current microrna target identification methods are heavily ... · microrna target identi cation...
TRANSCRIPT
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Current microRNA target identificationmethods are heavily contaminated with false
positives
Hervé Seitz
IGH (UMR 9002 CNRS and university of Montpellier), Montpellier, France
November 8, 2018
This slideshow is available at:http://www.igh.cnrs.fr/equip/Seitz/Sifr2018.pdf
http://www.igh.cnrs.fr/equip/Seitz/Sifr2018.pdf
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
miRNA sequences are known (for the most part); can weidentify their targets ?
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
Target geneTranscription mRNA Translation
Protein
mRNAdecay
Translationalrepression
miRNA
miRNA sequences are known (for the most part); can weidentify their targets ?
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
Ago protein+ miRNA
Target geneTranscription mRNA Translation
Protein
mRNAdecay
Translationalrepression
miRNA sequences are known (for the most part); can weidentify their targets ?
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
Ago protein+ miRNA
Target geneTranscription mRNA Translation
Protein
mRNAdecay
Translationalrepression
miRNA sequences are known (for the most part); can weidentify their targets ?
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
I Sequence complementarity to the miRNA.
I Physically bind the miRNA effector complex.
I mRNA and protein level respond to miRNA.
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
Stringent
Not stringent
(illustration by D. Goodsell, Scripps Research Institute andRutgers University)
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
Stringent
Not stringent
(illustration by D. Goodsell, Scripps Research Institute andRutgers University)
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
Comparative genomics dogma:if it is conserved, it must be functional.
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
Comparative genomics dogma:if it is conserved, it must be functional.
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
Comparative genomics dogma:if it is conserved, it must be functional.
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
.
||||||NNNNNNN NNNNNNNNNNNNNN5´ 3´miRNA:
target:
2 7
−→ the “seed”
Comparative genomics (if a seed match is conserved, it mustbe functional).
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
.
||||||NNNNNNN NNNNNNNNNNNNNN5´ 3´miRNA:
target:
2 7
−→ the “seed”
Comparative genomics (if a seed match is conserved, it mustbe functional).
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
.
||||||NNNNNNN NNNNNNNNNNNNNN5´ 3´miRNA:
target:
2 7
−→ the “seed”
Comparative genomics (if a seed match is conserved, it mustbe functional).
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
I Experimental methods (CLIP and avatars): hundreds oftargets per miRNA.
I Computational methods (seed matches, conserved inevolution): hundreds of targets per miRNA.
miRNA-mediated repression is very modest (usually< 2-fold): lower than well tolerated fluctuations in geneexpression (e.g., haplosufficiency). Why have these sitesbeen conserved if they are not functional?
CLIP and avatars
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
I Experimental methods (CLIP and avatars): hundreds oftargets per miRNA.
I Computational methods (seed matches, conserved inevolution): hundreds of targets per miRNA.
miRNA-mediated repression is very modest (usually< 2-fold): lower than well tolerated fluctuations in geneexpression (e.g., haplosufficiency). Why have these sitesbeen conserved if they are not functional?
CLIP and avatars
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
I Experimental methods (CLIP and avatars): hundreds oftargets per miRNA.
I Computational methods (seed matches, conserved inevolution): hundreds of targets per miRNA.
miRNA-mediated repression is very modest (usually< 2-fold): lower than well tolerated fluctuations in geneexpression (e.g., haplosufficiency).
Why have these sitesbeen conserved if they are not functional?
CLIP and avatars
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
I Experimental methods (CLIP and avatars): hundreds oftargets per miRNA.
I Computational methods (seed matches, conserved inevolution): hundreds of targets per miRNA.
miRNA-mediated repression is very modest (usually< 2-fold): lower than well tolerated fluctuations in geneexpression (e.g., haplosufficiency). Why have these sitesbeen conserved if they are not functional?
CLIP and avatars
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Results
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Robustness of predicted targets
Baek et al., 2008: quantification of miR-223-mediatedrepression in mouse neutrophils.
Neutrophil isolation
Blood collection
RNA extraction
cDNA labeling,
array hybridization
Blood collection
Pooled blood
Split in 5 replicates
Neutrophil isolation
RNA extraction
cDNA labeling,
array hybridization
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Robustness of predicted targets
Baek et al., 2008: quantification of miR-223-mediatedrepression in mouse neutrophils.
Neutrophil isolation
Blood collection
RNA extraction
cDNA labeling,
array hybridization
Blood collection
Pooled blood
Split in 5 replicates
Neutrophil isolation
RNA extraction
cDNA labeling,
array hybridization
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Robustness of predicted targets
Baek et al., 2008: quantification of miR-223-mediatedrepression in mouse neutrophils.
