searching for functional regions (coding or non-coding) in mammalian genomes

61
Searching for functional regions (coding Searching for functional regions (coding or non-coding) in mammalian genomes or non-coding) in mammalian genomes Human genome sequence: 1300 Mb (38 %) available in GenBank (November 22 1999) Draft sequence 90% complete in 2000 Finished sequence in 2003 Experimental characterization of all human genes. How many years ? Organization of the human genome Prediction of functional elements by computer analysis of genomic sequences State of the art Success and pitfalls of different approaches Prediction of function by homology Orthology/paralogy

Upload: ania

Post on 12-Jan-2016

30 views

Category:

Documents


0 download

DESCRIPTION

Searching for functional regions (coding or non-coding) in mammalian genomes. Human genome sequence: 1300 Mb (38 %) available in GenBank (November 22 1999) Draft sequence 90% complete in 2000 Finished sequence in 2003 Experimental characterization of all human genes. How many years ? - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Searching for functional regions (coding or non-coding) in mammalian genomes

Searching for functional regions (coding or non-coding) Searching for functional regions (coding or non-coding) in mammalian genomes in mammalian genomes

Human genome sequence: 1300 Mb (38 %) available in GenBank (November 22 1999) Draft sequence 90% complete in 2000 Finished sequence in 2003 Experimental characterization of all human genes. How many years ?

Organization of the human genome Prediction of functional elements by computer analysis of genomic sequences

State of the art Success and pitfalls of different approaches

Prediction of function by homology Orthology/paralogy

Page 2: Searching for functional regions (coding or non-coding) in mammalian genomes

Functional elements in the human genomeFunctional elements in the human genome

3.4 109 nt 50,000-100,000 protein-coding genes

81% no known function43%38%introns4%12%protein-coding regions

centromeres, telomeres,

RNA2%intergenic

Untranslated RNAs: Xist, H19, His-1, bic, etc.

Regulatory elements: promoters, enhancers, etc.

Repeated sequences (SINES, LINES, HERV, etc.) : 40% of the human genome

Page 3: Searching for functional regions (coding or non-coding) in mammalian genomes

Structure of human protein genesStructure of human protein genes

1396 complete human genes (exons + introns) from GenBank Average size (25%, 75%)

Gene 15 kb ± 23 kb (4, 16) (10% > 35 kb) CDS 1300 nt ± 1200 (600, 1500) Exon (coding) 200 nt ± 180 (110, 200) Intron 1800 nt ± 3000 (500, 2000) 5'UTR 210 nt (Pesole et al. 1999) 3'UTR 740 nt (Pesole et al. 1999)

Intron/exon Number of introns: 6 ±3 introns / kb CDS Introns / (introns + CDS): 80% 5' introns in 15% of genes (more ?), 3 ’introns very rare

Alternative splicing in more than 30% of human genes (Hanke et al. 1999)

Page 4: Searching for functional regions (coding or non-coding) in mammalian genomes

Structure of human protein genesStructure of human protein genes GenBank: bias towards short genes 1396 complete human genes (exons + introns)

≤949596979899Publication date48121620Gene size (coding exons+introns) kb

5101520253035≤949596979899Publication dateGene size (coding exons+introns) kb

Page 5: Searching for functional regions (coding or non-coding) in mammalian genomes

Structure of human protein genesStructure of human protein genes GenBank: bias towards short genes 1396 complete human genes (exons + introns) 9268 complete human mRNA

Sequence:cDNA

complete gene (exons+introns)

400800120016002000889092949698Average CDS size (nt)Publication date

Page 6: Searching for functional regions (coding or non-coding) in mammalian genomes

Isochore organization of the human genomeIsochore organization of the human genome

Insertion of repeated sequences (A. Smit 1996) Recombination frequency (Eyre-Walker 1993) Chromosome banding (Saccone, 1993) Replication timing (Bernardi, 1998) Gene density (Mouchiroud, 1991) Gene expression ?? -> No Gene structure (Duret, 1995)

isochore %C+G % total genomic DNA

L1+L2 : 33%-44% 62 %

H1+H2 : 44%-51% 31%

H3 : 51%-60% 3-5%

H1+H2L1+L2H3H1+H2L1+L2L1+L2>300 kbBernardi et al. 1985

Page 7: Searching for functional regions (coding or non-coding) in mammalian genomes

Isochores and insertion of repeat sequencesIsochores and insertion of repeat sequences

4%8%12%16%20%AluLINE-1LTR-

elements

Density in repeat sequencesG+C content of genomic sequence:G+C < 39%G+C > 47%G+C 39%-47%

4419 human genomic sequences > 50 kb4419 human genomic sequences > 50 kb

Page 8: Searching for functional regions (coding or non-coding) in mammalian genomes

