Slides from my Ph.D. defense that took place on June 23rd, 2017 at the University of Copenhagen

1. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Towards a systems-level understanding of RNA
secondary structure and interactions
PhD defense
Alexander Junge
Center for non-coding RNA in Technology and Health
Department of Veterinary and Animal Sciences
Faculty for Health and Medical Sciences
University of Copenhagen
June 23rd , 2017
1 / 42

2. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Non-coding RNAs (ncRNAs) are important cellular players
Nucleus
Cytoplasm
tRNA
rRNA
microRNA
mRNA
mRNA
2 / 42

3. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Non-coding RNAs (ncRNAs) are important cellular players
Xist lncRNA
Nucleus
Cytoplasm
tRNA
rRNA
microRNA
mRNA
snoRNA
Nucleolus
rRNA
mRNA
2 / 42

4. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Levels of RNA structure
UUGCGUGGAUAUGGCACGCAAGUUUCUAC Primary structure
• nucleotide sequence
• adenine (A), cytosine
(C), guanine (G),
uracil (U)
3 / 42

5. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Levels of RNA structure
UUGCGUGGAUAUGGCACGCAAGUUUCUAC
(((((((.......)))))))........
G
G
A
C
A
U
A
U
A
A
UU
G
C
G
U
G
G
A
U
A
U G
G
C
A
C
G
C
A A
G
U
U U C
U
A
C
C G G G
C
A
C
C G
U
A
AA
U
G
UCCG
A
C
U
A
U
G
U
C
C
Primary structure
Secondary structure
• A-U, G-C, G-U base
pairs, unpaired
nucleotides
• dot-bracket
notation/structure
diagram
3 / 42

6. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Levels of RNA structure
UUGCGUGGAUAUGGCACGCAAGUUUCUAC
(((((((.......)))))))........
G
G
A
C
A
U
A
U
A
A
UU
G
C
G
U
G
G
A
U
A
U G
G
C
A
C
G
C
A A
G
U
U U C
U
A
C
C G G G
C
A
C
C G
U
A
AA
U
G
UCCG
A
C
U
A
U
G
U
C
C
Primary structure
Secondary structure
Tertiary structure
• 3D shape
• PDB accession
number 4FE5
3 / 42

7. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Importance of RNA secondary structure
UUGCGUGGAUAUGGCACGCAAGUUUCUAC
C
A
U
A
U
A
A
UU
G
C
G
U
G
G
A
U
A
U G
G
C
A
C
G
C
A A
G
U
U U C
U
A
C
C G G G
C
A
C
C G
U
A
AA
U
G
UCCG
A
C
U
A
U
G
• secondary structure is basis for tertiary structure
• secondary structure important for function
• efficient to compute
4 / 42

8. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
RNA secondary structure prediction
For a single sequence:
• find energetically most
stable structure
• e.g., RNAfold [Hofacker et al.,
Monatsh Chem, 1994],[Lorenz et al.,
Algo Mol Biol 2011]
For multiple, related sequences:
• leverage evolutionary signal
in prediction
• e.g., PETfold [Seemann et al.,
Nucl Acids Res, 2008]
5 / 42

9. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Base pair conservation
((((((((.....(((((.......)))))..........(((((((.......))))))
ACCUCGUAUAAUAGCAGGGAUAUGGCUUGCAAGUUUCUACCCGACGACCCUAAAUCGUUG
CCAUCGUAUAUAUUCAGGGAUAUGGCCUGAACGUUUCUACCAAGCUGCCGUAAAUAGCUU
UGCCUAUAUAA-UGCCAUGAUAUGGAUGGCGAGUUUCUACCCAGUG-CCGUAAA-CACUG
CCCUCAUAUAU-AAUGGGAAUACGGCCCAUACGUCUCUACCCGGUUACCGUAAAUAAUCG
.........10........20........30........40........50........6
)..))))))))
GACUAUGGGGU
GACUACGAGGU
GACUAUAAGCG
GACUAUGAGGG
0........70
no sequence variation
2 types of base pairs
3 types of base pairs
6 / 42

10. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Base pair conservation
((((((((.....(((((.......)))))..........(((((((.......))))))
ACCUCGUAUAAUAGCAGGGAUAUGGCUUGCAAGUUUCUACCCGACGACCCUAAAUCGUUG
CCAUCGUAUAUAUUCAGGGAUAUGGCCUGAACGUUUCUACCAAGCUGCCGUAAAUAGCUU
UGCCUAUAUAA-UGCCAUGAUAUGGAUGGCGAGUUUCUACCCAGUG-CCGUAAA-CACUG
CCCUCAUAUAU-AAUGGGAAUACGGCCCAUACGUCUCUACCCGGUUACCGUAAAUAAUCG
.........10........20........30........40........50........6
)..))))))))
GACUAUGGGGU
GACUACGAGGU
GACUAUAAGCG
GACUAUGAGGG
0........70
no sequence variation
2 types of base pairs
3 types of base pairs
Evolutionarily conserved secondary structure results in
• consistent mutation C-G to U-G
• compensatory mutation C-G to A-U
6 / 42

11. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Covariation analysis to identify conserved structure
((((((((.....(((((.......)))))..........(((((((.......))))))
ACCUCGUAUAAUAGCAGGGAUAUGGCUUGCAAGUUUCUACCCGACGACCCUAAAUCGUUG
CCAUCGUAUAUAUUCAGGGAUAUGGCCUGAACGUUUCUACCAAGCUGCCGUAAAUAGCUU
UGCCUAUAUAA-UGCCAUGAUAUGGAUGGCGAGUUUCUACCCAGUG-CCGUAAA-CACUG
CCCUCAUAUAU-AAUGGGAAUACGGCCCAUACGUCUCUACCCGGUUACCGUAAAUAAUCG
.........10........20........30........40........50........6
)..))))))))
GACUAUGGGGU
GACUACGAGGU
GACUAUAAGCG
GACUAUGAGGG
0........70
no sequence variation
2 types of base pairs
3 types of base pairs
• identification of consensus structure
• build statistical models of conserved structure and sequence
→ covariance models [Durbin and Eddy, Nucl Acids Res, 1994]
7 / 42

12. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
ncRNAs function by interaction with other biomolecules
8 / 42

13. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
ncRNAs function by interaction with other biomolecules
• different types of interactions (ncRNA–protein,
ncRNA–mRNA, ncRNA–ncRNA)
• different types of evidence (experimental, predictions,
text-mining)
• focus on functional interactions
• physical interactions
• indirect associations (e.g. transcriptional regulation, same
pathway)
8 / 42

14. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
ncRNA–protein interactions
High-throughput experiment: CLIP (crosslinking immunoprecipitation)-seq
RBP
Cell lysis
Ultraviolet crosslinking
RBP
Immunoprecipitation, digestion
high-throughput sequencing,
read mapping, peak calling
Reverse transcription
9 / 42

15. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
ncRNA–mRNA interactions
microRNA–target prediction
microRNA
mRNA
translational
inhibition
mRNA
degradation
10 / 42

16. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
ncRNA–mRNA interactions
microRNA–target prediction
microRNA
mRNA
translational
inhibition
mRNA
degradation
3’-UAGCGUG-5’
5’-UAAGUCGCACUGC-3’ mRNA target
miRNA query
10 / 42

17. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
ncRNA–ncRNA interactions
Text-mining Xist–Tsix co-mentions
11 / 42

18. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
ncRNA–ncRNA interactions
Text-mining Xist–Tsix co-mentions
11 / 42

19. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Systems-level perspective on RNA biology
Study not on a single RNA (type) but the bigger picture:
• RNA sequence/structure → similarity clusters
• ncRNA interactions → interaction networks
• understand system’s behaviour → disease state
Image source: The Opte Project
12 / 42

20. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Main projects during my PhD
RNAscClust: clustering RNAs
using structure conservation and
graph based motifs
[Miladi*, Junge* et al., Bioinformatics,
2017]
RAIN: cataloging ncRNA–RNA
and ncRNA–protein interactions
[Junge*, Garde*, Refsgaard* et al.,
Database, 2017]
RUNX1T1 as a microRNA
sponge in t(8;21) acute myeloid
leukemia
[Junge et al., Gene, 2017]
* equal contribution
C
C
G
C GG
G
C
CC
GG
G
A
A
UA
G
G
5'
3'
1 2 3
4 5
6
graph
kernel
...
Normal
t(8;21) Acute Myeloid Leukemia
+
13 / 42

21. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Outline
Introduction
RNAscClust: clustering RNAs using structure conservation and
graph based motifs
RAIN: RNA–protein Association and Interaction Networks
miRNA sponge RUNX1T1 in t(8;21) acute myeloid leukemia
Conclusion
14 / 42

22. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
ncRNAs with well-defined secondary structure
let-7 microRNA
precursor
H/ACA box snoRNA tRNA
from Rfam database of RNA families [Nawrocki et al., Nucl Acids Res,
2015]
15 / 42

23. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Why clustering RNA secondary structures?
16 / 42

24. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Why clustering RNA secondary structures?
Clustering aims to identify groups of similar structures.
16 / 42

25. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
human
wrong clustering
clustering
GACACAGU
structure prediction
from single sequence
wrong structure
G
A
A
C
C
U
5'
3'
G A
GACCCAGUCCAUCG ACUCAGU
GACACAGU
CCAUCG
ACUCAGU
GACCCAGU
Single sequence clustering
17 / 42

26. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
human
wrong clustering
clustering
human
chimp
pig
mouse
covariation
consensus structure ( . ( . . . ) )
clustering
correct clustering
GACACAGU
UAGCCUCG
CAGUAUUG
A-AACUUU
GACACAGU
structure prediction
from single sequence
structure prediction
based on consensus structure
wrong structure correct structure
G
A
A
C
C
U
5'
3'
G A
A
GU
C
A
GU
C
C
A
G
GU
5'
3'
C
C
A A
A
G
GU
GACCCAGU GACCCAGU
C-GCCUCG
C-GCACUG
A-AACGUU
CCAUCG ACUCAGU ACUCAGU
GUGAUAC
UGCA-UA
GCGA-GC
CCAUCG
-CGUCG
UUGCAA
CGAACG
GACACAGU
CCAUCG GACACAGU
GACCCAGU
CCAUCG
ACUCAGU
GACCCAGU
ACUCAGU
Single sequence clustering Clustering using structure conservation
17 / 42

27. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
RNAscClust - main goals
1. cluster RNAs based on sequence-structure similarities
2. leverage structure conservation information
18 / 42

28. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
RNAscClust projects conserved structure onto a sequence
C
C
G
G
U C G U C C U U U C G A
U G A U C C U C U U C A
U C C A U C - U U G G A
U A C A C C - U U G U A
human
chimp
pig
mouse
threshold t
constrained
folding
multiple
alignment
base pair
reliabilities
human
constraints( ( ( ) ) )
U
A
U
U U
U
C
C
decorated structureplain structure
U
A
U
U U
U
C
CC
C
G
G
U
A
U
U U
U
C
C
N
N N
N
PETfold
RNAfold
R-scape
• extends GraphClust [Heyne et al., Bioinformatics, 2012]
• uses PETfold, RNAfold, R-scape [Rivas et al., Nat Methods, 2017]
19 / 42

29. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Graph kernel-based similarity measure
C
C
G
C GG
G
C
C
G
C
CC
GG
G
A
A
UA
G
G
H 12872
19744
373
9
28674
5'
3'
A
G
A G
A
A
G
1 2 3
4 5
6 0
1
.
.
9
.
.
373
.
.
12872
.
.
19744
.
.
28674
0
0
.
.
1
.
.
1
.
.
1
.
.
1
.
.
2
sparse
feature
vector
H
H
H
H
graph
kernel
N1
1
=
N1
2
=
N1
3
=
N1
5
=
N1
4
=N1
6
=
feature
index
feature
count
cosine
similarity
Neighborhood Subgraph Pairwise Distance Kernel [Costa and De Grave,
ICML, 2010]
20 / 42

30. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Full RNAscClust pipeline
Locality sensitive
hashing
Candidate clustersSparse feature
vectorization
Iterate until no candiate left
1 3 4 5
6 7 8
2
Conservation aware
secondary structure
Input multiple
alignments
Cluster refinement
and extension
Candidate clusters
of representatives
Report
clusters
21 / 42

31. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Constructing benchmark datasets
Split Rfam 12 family seed alignments into subalignments
Human
Mouse
Pig
Chimp
Rfam family
seed alignment
Family
subalignments
split
22 / 42

32. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Constructing benchmark datasets
Split Rfam 12 family seed alignments into subalignments
Human
Mouse
Pig
Chimp
Rfam family
seed alignment
Family
subalignments
split
1. similar sequences from different species form a subalignment
2. each subalignment contains one human sequence
3. datasets with different degrees of sequence identity
22 / 42

33. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Clustering performance increases with additional sequence
variation
50% 63% 73%
0.0
0.2
0.4
0.6
0.8
1.0
AdjustedRandIndex
RNAscClust GraphClust
Mean pairwise sequence identity per alignment
23 / 42

34. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Conclusion - RNAscClust
• graph kernel-derived similarity function
• highlight conserved base pairs in clustering
• potential to cluster large datasets (locality sensitive hashing)
24 / 42

35. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Outline
Introduction
RNAscClust: clustering RNAs using structure conservation and
graph based motifs
RAIN: RNA–protein Association and Interaction Networks
miRNA sponge RUNX1T1 in t(8;21) acute myeloid leukemia
Conclusion
25 / 42

36. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
The STRING database (http://string-db.org/)
• protein-protein interaction
networks for 2031 organisms
• integrates known and
predicted protein-protein
associations, e.g.,
• experiments
• text-mining
• genomic context
• interactions are scored
according to their reliability [Szklarczyk et al., Nucl. Acids Res., 2015]
26 / 42

37. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
The STRING database (http://string-db.org/)
• protein-protein interaction
networks for 2031 organisms
• integrates known and
predicted protein-protein
associations, e.g.,
• experiments
• text-mining
• genomic context
• interactions are scored
according to their reliability
• no such resource for
ncRNA interactions
[Szklarczyk et al., Nucl. Acids Res., 2015]
26 / 42

38. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
RAIN - main features
1. collects ncRNA-RNA and ncRNA-protein interactions from
various sources
2. unifies ncRNA and protein interactions
27 / 42

39. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
RAIN pipeline
m
iRanda
STarM
irDB
STRING:
Protein-Protein
Interactions
Text m
ining
Curated
Knowledge
m
iRDB
TargetScan
PITA
CLIP
based
exp.
NPinter
m
iRTarbase
Predictions
RNA-RNA and RNA-Protein
Interactions
Benchmarking
of Continuous Scores
Benchmarking
of Discrete Scores
Resource
Benchmarking
Resource
Integration
Evidence Channel
Benchmarking
Evidence Channel
Integration
RNA and Protein
Network Integration
[1]
[2]
[3]
[4]
[5]
RAIN:
Complete Networks of
RNA and Protein Interactions
Experiments
28 / 42

40. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
From raw interaction scores to confidences
Agreementwithgoldstandard
Raw interaction score
miRNA Target Raw score
miR-4685-5p KRAS 16.8
miR-548at-3p RUNX1 13.2
miR-205-3p ELF2 2.0
...
...
...
→ translate (predicted) raw scores into confidence scores ∈ [0, 1]
by fitting a calibration curve
29 / 42

41. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
From raw interaction scores to confidences
Gold standard: expert curated miRNA-mRNA interactions
TP FP Unknown
raw interaction score
sliding window
29 / 42

42. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
From raw interaction scores to confidences
Gold standard: expert curated miRNA-mRNA interactions
TP FP Unknown
mean raw interaction score in window
confidencescore(TP/P)
0
1
raw interaction score
sliding window
29 / 42

43. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
RAIN interaction counts
Interactions with confidence score > 0.15
miRNA–mRNA ncRNA–protein ncRNA–ncRNA Total
H. sapiens (Human) 174,853 11,026 2,507 188,386
M. musculus (Mouse) 77,270 469 35 77,774
R. norvegicus (Rat) 19,985 39 1 20,025
S. cerevisiae (Yeast) 0 640 85 725
Total 272,108 12,174 2,628 286,910
30 / 42

44. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
RAIN webinterface
http://rth.dk/resources/rain
31 / 42

45. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
RAIN webinterface
http://rth.dk/resources/rain
KLF4KLF4
FSCN1FSCN1
hsa-miR-145-5phsa-miR-145-5p SOX2SOX2
MYCMYC
IRS1IRS1
Experiments
Text mining
Predictions
Curated
31 / 42

46. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
RAIN webinterface
http://rth.dk/resources/rain
KLF4KLF4
FSCN1FSCN1
hsa-miR-145-5phsa-miR-145-5p SOX2SOX2
MYCMYC
IRS1IRS1
Experiments
Text mining
Predictions
Curated
31 / 42

47. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
RAIN webinterface
http://rth.dk/resources/rain
31 / 42

48. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Conclusion - RAIN
• integrate different sources of ncRNA-target interactions
• convert raw interaction scores to confidences
• provide research community with more complete picture
• protein-protein interactions
• ncRNA-protein interactions
• ncRNA-RNA interactions
32 / 42

49. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Outline
Introduction
RNAscClust: clustering RNAs using structure conservation and
graph based motifs
RAIN: RNA–protein Association and Interaction Networks
miRNA sponge RUNX1T1 in t(8;21) acute myeloid leukemia
Conclusion
33 / 42

50. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
t(8;21) acute myleoid leukemia
• most common form of acute leukemia
• chromosomal translocation t(8;21) → RUNX1-RUNX1T1
fusion transcript
34 / 42

51. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
t(8;21) acute myleoid leukemia
• most common form of acute leukemia
• chromosomal translocation t(8;21) → RUNX1-RUNX1T1
fusion transcript
• RUNX1T1 up to 1000-fold overexpressed
34 / 42

52. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Competing endogenous RNAs (ceRNAs)
= mRNAs competing for miRNA binding
Normal
Unknown ceRNA
miRNA
miRNA binding site
Legend
RUNX1T1
35 / 42

53. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Competing endogenous RNAs (ceRNAs)
= mRNAs competing for miRNA binding
RUNX1T1
overexpression
Normal
t(8;21) Acute Myeloid Leukemia
Unknown ceRNA
miRNA
miRNA binding site
+
Legend
RUNX1T1
35 / 42

54. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Competing endogenous RNAs (ceRNAs)
= mRNAs competing for miRNA binding
RUNX1T1
overexpression
Normal
t(8;21) Acute Myeloid Leukemia
Unknown ceRNA
miRNA
miRNA binding site
+
Legend
RUNX1T1
Expression levels
35 / 42

55. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Competing endogenous RNAs (ceRNAs)
= mRNAs competing for miRNA binding
RUNX1T1
overexpression
Normal
t(8;21) Acute Myeloid Leukemia
Unknown ceRNA
miRNA
miRNA binding site
+
Legend
RUNX1T1
Expression levels
• role of ceRNAs in melanoma, breast cancer [Poliseno and Pandolfi,
Methods, 2015]
RUNX1T1 as ceRNA involved in development of t(8;21)
acute myleoid leukemia?
35 / 42

56. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Competing endogenous RNAs (ceRNAs)
= mRNAs competing for miRNA binding
RUNX1T1 as ceRNA involved in development of t(8;21)
acute myleoid leukemia?
35 / 42

57. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Analysis approach
TCGA RNA-seq and miRNA-seq data for t(8;21) AML
605 genes significantly affected by ceRNA RUNX1T1
Cupid
[Chiu et al., 2015]
RUNX1T1
overexpression
Normal
t(8;21) Acute Myeloid Leukemia
Unknown ceRNA
miRNA
miRNA binding site
+
Legend
RUNX1T1
Expression levels
36 / 42

58. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Analysis approach
TCGA RNA-seq and miRNA-seq data for t(8;21) AML
605 genes significantly affected by ceRNA RUNX1T1
Overrepresentation of
cancer associated
genes among ceRNAs
(p<0.001)
DISEASES
[Pletscher-Frankild et al., 2015]
Cupid
[Chiu et al., 2015]
36 / 42

59. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Analysis approach
TCGA RNA-seq and miRNA-seq data for t(8;21) AML
605 genes significantly affected by ceRNA RUNX1T1
Overrepresentation of
cancer associated
genes among ceRNAs
(p<0.001)
DISEASES
[Pletscher-Frankild et al., 2015]
Cupid
[Chiu et al., 2015]
RAIN
[Junge et al., 2017]
Overrepresentation
of highly reliable miRNA
binding sites (p<0.001)
36 / 42

60. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Analysis approach
TCGA RNA-seq and miRNA-seq data for t(8;21) AML
605 genes significantly affected by ceRNA RUNX1T1
Overrepresentation of
cancer associated
genes among ceRNAs
(p<0.001)
DISEASES
[Pletscher-Frankild et al., 2015]
Cupid
[Chiu et al., 2015]
PANTHER
[Mi et al., 2016]
Gene Ontology and
pathway enrichment
RAIN
[Junge et al., 2017]
Overrepresentation
of highly reliable miRNA
binding sites (p<0.001)
36 / 42

61. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Summary of potential effects of RUNX1T1 microRNA
sponge in t(8;21) AML
RUNX1 RUNX1T1
3' UTR
37 / 42

62. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Summary of potential effects of RUNX1T1 microRNA
sponge in t(8;21) AML
RUNX1 RUNX1T1
3' UTR
Overexpression in
t(8;21) AML
PLAG1 TNFSF11
TCF4
AML oncogene
37 / 42

63. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Summary of potential effects of RUNX1T1 microRNA
sponge in t(8;21) AML
RUNX1 RUNX1T1
3' UTR
Overexpression in
t(8;21) AML
PLAG1 TNFSF11
Cadherin and Wnt
signaling pathways
TCF4
AML oncogene
Pathway
dysregulation
Integrin
signaling pathway
experimental validation of the ceRNA effects needed
37 / 42

64. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Outline
Introduction
RNAscClust: clustering RNAs using structure conservation and
graph based motifs
RAIN: RNA–protein Association and Interaction Networks
miRNA sponge RUNX1T1 in t(8;21) acute myeloid leukemia
Conclusion
38 / 42

65. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Conclusion
Sytems-level understanding of RNA structure and interactions
RNAscClust identifies recurrent, conserved
RNA structures by clustering
RAIN integrates scored ncRNA and protein
interactions
RUNX1T1 as a potentially oncogenic ceRNA
in t(8;21) acute myeloid leukemia
C
C
G
C GG
G
C
CC
GG
G
A
A
UA
G
G
5'
3'
1 2 3
4 5
6
graph
kernel
...
Normal
t(8;21) Acute Myeloid Leukemia
+
39 / 42

66. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Research perspectives
RNAscClust
• improve graph kernel by encoding covariation directly
• cluster RNA structures conserved in vertebrates [Seemann et al.,
Genome Research, 2017]
40 / 42

67. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Research perspectives
RNAscClust
• improve graph kernel by encoding covariation directly
• cluster RNA structures conserved in vertebrates [Seemann et al.,
Genome Research, 2017]
RAIN
• integrate more organisms and sources of interactions
40 / 42

68. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Research perspectives
RNAscClust
• improve graph kernel by encoding covariation directly
• cluster RNA structures conserved in vertebrates [Seemann et al.,
Genome Research, 2017]
RAIN
• integrate more organisms and sources of interactions
RUNX1T1 as ceRNA in t(8;21) AML
• experimental validation of the ceRNA effects
40 / 42

69. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Research perspectives
RNAscClust
• improve graph kernel by encoding covariation directly
• cluster RNA structures conserved in vertebrates [Seemann et al.,
Genome Research, 2017]
RAIN
• integrate more organisms and sources of interactions
RUNX1T1 as ceRNA in t(8;21) AML
• experimental validation of the ceRNA effects
investigate interplay between RNA structure and interactions
by genome-scale mapping of RNA interactions and structure
40 / 42

70. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Acknowledgements
RNAscClust
• Milad Miladi
• Fabrizio Costa
• Rolf Backofen
• Stefan E. Seemann
• Jakob Hull Havgaard
RAIN
• Jan C. Refsgaard
• Christian Garde
• Xiaoyong Pan
• Alberto Santos
• Ferhat Alkan
• Christian Anthon
• Christian von Mering
• Christopher T.
Workman
• Lars Juhl Jensen
ceRNAs in t(8;21) AML
• Roza Zandi
• Jack Cowland
• Jakob Hull Havgaard
Jan Gorodkin and the whole RTH group
Funding:
41 / 42

71. Introduction RNAscClust RAIN ceRNAs in leukemia Conclusion
Thank you for your attention!
42 / 42