Accurate and sensitive quantification of protein-DNA binding affinity

Rastogi, Chaitanya; Rube, H. Tomas; Kribelbauer, Judith F.; Crocker, Justin; Loker, Ryan; Martini, Gabriella; Laptenko, Oleg; Freed-Pastor, William A.; Prives, Carol; Stern, David L.; Mann, Richard S.; Bussemaker, Harmen J.

doi:10.1073/pnas.1714376115

Cited by 93 publications

(134 citation statements)

References 48 publications

(121 reference statements)

Supporting

Mentioning

120

Contrasting

Order By: Relevance

“…Through evaluation of neural network attribution methods, we observe that single-nucleotide saturated mutagenesis performs well, and similarly to integrated gradients on our dataset. This appears consistent with good performance of mono-nucleotide models (such as [42]) indicating that single-nucleotide perturbations have a strong effect on binding.…”

Section: Discussionsupporting

confidence: 78%

Uncovering tissue-specific binding features from differential deep learning

Phuycharoen¹,

Zarrineh

Bridoux³

et al. 2019

Preprint

View full text Add to dashboard Cite

Motivation: Transcription factors (TFs) can bind DNA in a cooperative manner, enabling a mutual increase in occupancy. Through this type of interaction, alternative binding sites can be preferentially bound in different tissues to regulate tissue-specific expression programmes. Recently, deep learning models have become state-of-the-art in various pattern analysis tasks, including applications in the field of genomics. We therefore investigate the application of convolutional neural network (CNN) models to the discovery of sequence features determining cooperative and differential TF binding across tissues. Results: We analyse ChIP-seq data from MEIS, TFs which are broadly expressed across mouse branchial arches, and HOXA2, which is expressed in the second and more posterior branchial arches. By developing models predictive of MEIS differential binding in all three tissues we are able to accurately predict HOXA2 co-binding sites. We evaluate transfer-like and multitask approaches to regularising the high-dimensional classification task with a larger regression dataset, allowing for creation of deeper and more accurate models. We test the performance of perturbation and gradient-based attribution methods in identifying the HOXA2 sites from differential MEIS data. Our results show that deep regularised models significantly outperform shallow CNNs as well as k-mer methods in the discovery of tissue-specific sites bound in vivo. Availability: For implementation and models please visit https://doi.org/10.5281/zenodo.2635463.

show abstract

Section: Discussionsupporting

confidence: 78%

Uncovering tissue-specific binding features from differential deep learning

Phuycharoen¹,

Zarrineh

Bridoux³

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Low-affinity binding sites facilitate the production of precise temporal and spatial patterns of expression 21,52 . This requirement for low-affinity binding sites, however, means that the entire enhancer may require many transcription factor binding sites, and a specific arrangements of binding sites 53,54 , to preserve kinetic efficiency 21,26,40 . In the case of the E3N enhancer, cooperativity between Hth and Ubx may enable the enhancer to drive precise patterns of 14 expression but make it fragile to mutations.…”

Section: Discussionmentioning

confidence: 99%

“…We examined this question first by analyzing the manifold consequences of mutations in Ubx binding sites. In E3N, Ubx acts as a transcriptional activator and we previously identified multiple low-affinity Ubx binding sites in this enhancer 26,40 .…”

Section: E3n Expressionmentioning

confidence: 97%

Dense encoding of developmental regulatory information may constrain evolvability

Fuqua

Jordan

Breugel

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Gene regulatory changes underlie much of phenotypic evolution. However, the evolutionary potential of regulatory evolution is unknown, because most evidence comes from either natural variation or limited experimental perturbations. Surveying an unbiased mutation library for a developmental enhancer in Drosophila melanogaster using an automated robotics pipeline, we found that most mutations alter gene expression. Our results suggest that regulatory information is distributed throughout most of a developmental enhancer and that parameters of gene expression-levels, location, and state-are convolved. The widespread pleiotropic effects of most mutations and the codependency of outputs may constrain the evolvability of developmental enhancers. Consistent with these observations, comparisons of diverse drosophilids reveal mainly stasis and apparent biases in the phenotypes influenced by this enhancer. Developmental enhancers may encode a much higher density of regulatory information than has been appreciated previously, which may impose constraints on regulatory evolution.

show abstract

“…3D). Strikingly, binding signal loss (defined as the Exd WT -V5 over Exd -shape -V5 coverage ratio at each Exd WT peak summit) showed a strong 265 correlation (r = 0.37, p < 2.2x10 -16 ) with predicted relative affinity for Hth HM -Exd-Antp (45), confirming that this aspect of in vitro binding is recapitulated in vivo; by contrast, ATAC-seq data from wing discs did not show an obvious correlation ( Fig. 3E).…”

Section: Combinatorial In Vitro Tf-complex Binding Behavior Is Recapimentioning

confidence: 98%

“…Finally, indirect pull-down at highly accessible sites (17) or experimental artifacts (18) may also contribute. Although the mechanism by which ChIP enrichment is accrued in the absence of 45 sequence-specific binding is not well understood, recent insights into the compartmentalized structure of eukaryotic nuclei and the formation of transcriptional hubs with high concentration of TFs (19,20) provide a potential explanation.…”

Section: Introductionmentioning

confidence: 99%

Context-dependent gene regulation by transcription factor complexes

Kribelbauer

Loker

Feng

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

Eukaryotic transcription factors (TFs) form complexes with various partner proteins to recognize 15 their genomic target sites. Yet, how the DNA sequence determines which TF complex forms at any given site is poorly understood. Here we demonstrate that high-throughput in vitro binding assays coupled with unbiased computational analysis provides unprecedented insight into how complexes of homeodomain proteins adapt their stoichiometry and configuration to the bound DNA. Using inferred knowledge about minor groove width readout, we design targeted protein 20 mutations that destabilize homeodomain binding in a complex-specific manner. By performing parallel SELEX-seq, ChIP-seq, RNA-seq and Hi-C assays, we not only reveal complex-specific functions, but also show that TF binding sites that lack a canonical sequence motif emerge as a consequence of direct interaction with functionally bound sites.

show abstract

Accurate and sensitive quantification of protein-DNA binding affinity

Cited by 93 publications

References 48 publications

Uncovering tissue-specific binding features from differential deep learning

Uncovering tissue-specific binding features from differential deep learning

Dense encoding of developmental regulatory information may constrain evolvability

Context-dependent gene regulation by transcription factor complexes

Contact Info

Product

Resources

About