Neutrophil isolation
Blood collection
RNA extraction
cDNA labeling,
array hybridization
Blood collection
Pooled blood
Split in 5 replicates
Neutrophil isolation
RNA extraction
cDNA labeling,
array hybridization
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Robustness of predicted targets
Baek et al., 2008: quantification of miR-223-mediatedrepression in mouse neutrophils.
Neutrophil isolation
Blood collection
RNA extraction
cDNA labeling,
array hybridization
Blood collection
Pooled blood
Split in 5 replicates
Neutrophil isolation
RNA extraction
cDNA labeling,
array hybridization
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Robustness of predicted targets
p: probability that the difference between two individualmice is smaller than miRNA-guided repression
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Robustness of predicted targets
p: probability that the difference between two individualmice is smaller than miRNA-guided repression
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Robustness of predicted targets
p: probability that the difference between two individualmice is smaller than miRNA-guided repression
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Robustness of predicted targets
p: probability that the difference between two individualmice is smaller than miRNA-guided repression
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Robustness of predicted targets
For 150 predicted targets out of 192:inter-individual fluctuations across 5 wild-type mice exceedsmiRNA-mediated regulation (p-value < 0.05).
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Robustness of predicted targets
For 150 predicted targets out of 192:inter-individual fluctuations across 5 wild-type mice exceedsmiRNA-mediated regulation (p-value < 0.05).
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Comparative genomics yield false positives
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Comparative genomics yield false positives
Taeniopygia guttata
Monodelphis domestica
Mus musculus
Homo sapiens
Danio rerio
Sus scrofa
Bos taurus
USP9X mRNA miR-134 site
Sarcophilus harrisii
Gallus gallus
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Comparative genomics yield false positives
:|| || |||||||
Taeniopygia guttata
Monodelphis domestica
Mus musculus
Homo sapiens
Danio rerio
Sus scrofa
Bos taurus
GACAUAACCAGCAAUGAAGA-UUUU-AGUCUCA
GGCAUGACACUG-AUUCAGG-AUUUCAGUCACA
GGCAUGAC----AAUUCAGG-CGUUCAGUCACA
GACAUGACCCUGAAUUCAGG-AGUUCAGUCACA
AGCAUGACACUGAAUUCAGGAAUUUCAGUCACA
AGCAUGACACUGAAUUCAGGAAUUUCAGUCACA
GGCACUAUGCUGAGUUCUGGAACAACAGUCACA
GGCACUAUGCUGAAUUCUGGAACAACAGUCACA
GAUGUGACGCUGAAUUCUGAAGCCACAGUCACA
USP9X mRNA miR-134 site
Sarcophilus harrisii
Gallus gallus
GGGGAGACCAGUUGGUCAGUGU5´3´
3´5´
miR-134:
: conserved in 8 or 9 species out of 9 : conserved in 6 or 7 species out of 9 : conserved in less than 6 species out of 9
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Comparative genomics yield false positives
:|| || |||||||
Taeniopygia guttata
Monodelphis domestica
Mus musculus
Homo sapiens
Danio rerio
miR-134
Sus scrofa
Bos taurus
GACAUAACCAGCAAUGAAGA-UUUU-AGUCUCA
GGCAUGACACUG-AUUCAGG-AUUUCAGUCACA
GGCAUGAC----AAUUCAGG-CGUUCAGUCACA
GACAUGACCCUGAAUUCAGG-AGUUCAGUCACA
AGCAUGACACUGAAUUCAGGAAUUUCAGUCACA
AGCAUGACACUGAAUUCAGGAAUUUCAGUCACA
GGCACUAUGCUGAGUUCUGGAACAACAGUCACA
GGCACUAUGCUGAAUUCUGGAACAACAGUCACA
GAUGUGACGCUGAAUUCUGAAGCCACAGUCACA
USP9X mRNA miR-134 site
Sarcophilus harrisii
Gallus gallus
GGGGAGACCAGUUGGUCAGUGU5´3´
3´5´
miR-134:
: conserved in 8 or 9 species out of 9 : conserved in 6 or 7 species out of 9 : conserved in less than 6 species out of 9
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Comparative genomics yield false positives
Hominidae
Catarrhini
Boreoeutheria
Euteleostomi
Comparison to prediction programs
Effect of tree architecture
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Comparative genomics yield false positives
Hominidae
Catarrhini
Boreoeutheria
Euteleostomi
3´ UTR seed matches:
Clade-specific miRNA families
Hominidae Catarrhini Boreo-eutheria
Euteleostomi
Prop
ortio
n of
ove
r-con
serv
ed s
eed
mat
ches
0.0
0.2
0.4
0.6
0.8
1.0(n=10 miRNA
families) (n=14 miRNAfamilies)
(n=14 miRNAfamilies)
(n=10 miRNAfamilies)
Comparison to prediction programs
Effect of tree architecture
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Conclusion
Most known miRNA “targets” are probably not functionallytargeted.