Isochores and gene densityIsochores and gene density

MHC locus (3.6 Mb) MHC locus (3.6 Mb) (The MHC sequencing consortium 1999)(The MHC sequencing consortium 1999)

Class I, class II (H1-H2 isochores): 20 genes/Mb, many pseudogenesClass I, class II (H1-H2 isochores): 20 genes/Mb, many pseudogenesClass III (H3 isochore): 84 genes/Mb, no pseudogeneClass III (H3 isochore): 84 genes/Mb, no pseudogene

Class II boundaries correlate with switching of replication timingClass II boundaries correlate with switching of replication timing

isochore % total genomic DNA %total genes

L1+L2 : 62 % 31%

H1+H2 : 31% 39%

H3 : 3-5% 30%

2060100140Number of genes / MbL1+L2H1+H2H3Mouchiroud et al. 1991

Page 9: Searching for functional regions (coding or non-coding) in mammalian genomes

Isochores and introns lengthIsochores and introns length

760 complete human genes L1L2: intron G+C content < 46% H1H2: intron G+C content 46-54% H3: intron G+C content >54%

Average intron length (bp)Gene compaction (intron length/coding region length)40080012001600200024681012L1L2H1H2H3L1L2H1H2H3

Duret, Mouchiroud and Gautier, 1995

Page 10: Searching for functional regions (coding or non-coding) in mammalian genomes

Prediction of functional elements (1)Prediction of functional elements (1) Ab initio methods

Ruled-based or statistical methods e.g.: protein genes prediction, promoter prediction, … Very useful but ...

Limits in sensibility/specificity No method available for many functional elements (non-coding RNA genes, regulatory elements, …)

Large scale transcriptome projects: ESTs, full-length cDNA Identification of transcribed genes (protein or non-coding RNA) Information on alternative splicing, polyadenylation (Hanke et al. 1999, Gautheret et al. 1998),

expression pattern Very useful but ...

Problems with genes expressed at low level, narrow tissue distribution, stage-specific expression, … Limited tissue sampling Artifacts in ESTs (introns, partially matured RNA, …) Limited to polyadenylated RNA

Page 11: Searching for functional regions (coding or non-coding) in mammalian genomes

Prediction of functional elements (2)Prediction of functional elements (2) Comparative sequence analysis (phylogenetic footprinting)

Function => selective pressure

Corollary Sequence conservation = selective pressure = function

provided the number of aligned homologous sequences represents enough evolutionary time for the accumulation of mutations at the less constrained (presumably selectively neutral)

base positions.

Evolutionary rate in non-functional DNA: ~ 0.3% / My (± 0.069)

Man/Mouse: ~ 80 Myrs 46-58% identity

Mammals/Birds: ~ 300 Myr 26-28% identity

Random sequences 25% identity

Page 12: Searching for functional regions (coding or non-coding) in mammalian genomes

Analyse comparative des gènes de Analyse comparative des gènes de -actine de l'homme et de la carpe-actine de l'homme et de la carpe

CarpeHomme5’UTR 3’UTR site polyA échelle de similarité: pas de similarité significative70 - 80% identité80 - 90% identitérégions codantes: éléments régulateurs:introns:ATGcodon stop

Page 13: Searching for functional regions (coding or non-coding) in mammalian genomes

Phylogenetic footprintingPhylogenetic footprinting Advantages

Works for all kinds of functional elements (transcribed or not, coding or not) as far as the information is in the primary sequence

Does not require any a priori knowledge of the functional elements

Limits Absence of evolutionary conservation does not mean absence of function No efficient method to detect unknown conserved secondary structure in RNA Function, but what function ? Depends on the sequencing status of other genomes

Human, mouse, fugu, C. elegans, drosophila, yeast, A. thaliana Number of sequences to compare : > 200 Myrs of evolution

Mammals/birds: 310 Myrs Human + mouse + bovine : 240 Myrs

Page 14: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction de fonction par homologie ?Prédiction de fonction par homologie ? Similarité entre séquences homologie Homologie structure conservée Structure conservée fonction conservée

Oui, mais … Fonction: concept flou

– activité biochimique identique ? e.g. même ligand pour un récepteur, même substrat pour une enzyme, même gènes cibles pour un facteur de transcription.

– distribution tissulaire ? (isoformes tissu-spécifiques).– compartimentalisation cellulaire: cytoplasme, mitochondrie, etc.

Protéines homologues de fonction différentes – Protéines homologues ligands (activateur/répresseur) d ’un même récepteur– Recrutement pour une fonction totalement différente: -cristalline / -énolase

Orthologie/paralogie

Évolution modulaire

Page 15: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction de fonction par homologie ?Prédiction de fonction par homologie ?

MZEORFG: 1 ILNSPDRACNLAKQAFDEAISELDSLGEESYKDSTLIMQLLXDNLTLWTSDTNEDGGDE 59

I N+P++AC LAKQAFD+AI+ELD+L E+SYKDSTLIMQLL DNLTLWTSD ++ E

BOV1433P: 186 IQNAPEQACLLAKQAFDDAIAELDTLNEDSYKDSTLIMQLLRDNLTLWTSDQQDEEAGE 244

Score = 87.4 bits (213), Expect = 1e-17

Identities = 41/59 (69%), Positives = 50/59 (84%)

LOCUS BOV1433P 1696 bp mRNA MAM 26-APR-1993

DEFINITION Bovine brain-specific 14-3-3 protein eta chain mRNA, complete cds.

ACCESSION J03868

LOCUS MZEORFG 187 bp mRNA PLN 31-MAY-1994

DEFINITION Zea mays putative brain specific 14-3-3 protein, tau protein

homolog mRNA, partial cds.

Page 16: Searching for functional regions (coding or non-coding) in mammalian genomes

Orthologie/paralogieOrthologie/paralogiespéciationduplicationPrimatesRongeursHommeRatGène ancestral

de l’insulineSourisRatSourisINSINS1INS1INS1INS2INS2INS2Homologie: deux gènes sont homologues si ils ont un ancêtre commun

Orthologie: deux gènes sont orthologues si ils ont divergé à la suite d’un évènement de spéciation

Paralogie: deux gènes sont paralogues si ils ont divergé à la suite d’un évènement de duplication

Orthologie ≠ équivalence fonctionnelle

!

Page 17: Searching for functional regions (coding or non-coding) in mammalian genomes

Diversification fonctionnelle par duplication de gènes au cours de l'évolution des métazoaires

Iwabe et al. MBE 13:483-493 (1996):

Analyse phylogénétique de 25 familles de gènes

isoformes compartimentalisées: duplications avant la divergence animaux/champignons

paralogues avec activités différentes: duplication avant la divergence vertébrés/arthropode

isoformes tissu-spécifiques: duplication avant la divergence poissons/tétrapodes

Page 18: Searching for functional regions (coding or non-coding) in mammalian genomes

Approche phylogénétique pour la prédiction de fonction

1) Identifier les homologues

2) Aligner les séquences

3) Calculer l’arbre phylogénétique

2A3A1A1B2B3B2A3A1A1B2B3B2A3A1A1B2B3B2A3A1A1B2B3B2ADuplication de gènes4) Placer les fonctions connues sur l’arbre

5) Inférer la fonction probable des gènes

Page 19: Searching for functional regions (coding or non-coding) in mammalian genomes

Orthologie/paralogie: abus de langageOrthologie/paralogie: abus de langage

Fitch (1970) Syst. Zool. 19:99-113:

"Where the homology is the result of gene duplication [...] the genes should be called paralogous. Where the homology is the result of speciation [...] the genes should be called orthologous."

Koonin (1996) TIG, PNAS,Curr. Opin. Genet. Dev.

"By definition, orthologs are genes that are related by vertical descent from a common ancestor and encode proteins with the same function in different species. By contrast, paralogs are homologous genes that have evolved by duplication and code for proteins with similar, but not identical functions".

Corrigé dans Science 1997

Page 20: Searching for functional regions (coding or non-coding) in mammalian genomes

Evolution modulaire

ABC

Page 21: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction de régions régulatricesPrédiction de régions régulatrices

Méthodes ab initio

Prédiction de promoteurs Îlots CpG

Approche comparative

Page 22: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction de promoteurs eucaryotesPrédiction de promoteurs eucaryotes

Combinaison de sites de fixation de facteur de transcription (ordre, orientation, distance)

Motifs courts, dégénérés Difficile de distinguer les vrais sites des faux positifs: Motif à 4 bases: ≈1/256 pb (1/128 pb sur les deux brins)

Boîtes TATA, CAAT , GC: absents dans beaucoup de promoteurs

Banques de données de sites de fixation de facteurs de transcription (TRANSFAC), de promoteurs caractérisés expérimentalement (EPD)

PromoterScan (Prestridge 1995): Mesure de la densité en sites potentiels de fixation de facteurs de transcription de long de la séquence (pondération en fonction de la fréquence des sites dans ou en dehors des vrais promoteurs)

Page 23: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction de promoteurs: sensibilité, spécificitéPrédiction de promoteurs: sensibilité, spécificité

Sensibilité: fraction des promoteurs qui sont trouvés par le logiciel

PromoterScan: sensibilité = 70% (promoteurs à boîte TATA) Spécificité: fraction des vrais promoteurs parmi ceux qui ont été prédits

PromoterScan: spécificité = 20% Un faux positif / 10 kb

Génome humain: ≈100 000 gènes, ≈1 promoteur/30 kb

sensibilité=vrais_ positifs

vrais_ positifs+faux_ négatifs

spécificité=vrais_ positifs

vrais_ positifs+faux_ positifs

SpécificitéSensibilité1000Seuil

Page 24: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction de promoteurs eucaryotes: Prédiction de promoteurs eucaryotes: recherches en coursrecherches en cours

Prise en compte de l'orientation relative et des distances entre sites de fixation de facteurs de transcription COMPEL (Kolchanov 1998): banque de données d'éléments composites FastM : recherche dans une séquence génomique d'une combinaison de deux sites

de fixation de facteurs de transcription à une distance définie l'un de l'autre

Recherche de corrélations entre sites Prospector (Werner 2000)

– Sensibilité: 50%– Spécificité: 80%

http://www.gsf.de/biodv/index.html

Combinaison recherche ab initio / approche comparative: recherche de sites potentiels parmi les régions conservées

Page 25: Searching for functional regions (coding or non-coding) in mammalian genomes

Îlots CpGÎlots CpG Génome de vertébrés :

méthylation des C dans les dinucléotides 5 ’-CG-3 ’(CpG) Me-C fortement mutable -> T

5 ’-CG- 3 ’ 5 ’-TG-3 ’ 5 ’-CA-3 ’

3 ’-GC- 5 ’ 3 ’-AC-5 ’ 3 ’-GT-5 ’

Génome des vertébrés: globalement dépourvu en CpG (excès de TG, CA)

Certaines régions (200 nt à plusieurs kb) échappent à la méthylation Pas de déplétion en CpG: CpGo/e proche de 1 Riche en G+C Îlot CpG:

Longueur > 500 nt

CpGo/e > 0.6

G+C > 50%

ouou

CpGo /e =Nombre_ de_ CpG_ observéNombre_ de_CpG_ attendu

=0.25

01CpGo/e

Page 26: Searching for functional regions (coding or non-coding) in mammalian genomes

Îlots CpG: associé aux régions promotrices ?Îlots CpG: associé aux régions promotrices ?

Bird (1986), Gardiner-Garden (1987) Larsen (1992) ref 40% des gènes tissu-spécifiques possèdent un îlot CpG en 5 ’ 100% des gènes ‘ housekeeping ’ possèdent un îlot CpG en 5 ’

Rechercher des îlots CpG pour prédire des régions promotrices ? Sensibilité: 40-100% Spécificité ?? (Quelle fraction des îlots CpG correspond effectivement à des

régions promotrices ?)

Ponger (1999): comparaison des îlot CpG qui recouvre ou non le site d ’initiation de la transcription

Page 27: Searching for functional regions (coding or non-coding) in mammalian genomes

Fréquence des gènes humains avec un îlot CpG Fréquence des gènes humains avec un îlot CpG recouvrant le site d ’initiation de la transcriptionrecouvrant le site d ’initiation de la transcription

800 gènes humains avec promoteur décrit Mesure de la distribution tissulaire à l ’aide d ’EST (20 tissus)

0%20%40%60%80%0-23-67-20Nombre de tissus où le gène est exprimé

Page 28: Searching for functional regions (coding or non-coding) in mammalian genomes

Comparaison des îlots CpG recouvrant ou non le site Comparaison des îlots CpG recouvrant ou non le site d ’initiation de la transcriptiond ’initiation de la transcription

272 îlots start CpG recouvrant le site d ’initiation de la transcription

1078 îlots CpG en dehors d ’un promoteur connu

0 2460.50.70.91.11.350%60%70%80%otherstartG+C%CpGo/eLongueur(kb)

Page 29: Searching for functional regions (coding or non-coding) in mammalian genomes

Recherche de régions régulatrices par analyse Recherche de régions régulatrices par analyse comparative (empreintes phylogénétiques)comparative (empreintes phylogénétiques)

Goodman et al. 1988: régulation de l’expression des gènes du cluster -globine au cours du développement

– Alignement de séquences orthologues de 6 mammifères (> 270 Ma d’évolution)

– 13 empreintes phylogénétiques: ≥ 6 nt, conservation 100%– Analyse par retard de bande sur gel: – 12/13 (92%) correspondent à des sites de fixation de protéines

1996: 35 empreintes phylogénétiques avec protéines fixatrices identifiées

Enhancers de gènes HOX (Fugu/souris) (Aparicio et al. 1995)

enhancer TCR (homme/souris) (Luo, 1998)

promoteur COX5B (11 primates) (Bachman, 1996)

promoteur uPAR (homme/souris) (Soravia, 1995)

Page 30: Searching for functional regions (coding or non-coding) in mammalian genomes

Large scale phylogenetic Large scale phylogenetic footprintingfootprinting

Non-coding sequences : 325,247 sequences 145 Mb

everything except protein-coding regions and structural RNA genes (rRNA, tRNA, snRNA, scRNA)

Introns, 5' and 3' untranslated regions, intergenic sequences

Filtering of microsatellite repeats and cloning vectors: XBLAST

Similarity search: BLASTN + LFASTA

Vertebrates, insects, nematode

Page 31: Searching for functional regions (coding or non-coding) in mammalian genomes

Metazoan Genome ProjectsMetazoan Genome ProjectsMillion yearsPorifera (sponge)Nematodes (C. elegans)Arthropods (Drosophila)EchinodermsUrochordataCephalochordata (amphioxus)Jawless fisheschondrichthyes (ray, shark)actinopterygii (bony fishes)amphibians mammals birds reptiles600400200800VertebratesSequencing effort: 9 to 100 Mb 0.8 to 2.4 Mb less than 0.2 Mb

Page 32: Searching for functional regions (coding or non-coding) in mammalian genomes

Sequence SimilaritiesSequence Similarities1- Identification of new genes

protein-genes, RNA-genes: intronic snoRNA genes

2- Retroviral elements, retrotransposons

3- Low complexity sequences:

GC-rich, AT-rich, cryptic microsatellites

4- Artefacts:

annotation errors, sample contamination (sponge insulin, ascidian RNA, chicken TGFB1)

5- 326 highly conserved regions (HCRs)

- do not code for proteins

- do not correspond to any known structural RNA

Page 33: Searching for functional regions (coding or non-coding) in mammalian genomes

326 Highly Conserved 326 Highly Conserved Regions (HCRs)Regions (HCRs)

• > 70% identity over 50 to 2000 nt after more than 300 Myrs

• Unique sequences

• Generally specific of only one gene

• Longest HCR:

84% identity over 1930 nt after 300 Myrs

3’UTR deltaEF1 transcription factor

• Oldest HCRs: 500 to 600 Myrs

• No HCR between vertebrates and insects or nematode

Page 34: Searching for functional regions (coding or non-coding) in mammalian genomes

Oldest HCRsOldest HCRsMillion yearsPorifera (sponge)Nematodes (C. elegans)Arthropods (Drosophila)EchinodermsUrochordataCephalochordata (amphioxus)Jawless fisheschondrichthyes (ray, shark)actinopterygii (bony fishes)amphibians mammals birds reptiles600400200800Sequencing effort: 9 to 100 Mb 0.8 to 2.4 Mb less than 0.2 MbHistone 3’UTR- actin3’UTR

3 5’HOX UTRVertebrates

Page 35: Searching for functional regions (coding or non-coding) in mammalian genomes

Conservation pattern in Conservation pattern in 3’UTRs3’UTRs

position relative to the stop codon (nt)10005000150020002400c-fosTransferrin receptorbirdmammalEndoplasmic-reticulum Ca2+ ATPase birdmammalbirdmammalsimilarity: <60% ≥60% ≥70% ≥80%

Page 36: Searching for functional regions (coding or non-coding) in mammalian genomes

Distribution of HCRs within Distribution of HCRs within genesgenes3'-non-coding5'-non-codingintrons0%10%20%30%40%mammals / birdsmammals / amphibiansmammals / bony fishes2841917296512563812 Frequency of orthologous

genes containing HCRs

Page 37: Searching for functional regions (coding or non-coding) in mammalian genomes

HCRs and multigenic familiesHCRs and multigenic familiesHistone replacement variant H3.3A0400600100014001800AAAAAAAAUGStopAUGStopAAAAAAAHistone replacement variant H3.3BHistone replacement variant H3.3A and H3.3B, Calmodulinsnt• several genes coding for a same protein

• non-coding sequences are distinct, and conserved

Page 38: Searching for functional regions (coding or non-coding) in mammalian genomes

Function of 3’HCRs: Function of 3’HCRs: mRNA stability, translationmRNA stability, translationA+U-rich element: stability, translationposition relative to the stop codon (nt)10005000150020002400c-fosTransferrin receptorbirdmammalbirdmammalsimilarity: <60% ≥60% ≥70% ≥80%IRE : Iron Responsive Element

IRP : Iron Regulatory Protein

CCAGUGN5'3'

Page 39: Searching for functional regions (coding or non-coding) in mammalian genomes

Function of 3’HCRs:Function of 3’HCRs:mRNA subcellular localizationmRNA subcellular localization

Myosin heavy chain, c-myc, vimentin, -actin

chickencarp (bony fish)site poly(A)site poly(A)0200400600800position relative to the stop codon (nt)localization signalssimilarity: <60% ≥60% ≥70% ≥80%- 3’actin UTR

Page 40: Searching for functional regions (coding or non-coding) in mammalian genomes

ACUTS: compilation of ACUTS: compilation of Ancient Conserved Ancient Conserved

UnTranslated SequencesUnTranslated Sequences

Annotated multiple alignments:

