ZBTB Transcription Factors Refine Cohesin Recruitment at Gene Promoters beyond CTCF

Cohesin complexes organize genomes through loop extrusion, with CTCF positioning cohesin at chromatin boundaries. Whether other transcription factors independently contribute to cohesin recruitment at promoters remains unclear. We developed an integrative machine learning framework to evaluate cohesin binding across 61.471 promoters in HEK293T cells, combining ChIP–seq data for CTCF, cohesin subunits (RAD21, SMC3), and five ZBTB transcription factors. While CTCF alone achieved near–perfect prediction (AUROC > 0.98), ZBTB inclusion yielded significant improvements (∆AUROC = 0.006–0.011, $$p\,~ \leqslant \,~0.015$$ ), driven by ZBTB7B and ZBTB21. Partial correlations confirmed these contributions were CTCF–independent (partial ρ = 0.31–0.43), and SMC3 benefited twice as much as RAD21, suggesting subunit–specific regulation. Findings were robust across chromosome blocked validation, permutation testing, and spatial shuffling controls. These results reveal ZBTB factors as a secondary regulatory axis refining cohesin positioning, with implications for cell-type-specific chromatin architecture. This work demonstrates how integrative modeling can uncover hidden regulatory layers within well characterized systems, an approach broadly applicable across functional genomics.