Observing a molecular does not mean that it has a biologicalfunction.
Homeostasis attenuates changes of external origin: it alsobuffers internal changes.
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Conclusion
Most known miRNA “targets” are probably not functionallytargeted.
Observing a molecular event does not mean that it has abiological function.
Homeostasis attenuates changes of external origin: it alsobuffers internal changes.
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Conclusion
Most known miRNA “targets” are probably not functionallytargeted.
Observing a molecular object does not mean that it has abiological function.
Homeostasis attenuates changes of external origin: it alsobuffers internal changes.
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Conclusion
Most known miRNA “targets” are probably not functionallytargeted.
Observing a molecular object does not mean that it has abiological function.
Homeostasis attenuates changes of external origin: it alsobuffers internal changes.
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Acknowledgements
I. Busseau A. SergeevaL. Martinez N. Pinzón
B. Li S. Mockly
Collaborators: F. Apparailly, J. Presumey (INM, Montpellier)
October 2019: CNRS � école thématique � on smallregulatory RNAs (miRNAs, siRNAs, piRNAs) in Sète.
This slideshow is available at:http://www.igh.cnrs.fr/equip/Seitz/Sifr2018.pdf
http://www.igh.cnrs.fr/equip/Seitz/Sifr2018.pdf
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Acknowledgements
I. Busseau A. SergeevaL. Martinez N. Pinzón
B. Li S. Mockly
Collaborators: F. Apparailly, J. Presumey (INM, Montpellier)
October 2019: CNRS � école thématique � on smallregulatory RNAs (miRNAs, siRNAs, piRNAs) in Sète.
This slideshow is available at:http://www.igh.cnrs.fr/equip/Seitz/Sifr2018.pdf
http://www.igh.cnrs.fr/equip/Seitz/Sifr2018.pdf
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Supplementary data
CLIP and avatars
Overconservation statistics
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Introduction
Reverse transcription,amplification, sequencing
mRNA
Tag clusters
Partial RNA digestion
Immuno-precipitation
Ago
target RNA
miRNA
U.V. crosslinking
Gel purification,proteolysis
Adapter ligation,gel purification
Return
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Over-conservation in prediction programs
Predictions by TargetScan 7.0:
Clade-specific miRNA families
Hominidae Catarrhini Boreo-eutheria
EuteleostomiProp
ortio
n of
ove
r-con
serv
ed s
eed
mat
ches
0.0
0.2
0.4
0.6
0.8
1.0Predictions by miRanda (aug2010):
Clade-specific miRNA families
Hominidae Catarrhini Boreo-eutheria
EuteleostomiProp
ortio
n of
ove
r-con
serv
ed s
eed
mat
ches
0.0
0.2
0.4
0.6
0.8
1.0
Predictions by PicTar2:
Clade-specific miRNA families
Prop
ortio
n of
ove
r-con
serv
ed s
eed
mat
ches
Hominidae Catarrhini Boreo-eutheria
Euteleostomi0.0
0.2
0.4
0.6
0.8
1.0
Predictions by microT 5.0:
Clade-specific miRNA families
Hominidae Catarrhini Boreo-eutheria
EuteleostomiProp
ortio
n of
ove
r-con
serv
ed s
eed
mat
ches
0.0
0.2
0.4
0.6
0.8
1.0
(conservationnot considered)
Return
-
False positives inmicroRNA target
identification
Introduction
Results
Robustness of predictedtargets
Comparative genomics yieldfalse positives
Conclusion
Acknowledgements
Supplementarydata
Effect of tree architecture
Hominidae Catarrhini Boreoeutheria Euteleostomi
Clade of interest
0.0
0.2
0.4
0.6
0.8
1.0
Prop
ortio
n of
see
d m
atch
es c
onse
rved
outs
ide
clad
e of
inte
rest
Hominidae-specific seedsCatarrhini-specific seedsBoreoeutheria-specific seedsEuteleostomi-specific seeds
Non-seed hexamers3´ UTR seed matches:
Return
IntroductionResultsRobustness of predicted targetsComparative genomics yield false positives
ConclusionAcknowledgementsSupplementary data
anm0: 0.14: 0.13: 0.12: 0.11: 0.10: 0.9: 0.8: 0.7: 0.6: 0.5: 0.4: 0.3: 0.2: 0.1: 0.0: