Dynamics of 3D Enhancer-Gene Associations Across Diverse Human Cell Types and Tissues

on YouTube ↗37m

" summary

A hierarchical probabilistic mixed‑membership model links chromatin and expression across 127 Roadmap+ENCODE cell types to map enhancers to genes. The pipeline calls ~2 million candidate enhancers, learns 61 enhancer and 65 gene modules, and predicts ~250,000 enhancer–gene links at 1% FDR connecting ~132,000 enhancers to ~15,000 genes; predictions show tissue-specific enrichment versus Hi‑C, ChIA‑PET and eQTLs, often skips nearest gene, and initial CRISPR validations are underway.

AI-written summary · about

" chapters (29)

1. 0:00Start
2. 1:01Epigenomes and transcriptomes of 127 human tissues/cel-types
3. 1:51Combinatorial chromatin states define regulatory elements
4. 2:27Joint learning of regulatory chromatin states
5. 5:06Chromatin state dynamics across cell-types/tissues
6. 5:49Modular chromatin activity dynamics of 2 million enhancers
7. 7:29Chromatin contact mapping experiments
8. 10:34Computational methods for linking distal elements to genes
9. 12:51What is the relationship between chromatin and expression co-dynamics at linked enhancer- promoter pairs across cell types
10. 14:05A novel prob. model for enhancers-gene linking using chromatin-expression dynamics
11. 14:52Mixed membership model for enhancer/gene modules
12. 16:11Non-linear linking of enhancer-gene modules
13. 19:40Joint learning of modules and associations
14. 20:52Cell-type specific prob. enhancer-gene links
15. 21:43Calling statistically significant links
16. 24:50Learned enhancer-gene modules and their links
17. 25:2361 enhancer modules and their tissue-specificity
18. 25:3865 gene modules and their tissue-specificity
19. 26:28Module linking probabilities
20. 26:51Predicted cell-type specific enhancer-gene are globally predictive of Hic, CHIA-PET data +eQTLs
21. 28:25Comparison to other methods
22. 29:02Properties of long-rang enhancer-gene network
23. 30:21Enhancers often do NOT associate with their nearest promoters
24. 30:44Pathway enrichments based on target genes associated with enhancers overlapping non-coding GWAS variants
25. 31:38Case Study: Predict target genes of CRC variants using enhancer gene links
26. 32:44Case Study: Target genes of key HDL cholesterol GWAS SNP
27. 33:11Case Study: Novel target gene of SNP associated with Ulcerative Colitis
28. 34:10Summary
29. 35:09Acknowledgements

" related talks

• 01h 13mIntroduction to Regulatory Genomics and Epigenomics I: Intro to the Biology of Gene RegulationNadav Ahituv · Algorithmic Challenges in Genomics Boot Camp
• 35mMachine Learning for Single-Cell 3D EpigenomicsJian Ma · From Algorithms to Discovery in Genome-Scale Biology and Medicine 2022
• 48mVariation in Regulatory Information Within and Between SpeciesMike Snyder · Computational Theories of Evolution
• 38mInterpreting Deep Learning Models Of Functional Genomics Data To Decode Regulatory Sequence...Anshul Kundaje · Interpretable Machine Learning in Natural and Social Sciences (2022)
• 33mUnderstanding Molecular Complexity for Precision MedicineNatasa Przulj · From Algorithms to Discovery in Genome-Scale Biology and Medicine 2022
• 05mDecoding the Human Genome with Deep LearningOlga Troyanskaya · O'Reilly Artificial Intelligence Conference 2019 - New York, New York
• 43mAlgorithms for Inferring Phenotypes from Ancient DNATony Capra · From Algorithms to Discovery in Genome-Scale Biology and Medicine 2022
• 57mBig Data and New Models Needed to Study DNA Variation in Evolution and CancerDavid Haussler · Symposium on Visions of the Theory of Computing 2013