˘ age of the conserved element

˘ gene function

˘ function of the conserved element

˘ bibliographic references (MEDLINE)

˘ sequences available from different species (EMBL)

˘ description of sequence features

http://pbil.univ-lyon1.fr/acuts/ACUTS.html

Page 41: Searching for functional regions (coding or non-coding) in mammalian genomes

Comparaison des régions non-codantes de 77 gènes Comparaison des régions non-codantes de 77 gènes orthologues homme/souris orthologues homme/souris (Jareborg et al. 1999)(Jareborg et al. 1999)

0.20.40.60.81Upstream(1 kb)5’UTRIntrons3’UTRs

Upstream015’UTRcoding exon

intron3’UTR

Fraction des régions non-codantes conservées entre homme et souris

Page 42: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction Prédiction ab initioab initio de gènes eucaryotes de gènes eucaryotes

Prédiction d ’exons codants Recherche de phases ouvertes de lecture (ORF: open reading frame)

– Taille moyenne des exons: ± 150 nt Statistiques sur les nucléotides, usage des codons

– Périodicité d'ordre 3, fréquence d ’hexamères– Modèles de Markov cachés

Signaux d ’épissage– Profils, modèles de Markov cachés, réseau neuronaux

Construction d ’un modèle de gène protéique Combinaison d ’exons de phases compatibles (pondération en fonction des scores de chaque exon

potentiel) Recherche de limites de gènes

– Exons terminaux (5 ’, 3 ’)– Promoteur– Signal de polyadénylation

Epissage alternatif ?? Exons non codants ?? Gène transcrits non codants (Xist, …) ??

Page 43: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction de gènes eucaryotes: Prédiction de gènes eucaryotes: qualité de la prédictionqualité de la prédiction

Comparaison des différents logiciels: sensibilité/spécificité Sn: sensibilité Sp: spécificité par exon (sn_e, sp_e) ou par nucéotide (sn_e, sp_e)

Jeu de données Burset-Guigo (1996): 570 gènes de vertébrés

Jeu de données Salamov et al (1998): 660 gènes humains

Sn_e Sp_e Sn_n Sp_nGenScan 0.78 0.81 0.93 0.93FGENES 1.6 0.83 0.82 0.92 0.93Grail2 0.36 0.43 0.72 0.87

Sn_e Sp_e Sn_n Sp_nGenScan 0.70 0.71 0.92 0.90FGENES 1.6 0.77 0.77 0.90 0.91

Page 44: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction de gènes eucaryotes: Prédiction de gènes eucaryotes: qualité de la prédictionqualité de la prédiction

Comparaison des différents logiciels: sensibilité/spécificité Sn: sensibilité Sp: spécificité par exon (sn_e, sp_e) ou par nucéotide (sn_e, sp_e)

Locus BRCA2 (1.4 Mb, chrom. 13q) (Sanger Centre 1999): région "difficile" pour les logiciels de prédiction. 159 exons

Sn_e Sp_e Sn_n Sp_nGenScan 0.66 0.36 0.81 0.44FGENES 1.6 0.69 0.57 0.79 0.66FGENES 1.6 masked 0.69 0.65 0.79 0.74GenScan+FGENES 0.61 0.82 0.67 0.90

Page 45: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction de gènes protéiques completsPrédiction de gènes protéiques complets C. elegans: la plupart des ‘ gènes ’ annotés sont seulement des prédictions Peut-on utiliser ces méthodes pour annoter les séquences génomique humaines ?

+ les faux positifs !

00.20.40.60.8113579111315Sensibilité par exon:90%80%

Probabilité de détecter tous les exons d’un gènesNombre d’exons du gène

Page 46: Searching for functional regions (coding or non-coding) in mammalian genomes

Un peu d ’optimismeUn peu d ’optimisme Fraction de la longueur des gènes correctement prédits:

70-80%

Probabilité que deux exons potentiels consécutifs soient réels (et donc positifs en RT-PCR)

0.5

Page 47: Searching for functional regions (coding or non-coding) in mammalian genomes

Prédiction de gènes eucaryotes (suite)Prédiction de gènes eucaryotes (suite)

Utilisation des EST Alignement séquence génomique / cDNA (EST): SIM4 (alignement de séquences

nucléiques très fortement similaires avec ‘ épissage ’)

Approche comparative Comparaison d ’une séquence génomique avec des gènes déjà caractérisés dans d ’autres

espèces (WISE2: alignement ADN/protéine avec épissage) Comparaison de séquences génomiques (non-annotées) homologues

– Locus mnd2 (homme souris) (Jang et al. 1999): >80 kb– Prédiction d ’exons internes basée sur la conservation de séquence

ORF ≥ 80 nt

Séquence protéique ≥ 70% similarité

Séquence ADN ≥50% identité

GT AG conservés

=> détection de tous les exons internes du gène D6Mm5e

– Généralisation de la méthode (Guigo 2000). Sensibilité ? Spécificité ?

