Identification of Cell States Using Super-Enhancer RNA

1. Identification of Cell States Using
Super-Enhancer RNA
Speaker: Yueh-Hua Tu (杜岳華)
Advisor: Hsuan-Cheng Huang

2. Enhancer
2

3. Super-enhancer (SE)
3
Pott, S., & Lieb, J. D. (2014). What are super-enhancers? Nature Genetics, 47(1), 8–12.
ChIP-seqsignal
Super-enhancer
A B

4. Super-enhancers are cell-specific
4
Whyte, W. A., Orlando, D. A., Hnisz, D., Abraham, B. J., Lin, C. Y., Kagey, M. H., … Young, R. A. (2013). Master
Transcription Factors and Mediator Establish Super-Enhancers at Key Cell Identity Genes. Cell, 153(2), 307–319.

5. Super-enhancer v.s. typical
enhancer
5
Whyte, W. A., Orlando, D. A., Hnisz, D., Abraham, B. J., Lin, C. Y., Kagey, M. H., … Young, R. A. (2013). Master
Transcription Factors and Mediator Establish Super-Enhancers at Key Cell Identity Genes. Cell, 153(2), 307–319.

6. 6
Kim, T.-K., et al (2010). Widespread transcription at neuronal activity-regulated
enhancers. Nature, 465(7295), 182–7.
Enhancer RNA (eRNA)

7. Rationale & Hypothesis
• Define super-enhancer RNA using eRNA
• There exist different cell states between cell types
• Each cell state can be represented by distinct
super-enhancer RNA profile
7

8. Purpose
• To identify cell states using super-enhancer (SE)
RNA.
8

9. FANTOM5 dataset
• Why use this dataset?
• Cap Analysis of Gene Expression (CAGE-seq) can detect
eRNA (without poly-A tail)
• 1829 samples containing cell lines, primary cells, tissues,
time-course (785 samples, 12 cell types)
• Contains stimulated and unstimulated profiles
9
FANTOM5: http://fantom.gsc.riken.jp/

10. CAGE sequencing
10

11. Method
11
eRNA profile
Gene expression
profile
Super-enhancer
RNA profile
From FANTOM5

12. Expression of SE and proximal gene
12

13. Clustering cell types using super-
enhancer RNA
13

14. Cell type classification power
14

15. Apply non-negative matrix
factorization
15

16. Non-negative matrix factorization
16
Customer
rating
matrix
n movies
mcustomers
Super-
enhancer
RNA
profile
n samples
msuper-enhancerRNA
𝑀 = 𝑂𝑅

17. Non-negative matrix factorization
17
W H×
k peferences
mcustomers
n movies
kpeferences
k states
msuper-enhancerRNA
n samples
kstates
𝑀 =

18. NMF on all cell types
18
kstates
n samples

19. iPS differentiate to neuron
19

20. iPS differentiate to neuron
20

21. iPS differentiate to neuron
21

22. Conclusions
• Our proposed super-enhancer RNA can act as a
good alternative for classification of cell type
specification, without complicated measurements
of histone modifications by ChIP-seq.
• Super-enhancer RNA profiles can be used to
identify cell states between cell types.
• Non-negative matrix factorization is a good method
for decomposing large biological data to reveal the
interpretable hidden states
22

23. Future work
• Find out the core regulatory circuitry in each cell or
cancer types.
• How cell migrate from one state to another state?
• Use Markov model or Bayesian network to construct and
organize cell states.
• How core regulatory circuitry behave during migration?
• Construct a virtual cell with dynamic gene
regulatory circuitry.
• If we know how it behave, we can get deeper into
tumorigenesis.
23

24. Thank you for attention
24