Fast and effective prediction of miRNA targets

1
Fast and effective prediction of miRNA targets

• Marc Rehmsmeier
• CeBiTec, Bielefeld University, Germany
• Junior Research Group Bioinformatics of Regulation

17.02.2005
2
Small interfering RNAs versus small temporal RNAs
Hannon. Nature. 418244-251, 2002.
3
miRNA/target duplexes
Grosshans and Slack. The Journal of Cell Biology,
156(1)17-21, 2002.
4
A direct approach
Given a miRNA and a potential target What are
the energetically most favourable binding sites?
Calculation of multiple mfe secondary structure
duplexes
5
The language of RNA duplexes
hybrid nil unpaired_left_top closed
... h unpaired_left_top ult (base,empty) unpaired_left_top

unpaired_left_bot ... h unpaired_left_bot
ulb unpaired_left_bot
edangle ...
h edangle eds closed edt closed edb closed ... h
6
The language of RNA duplexes
closed stacking_region
bulge_top
bulge_bottom internal_loop end
_loop ... h stacking_region sr basepair closed bulge_top (bt basepair tt (uregion, empty)) topbound
closed bulge_bottom (bb (empty, uregion)) botbound closed
internal_loop (il (uregion,uregion)) symbound closed end_loop
el 7
Dynamic Programming recurrences
Time/memory complexity linear in target length
8
let-7/lin-41 binding sites
9
Requirements
For prediction of miRNA targets in large
databases we need

• A fast program ?
• Good statistics

10
Length normalisation of minimum free energies
11
p-values of individual binding sites
12
Poisson statistics of multiple binding sites
13
Comparative analysis of orthologous targets
14
Multi-species p-values
Poisson p-values
multi-species p-value
General case k species
15
A dependence problem
We should see a p-value as often as it says (blue
curve), but we dont (red curve).
16
let-7b/NME4 (human/mouse) binding sites
-GGCTCAAGCTGCCCTTACCACCCCATCCCCCACGCAGGACCAACTACCT
CCGTCAGCAAGAACCCAAGCCCACATCCAAACCTGCCTGTCCCAAACCAC
GGGCTTGCACTGCCTTCTGCACTTCAGGTCT-ACCCATGACCTACTACC
TCTGTCAACAAGAAGTCAAGCCCCCATGC---TTCCCATGTCCCCAAAC-
-



TTACTTCCCTGTTCACCTCTGCCCCACCCCAGCCCA
GAGGAGTTTGAGCCACCAACTTCAGTGCCTTTCTGTACCCCAAGCCAGCA
CAAGATTGGACCAA -CACTCCCTACTCCCGCTCTACCCAACTCCAGCCC
AGGGGAGTCTAAGCCTCAACTCTATGTGCCTTTTTGTATCCTAAGTCAAT
ACAATATTGGACCAT

TCCTTTTTGCACCAAAGTGCCGGACA
ACCTTTGTGGTGGGGGGGGGTCTTCACATTATCATAACCTCTCCTCTAAA
GGGGAGGCATTAAAATTCACTGTG GTCCTTGTGTACAAAAGTGCCAGAC
AACCTTTG--------GGGCATTGTCA-AAGGTGACTTCACCTGCCTCAA
AGGAGAGACATTAAAATTT--TATG

CCCAGCACATGGGTGGTACAC
TAATTATGACTTCCCCCAGCTCTGAGGTAGAAATGACGCCTTTATGCAAG
TTGTAAGGAGTTGAACAGTAAAGAGGAAG CTTAAAAT------------
--------------------------------------------------
------------------------------


k 1.1 is the effective k
17
Effective number of orthologous targets
18
Requirements
For prediction of miRNA targets in large
databases we need

• A fast program ?
• Good statistics ?

19
True and false positives and negatives
Positives Negatives

20
Sensitivity and specificity
Spec
p-values control specificity
21
Target prediction workflow
target db
multi-species p-values
22
Prediction of Drosophila miRNA targets

• 78 miRNAs
• 28,645 3UTRs (1/3 from D. mel, 1/3 from D.
  pseu, 1/3 from A. gamb)

23
Bantam hits
24
miR-7 hits
25
miR-2 hits
plus a number of others
26
RNAhybrid functionality
comparative analysis
effective k
27
miRNA target selection
p-values indicate not only biochemical
possibility, but also biological function.
28
Acknowledgements

• Peter Steffen, Robert Giegerich, Jan Krüger
• Matthias Höchsmann
• Alexander Stark, Julius Brennecke, Stephen M.
  Cohen
• Sven Rahmann
• Gregor Obernosterer
• Robert Heinen
• Leonie Ringrose

29
References
Rehmsmeier M, Steffen P, Höchsmann M and
Giegerich R. Fast and effective prediction of
microRNA/target duplexes. RNA, 101507-1517, 2004.
bibiserv.techfak.uni-bielefeld.de/rnahybrid