Page 48: Searching for functional regions (coding or non-coding) in mammalian genomes

Stratégies de recherche de similarités: ADN ou protéine ?Stratégies de recherche de similarités: ADN ou protéine ? Limites des recherches de similarité au niveau ADN

Alphabet réduit (4 lettres) Dégénérescence du code génétique

Mais … tout n'est pas codant régions régulatrices, ARN structuraux, ...

Deux brins!Deux brins!

Page 49: Searching for functional regions (coding or non-coding) in mammalian genomes

Traitement du bruit de fond: filtres et masquesTraitement du bruit de fond: filtres et masques Séquences de faible complexité (protéines, ADN):

40% des protéines ADN: microsatellites

15% du total des résidus exemple: CACACACACACACACACA

Ala, Gly, Pro, Ser, Glu, Gln

logiciels de filtrage: SEG, XNU, DUST

RSPPR--KPQGPPQQEGNNPQGPPPPAGGNPQQPQAPPAGQPQGPP . ::: : :: : : ::::: : :: :.: :: : :::::QGPPRPGNQQCPPPQGG--PQGPPRP--GNQQRP--PPQGGPQGPP

(filtré par défaut par BLAST)

Séquences abondantes

3000 Immunoglobulines dans GenBank

106 Alu, 105 L1 dans le génome humain

logiciels de masquage: XBLAST, RepeatMasker

NNNNNNNNNNNNN

Page 50: Searching for functional regions (coding or non-coding) in mammalian genomes

Quelle approche adopter ?Quelle approche adopter ? Recherche rapide de similarité dans les banques

algorithme (BLAST, FASTA) matrices de substitution, pondération des gaps stratégie de recherche (nucléique, protéique) traitement du bruit de fond complétude des banques de données, répéter la recherche régulièrement

Comparaison d'une séquence génomique à un cDNA (EST) BLASTN < LFASTA << SIM4 (épissage)

Comparaison d'une séquence génomique à une protéine BLASTX, FASTX << WISE2 (épissage)

Comparaison de deux séquences génomiques ADN: BLASTN < LFASTA << Smith-Waterman (SIM, LALIGN) Protéique: TBLASTX

Page 51: Searching for functional regions (coding or non-coding) in mammalian genomes

Alignement par bloc ou alignement global : comparaison BLAST / FASTA

protéine 1protéine 2A1B1C1A'1A2C2B2A2A1B1C1A'1A2C2B2A1B1C1A'1A2C2B2Recherche desimilitudeFASTABLAST

Page 52: Searching for functional regions (coding or non-coding) in mammalian genomes

Représentation graphique des similarités entre séquences: DOTTERReprésentation graphique des similarités entre séquences: DOTTER

Page 53: Searching for functional regions (coding or non-coding) in mammalian genomes

Représentation graphique des similarités entre séquences: LALNVIEWReprésentation graphique des similarités entre séquences: LALNVIEW

Page 54: Searching for functional regions (coding or non-coding) in mammalian genomes

Représentation graphique des similarités entre séquences: PIP-makerReprésentation graphique des similarités entre séquences: PIP-maker

Page 55: Searching for functional regions (coding or non-coding) in mammalian genomes

Limitation des comparaisons deux à deux (BLAST, Limitation des comparaisons deux à deux (BLAST, FASTA, ...)FASTA, ...)

Seq A CGRRLILFMLATCGECDTDSSE … HICCIKQCDVQDIIRVCC

:: : ::: :: : :

Insuline CGSHLVEALYLVCGERGFFYTP … EQCCTSICSLYQLENYCN

::: : : : :: : :

Seq B YQSHLLIVLLAITLECFFSDRK … KRQWISIFDLQTLRPMTA

Comparaisons 2 à 2:

Insuline / Seq A : 25% d'identité

Insuline / Seq B : 25% d'identité

Page 56: Searching for functional regions (coding or non-coding) in mammalian genomes

Alignement de séquences de la famille des insulinesAlignement de séquences de la famille des insulines

