools to engineer sequences with desired properties. For example, we designed mutations to alter a nucleosome’s presence in vivo, and our design was corroborated experimentally.

(8/10) These models detect barrier elements in nucleosome occupancy and chromatin organization. Around barriers, motif effects are asymmetric, and the most asymmetric regions align with known domain boundaries.

(7/10) Overcoming enzymatic sequence biases using PISA reveals the nucleosome motif grammar. Upon correcting the AT-rich sequences of a BPNet model trained to predict MNase-seq nucleosome occupancy, resulting attribution scores clearly identify motifs.

(6/10) PISA can detect enzymatic biases in various sequencing data.

Sequence-to-function models learn both enzymatic bias of the sequencing experiment and the underlying biology. Here, a BPNet MNase-seq model trained in yeast exhibits expected preference for AT-rich sequences.

(5/10) PISA reveals complex motif effects on histone modification ChIP-seq data.

In a H3K27ac BPNet model predicting activity in early embryo fly, a pioneering Zelda motif produces a dual response in H3K27ac profile: a central depletion, flanked by an increase in activity.

(4/10) PISA can reveal “hidden” motifs that possess disparate effects on the output window.

In a ChromBPNet model predicting accessibility in early embryo fly, a CA-rich “Cackle” motif is revealed to possess a positive and negative contributions depending on the output locus.

(3/10) PISA can visualize the influence range of sequences, including TF motifs.

In a mESC BPNet model predicting Nanog binding, PISA can distinguish the nucleosome-range effects of pioneering motif Oct4-Sox2 versus the binding-range effects of the Nanog motif itself.

(2/10) Our new interpretation tool called PISA (pairwise influence by sequence attribution) overcomes this limitation and quantifies how each individual base impacts the predicted readout at each genomic coordinate.

(1/10) When interpreting sequence-to-function models, current attribution methods typically summarize an input base’s effect on the entire output. But what happens if a single nucleotide causes one effect at one output position, but a different effect elsewhere?

The new updates for Charles McAnany’s preprint “Positional Interpretation of Cis-Regulatory Code and Nucleosome Organization with Deep Learning Models” (www.biorxiv.org/content/10.1...) are up!

We introduce PISA, a tool to visualize the cis-regulatory code. See a recap below:

(7/7) Our Model: All promoters use TFIID to load TBP, but TATA promoters additionally allow direct TBP binding to the TATA box. Such dual initiation likely enables faster TBP re-loading and larger transcriptional bursts at TATA promoters.

For more details, check out our work!

(6/7) DPR promoters, which contain downstream sequences favorable for TFIID binding, show the highest levels of downstream TBP. Downstream TBP shows the strongest correlation with TAF2, TAF1 and TAF7, consistent with this being the promoter loading state of TFIID.

(5/7) When binding is normalized by transcription output, we see that TAFs are significantly depleted at TATA promoters. Here, TBP correlates more strongly with TFIIA, TFIIB, TFIIF, NC2, and Mot1 than with the TAFs. Binding at DPR promoters is more homogeneous.

(4/7) What makes the TBP profile promoter-specific? Multiple lines of evidence suggest that TATA promoters show TAF-dependent and TAF-independent initiation.

(3/7) Across promoter types, TAF footprints are strikingly similar, but TBP shows strong promoter-type–specific binding patterns. Using TBP binding patterns alone, we could classify promoters de novo into TATA, DPR, and TCT/housekeeping—and recover their core promoter motifs.

(2/7) We observe all TFIID subunits at active promoters with highly correlated binding, arguing against promoter-specific partial TFIID complexes. Also, our high-resolution DNA footprints match cryo-EM structures —validating them in vivo.

(1/7) How does TFIID function across promoter types in vivo? We mapped all 14 TFIID subunits at base-pair resolution using ChIP-nexus in Drosophila. This lets us directly connect cryo-EM structures, biochemistry, and genetics to promoter behavior in vivo.

High-resolution binding data of TFIID and cofactors show promoter-specific differences in vivo TFIID is instrumental in recognizing promoter sequences and initiating transcription, yet a cohesive understanding of how this complex interacts with and functions at different promoter types in vivo ...

The @zeitlingerlab.bsky.social is pleased to announce @sergio-gma91.bsky.social’s preprint “High-resolution binding data of TFIID and cofactors show promoter-specific differences in vivo” (www.biorxiv.org/content/10.6...).

TLDR; TFIID behaves differently depending on promoter type. More below:

Thanks to coauthors Kaelan Brennan, @kats0805.bsky.social, @hainingjiang.bsky.social, and Sabrina Krueger. Thanks again @rmartinezcorral.bsky.social for your mechanistic modeling, we learned so much from you! Finally, thank you to @juliazeitlinger.bsky.social for your guidance along this journey!

(13) Putting it together, it seems that low-affinity motifs likely evolve easily in enhancers because (1) they arise often (futility theorem), (2) the syntax is flexible, (3) the effect is relatively large. Due to motif cooperativity, even small changes can affect enhancer function.

(12) We tested this idea to an accessibility time-course on decreasing Oct4 concentrations (Xiong et al, from Hans Schöler's lab). When a pioneer motif was in a cooperative vs. single configuration, the enhancer was more sensitive to changing Oct4 conc., regardless of affinity.

(11) We found that the regulatory potential increases when two pioneer TFs cooperate. Motif affinity shifts the curve towards higher or lower TF concentrations, but does not change the regulatory potential. Thus, cooperativity and motif affinity have distinct effects.

(10) By simulating pioneering across changing TF concentrations, we found that if a pioneer TF is bound 100%, it does not guarantee 100% accessibility. We referred to how open chromatin could be at full TF occupancy as the “regulatory potential”.

(9) To understand the implications of pioneering cooperativity, we collaborated with @rmartinezcorral.bsky.social at @crg.eu. We showed that a kinetic modeling framework of nucleosome-mediated TF cooperativity could readily simulate the data from the ChromBPNet model.

(8) This means that low-affinity motifs cooperate as readily as high-affinity motifs, but their relative gain is higher, which is why they produce strong effects.

(7) Looking further, we find that arrangements of pioneer motifs tend to cooperate within nucleosome distances (~200 bp). This cooperative soft syntax applies to every examined pioneering motif pair and all mixtures of motif pair affinities.

(6) Depending on the distance to a strong pioneer motif, the same motif sequence may have different effects on accessibility. This was validated with CRISPR/Cas9 editing on the Akr1cl enhancer, where two identical and bound Sox2 motifs have very different effects on pioneering.

(5) We then found that low-affinity motifs are predicted to have outsized effects on pioneering due to the motif’s arrangement in the genomic region. Surprisingly, this context is a stronger determinant of pioneering than the motif’s affinity alone, as confirmed with CRISPR/Cas9 editing.

(4) To map low-affinity motifs in their genomic context, we trained ChromBPNet (from the lab of @anshulkundaje.bsky.social) deep learning models in mESCs, learning the expected pluripotency TF motifs. We then validated the Oct4-Sox2 motif mappings through high-resolution TF binding footprints.