Rendered with Cayman | Work in progress

Project Title

Genome-wide profiling of H3K27me3 and its functional integration with 3D genome architecture in human cells

Location

Papantonis Lab, Universitätsmedizin Göttingen

Background

To be updated

Data

• Micro-C, CUT&Tag, RNA-Seq: In-house generated datasets
• ATAC-Seq: Public dataset from Lyu, Huijue, et al.
  “GATA6 is a novel regulator of gene expression and 3D genome in colorectal cancer.”
  Cancer Research 84.6_Supplement (2024): 4401–4401
• ChIP-Seq: Retrieved from ChIP-Atlas using query: DLD-1

Chapters

• Chapter.1 - Chromatin Architecture in Wild-Type and EED-Mutant Conditions
• Chapter.2 - Differential analysis - How stable is chromatin architecture in the absence of PRC2?
• Chapter.3 - Integrating Chromatin Architecture with Transcriptional Dynamics
• Chapter.4 - Methodological Framework: Tools and Optimization
• Chapter.5 - Exploratory Analyses

Micro-C and Loop Analysis

Category	Tool(s) / Documentation
QC & .pairs generation	Documentation
Loop calling	Mustache, Cooltools-dots, Peakachu
TAD calling	RobusTAD, Arrowhead
A/B compartment analysis	Cooltools eigs-cis, Calder2
Stripes	Stripenn, Stripecaller
Differential analysis	diffDomain

RNA-Seq Analysis

Category	Tool(s)
Gene body analysis	DESeq2
Intron & Exon	iRNASeq
GO & Genomic coordinates	clusterProfiler, TxDb, MyGene.py

ML Modeling

Task	Model(s)	Purpose
Classification	Logistic Regression(baseline), XGBoost	Classify loop status (gained, lost, shared)
Regression	XGBoost, LightGBM, CatBoost	Predict gene expression (log₂FC) from features

Questions

In case of questions about the code or the study, please reach out here

References

Acknowledgments

Many thanks to everyone at Papantonis lab. Also grateful for support from Johannes lab

TO DOs

• HiGlass server
• Nextflow for ML feature table generation
• Ablation testing in models

---

 Turns out PRC2 wasn't repressing genes…it was just trying to exploit some chromatin loop holes.