B-chain A-chain

INSL4 Q14641 ELRGCGPRFGKHLLSYCPMPEKTFTTTPGG...[x]58 ....SGRHRFDPFCCEVICDDGTSVKLCT

INSL3 P51460 REKLCGHHFVRALVRVCGGPRWSTEA.......[x]51 ....AAATNPARYCCLSGCTQQDLLTLCPY

RLN1 P04808 VIKLCGRELVRAQIAICGMSTWS..........[x]109 ....PYVALFEKCCLIGCTKRSLAKYC

BBXA P26732 VHTYCGRHLARTLADLCWEAGVD..........[x]25 ........GIVDECCLRPCSVDVLLSYC

BBXB P26733 ARTYCGRHLADTLADLCF--GVE..........[x]23 ........GVVDECCFRPCTLDVLLSYCG

BBXC P26735 SQFYCGDFLARTMSILCWPDMP...........[x]25 ........GIVDECCYRPCTTDVLKLYCDKQI

BBXD P26736 GHIYCGRYLAYKMADLCWRAGFE..........[x]25 ........GIADECCLQPCTNDVLLSYC

LIRP P15131 VARYCGEKLSNALKLVCRGNYNTMF........[x]58 ........GVFDECCRKSCSISELQTYCGRR

MIP I P07223 RRGVCGSALADLVDFACSSSNQPAMV.......[x]29 ....QGTTNIVCECCMKPCTLSELRQYCP

MIP II P25289 PRGICGSNLAGFRAFICSNQNSPSMV.......[x]44 ....QRTTNLVCECCFNYCTPDVVRKYCY

MIP III P80090 PRGLCGSTLANMVQWLCSTYTTSSKV.......[x]30 ....ESRPSIVCECCFNQCTVQELLAYC

MIP V P31241 PRGICGSDLADLRAFICSRRNQPAMV.......[x]44 ....QRTTNLVCECCYNVCTVDVFYEYCY

MIP VII P91797 PRGLCGNRLARAHANLCFLLRNTYPDIFPR...[x]86 ..EVMAEPSLVCDCCYNECSVRKLATYC

ILP P22334 AEYLCGSTLADVLSFVCGNRGYNSQP.......[x]31 ........GLVEECCYNVCDYSQLESYCNPYS

INS P01308 NQHLCGSHLVEALYLVCGERGFFYTPKT.....[x]35 ........GIVEQCCTSICSLYQLENYCN

IGF1 P01343 PETLCGAELVDALQFVCGDRGFYF.........[x]12 ........GIVDECCFRSCDLRRLEMYCAPLK

IGF2 P01344 SETLCGGELVDTLQFVCGDRGFYF.........[x]12 ........GIVEECCFRSCDLALLETYCATPA

*. .* ** * . *

Page 57: Searching for functional regions (coding or non-coding) in mammalian genomes

NH3-SSSSSSA-chain- COOHB-chainC-peptideSSSSSS signal peptideB chainA chainC peptide

Structure des insulines

Page 58: Searching for functional regions (coding or non-coding) in mammalian genomes

Représentation d ’un motif conservé par une Représentation d ’un motif conservé par une matrice de matrice de fréquences (profil)fréquences (profil)

Exemple: site donneur d ’épissage (vertébrés)

Matrice de fréquence (pourcentage):

Base Position

-3 -2 -1 +1 +2 +3 +4 +5 +6

A 33 60 8 0 0 49 71 6 15

C 37 13 4 0 0 3 7 5 19

G 18 14 81 100 0 45 12 84 20

T 12 13 7 0 100 3 9 5 46

Cons. M A G G T R A G T

Exon IntronExon Intron

Page 59: Searching for functional regions (coding or non-coding) in mammalian genomes

Recherche d ’un motif dans une séquence à l ’aide d ’un Recherche d ’un motif dans une séquence à l ’aide d ’un profilprofil

Calcul des scores de similarité en faisant glisser une fenêtre de la longueur du motif le long de la séquence. Exemple:

A 33 60 8 0 0 49 71 6 15

C 37 13 4 0 0 3 7 5 19

G 18 14 81 100 0 45 12 84 20

T 12 13 7 0 100 3 9 5 46

GAAAGGTGAGTCAT...

GAAAGGTGA S=18+60+8+0+0+45+9+84+15=239

.AAAGGTGAG S=33+60+8+100+0+3+12+6+20=242

..AAGGTGAGT S=33+60+81+100+100+45+71+84+46=620

...AGGTGAGTC S=33+14+81+0+0+49+12+5+19=213

....GGTGAGTCA …etc

Page 60: Searching for functional regions (coding or non-coding) in mammalian genomes

PSI-BLASTPSI-BLAST

Position-Specific Iterated BLAST 1-recherche BLAST classique (protéine) 2-construction d'une matrice de pondération (profil) avec les séquences

similaires détectées 3-recherche BLAST à partir de ce nouveau profil 4-itération des étapes 2-3 jusqu'à convergence

plus sensible que Smith-Waterman 40 fois plus rapide

Page 61: Searching for functional regions (coding or non-coding) in mammalian genomes

Comparaison d ’une séquence génomique à un Comparaison d ’une séquence génomique à un motif ou à une banque de motifsmotif ou à une banque de motifs

Banques de données de motifs protéiques PROSITE BLOCKS PFAM PRODOM PRINTS

Logiciels de comparaison d ’une séquence à une banque de motifs, ou de comparaison d ’un motif à une banque de séquences Pfscan: profils WISE2