Index of /SISTA/aerts/temp

Hele tekst

(1)Integrative CRM prediction approaches uncover new direct target genes of retinal determination factors in fly and of cancer-related transcription factors in human. Bram van de Sande, Marina Naval Sanchez, Delphine Potier, Rekin's Janky, Zeynep Kalender, Annelien Verfaillie, Valerie Christiaens, Lotte Haagen, Carl Herrmann$, and Stein Aerts Laboratory of Computational Biology, Department for Human Genetics, University of Leuven, Belgium. i-cisTarget: motifs and in vivo Tracks. de la Méditerranée & TAGC – Inserm U928, France. Web application: example Zelda LOF data. Motifs 6383 Position Weight Matrices (PWM).

(2) ' (.

(3)

(4) . $ Université. &. .

(5) . *

(6) . '( $%& . JASPAR, TRANSFAC, FlyFactorSurvey, Tiffin Database, O. Elemento et al. conserved motifs, Stark et al. conserved motifs, SelexConsensus, YeTFaSCo, hDPI. ,. $%& .

(7) . .

(8) . ')( $%& !!! " " $ " " . In vivo tracks.

(9) .

(10) +.

(11) . . .

(12) . !!! # . !" #$%! . # $ # # . . . . 109 TF ChIP tracks modEncode (42), BDTNP (41), E. Furlong lab (26). 216 non-TF ChIP & DHS tracks. ! ! !. . . .

(13) . . modEncode (208), BDTNP (7), E. Furlong lab (1). 211 histone modification tracks.

(14) .

(15) . modEncode (205), BDTNP (0), E. Furlong lab (6).

(16) . Validation on ChIP peak sets and co-expressed gene sets. & % $. . $. &. The scatterplots on the left show that in vivo tracks (IVT) are found enriched in truly bound regions, while not in unbound regions. The bound regions are derived from ChIP peaks, and the unbound regions are genomic regions with high-scoring PWM matches for the TF under study, but without a ChIP peak.. '. . $%. #$%&'. %. . /21. ( ") *# %'%'%-".%&. + . . . ' &. . %. &. '. $'. . $*. . . /"<=== & -:- %. : 6 . . 2 ). .*) .*$. %*. ) /%. 6 . . % 0%. +. 1./. ) %* 4-. 1%. %- . !. . ,* . ). . & . %. . + . +. . 9:%+ 9:%+ 9: /%:

(17) : : '4/: 6: 2 %:798 &9: /%:

(18) : : '4/: 6: -% : 7/6: / %: -%6: 8 9: :

(19) ;'4/:

(20) %: : :/::%:

(21) : :-: 9: : : -:

(22) : % :%+:-+:)5$. 5. . $#('' 6. .*) . %. 0%. 1 . ,. %

(23) & '"

(24) & . %. 0 . ,. )+. +. ,. -. .. /.

(25) ##). 1 /0. -. . . !"#$%# $ & #'(. . 20 -. )* $( ,. .. 1. 20. . . . . . . . Combinations of motifs and In Vivo Tracks improve the accuracy of CRM predictions (TPR and PPV). In this example we analyze a set of downregulated genes in the Zelda LOF (Liang et al.), and compare the predicted CRMs with Zelda ChIP peaks. In this example, the ChIP peaks are used as true positives, and not used as iVT.. 2-. 0.

(26)

(27)

(28). %. %((. . . )#%#* +, #%#* *-. /'0. 0. 0. . . 2. %

(29) & '"

(30) & 2,. 2.. 21.

(31) . -. Comparative Transcriptomics + Comparative motif discovery Validation of the predicted Glass targetome using RNA-Seq in homozygous glass mutant discs. Validation of predicted Glass target CRMs using in vivo enhancer-reporters. . . #

(32) . 6 . 1

(33) . 2 ). 5. 20.

(34)

(35)

(36) . . %*. . 1% 1./. +. /. . ). . !. %. .*$. 2. Derive a conserved set of eye-specific genes - “species as replicates”. . 4-. 1. . $#(''. ,*. +. . )* $* . Tag-Sequencing in two tissues and three species. . . 1

(37) . /. +. . #

(38) . 1. / 2)3&.

(39) %-. 6. + ) . %-. 2,. . .

(40)

(41)

(42) . The scatterplots on the right show the correlation between the enrichment score for in vivo tracks and motifs, obtained by analyzing a gene set (e.g., Zelda LOF) and a related ChIP peak set (e.g., Zelda ChIP). Agreement between these indicates that the features we discover on the gene set are indicative of true regulatory regions.. Application to 628 co-expressed gene sets 78. .. ! " !. "

(43) " #

(44) ". *. $. )*+. ,. + )#%#* ,% /10. . .

(45)

(46)

(47)

(48)

(49)

(50) . " *# %'%'%-".%! " $ "

(51) " #

(52) " . )+. !"#$

(53) . . '.

(54) . $%& /0. !"#$

(55) . . .

(56)

(57)

(58) . ! " . ()*+. In the heatmaps, a red square indicates that the expected motif (on the right) was found among the top 3 motifs. A pink square indicates that the expected motif was found to be significantly enriched. A black square indicates that the expected motif was not found enriched.. ( ") . *# %'%'%-".%! ". (,-$%. /(1. , . .

(59).

(60) !. ". " #$%& "'&(. #. . /01. 2)3&. ). & /.

(61) %-. +. %+:% '% ::% '> 1 :%. '4/. 7"8. /%. 1 :%-. 2+:%-. .?#. /:. '4/:@:.?#. /6+. /"<===== A:% %: . /:% ::%+- %: 6:%+. %*. / 5 %:+. 7$8. /"<====A:$% &!. B%&. $!!.

(62) !". . . '% . #

(63) . . % 6 B. #

(64) . .

(65) . .

(66). %.

(67) . . . Top 245 conserved eye-specific genes => cisTargetX (cutoff determined by GO “photoreceptor differentiation e-12”). % . ()' . . 78 . .

(68) . . . . . . *" . Meta-Targetomes obtained from 2211 cancer gene signatures. cisTargetNET for Human and Mouse. NFkB CTDB ranking. PPARA activation; 698 up-reg. genes. vs.. 104 known NFkB-targets. TP53 CTDB ranking. vs.. 176 known TP53-targets. Rakhshandehroo et al., 2009. NES=2.0 FDR < 0.00001. NES=2.28 FDR < 0.00001 64 in top 1429. 1st motif (NES=6.31). Validated (13/19) (incl. AMADEUS Benchmark) PPAR, TP53, NFkB, LEF1, E2F, ETS1, HNF1A, HSF1, NFY, NRF1, OCT4, SRF, YY1. MiR. microRNA targets. 107 predicted direct targets. TP53 activation; 668 up-reg. genes LCB, MCF7+Nutlin+RNA-Seq, unpublished. 1st motif (NES=7.59). Meta-Targetomes: cisTargetNET was applied to 2211 cancer gene signatures. For each significant motif and linked TF, the number of target gene predictions across gene signatures are counted. The meta-targetome is the ranking of predicted targets based on the recurrence of finding the same target across gene signatures.. 3rd motif (NES=4.56) On 159 miR target sets: 61 result in significant enrichment of TF regulators. STRING network TP53: 500 associations. Availability: http://med.kuleuven.be/lcb/i-cisTarget. Cytoscape Plugin.

(69)

No results found