A DNA recognition code for probing the in vivo functions of zinc finger transcription factors at domain resolution
Dogan, B.; Kailasam, S.; Corchado, A. H.; Nikpoor, N.; Najafabadi, H. S.
Show abstract
Multi-zinc finger proteins constitute the largest class of human transcription factors. Their DNA-binding specificity is usually encoded by a subset of their tandem Cys2His2 zinc finger (ZF) domains - the subset that binds to DNA, however, is often unknown. Here, by combining a context-aware machine-learning-based model of DNA recognition with in vivo binding data, we characterize the sequence preferences and the ZF subset that is responsible for DNA binding in 209 human multi-ZF proteins. We show that in vivo DNA binding is primarily driven by [~]50% of the ZFs - these DNA-binding ZFs are under strong selective pressure within and across species, and their mutations affect the expression of hundreds of genes as revealed by pan-cancer trans-eQTL analysis across 18 tissues. Among the genes affected by mutations in multi-ZF proteins, we identify several oncogenic factors regulated by SP1, and show that SP1 up-regulation in cancer promotes the expression of these genes while mutations in SP1 ZFs lead to their repression. Together, these analyses suggest that mutations in DNA-binding ZFs have distinct and widespread regulatory consequences that contribute to transcriptome remodelling in cancer.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Recruitment of Homodimeric Proneural Factors by Conserved CAT-CAT E-Boxes Drives Major Epigenetic Reconfiguration in Cortical Neurogenesis 96%
- Deciphering the 3D genome organization across species from Hi-C data 95%
- Single-molecule imaging reveals a direct role of CTCF's zinc fingers in SA interaction and cluster-dependent RNA recruitment 95%
Similar papers in this journal
- An interpretable bimodal neural network characterizes the sequence and preexisting chromatin predictors of induced TF binding 97%
- CREaTor: zero-shot cis-regulatory pattern modeling with attention mechanisms 96%
- Evaluating the representational power of pre-trained DNA language models for regulatory genomics 96%
Similar papers in this journal
- Quantitative occupancy of myriad transcription factors from one DNase experiment enables efficient comparisons across conditions 96%
- DeepArk: modeling cis-regulatory codes of model species with deep learning 95%
- Enhancers display constrained sequence flexibility and context-specific modulation of motif function 94%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.