Eukaryotic Gene Prediction. Wei Zhu May 2007

Transcription

1 Eukaryotic Gene Prediction Wei Zhu May 2007

2 In nature, nothing is perfect... - Alice Walker

3 Gene Structure

4 What is Gene Prediction? Gene prediction is the problem of parsing a sequence into nonoverlapping coding segments (CDSs) consisting of exons separated by introns.

5 Signal Sensors A signal sensor evaluates fixed-length features in DNA. Start codons Stop codons Donor sites Acceptor sites Promoters Poly-A signals

6 Content Sensors A content sensor evaluates variable-length features which extend from one signal to another: Exons Introns Intergenic regions UTRs

7 Gene Prediction Approaches Intrinsic (ab( initio) GENSCAN, FGENESH, GeneMark.hmm GlimmerM, Genie; Extrinsic (similarity-based) Spliced alignment: GenomeScan, EuGene,, FGENESH+, FGENESH_C, GeneId+, etc; Genomic comparison: TwinScan,, TWAIN, SLAM, SGP, FGENESH_2, etc; Integrated GeneScope, GeneMachine,, JIGSAW, RiceGAAS, Ensembl,, EVM etc.

8 ab initio Gene Prediction Adopt a rigorous probabilistic model of sequence structure and choose the most probable parse according to that probabilistic model. Pros Cons Fast and efficient Remarkable accuracy at the nucleotide level Less than 50% accuracy at the gene level

9 Development of a Gene Finder Build the model Train the model to generate the related parameters Predict/Evaluate

10 Imperfect Model GT..AG A G G T 1-bp intron

11 Nucleotide level Exon level Gene level Accuracy Evaluation

12 Nucleotide/Base Level Prediction accuracy per base coding/non-coding

13 Exon Level Prediction accuracy with respect to exact prediction of exon start and end points

14 Gene/Protein Level Prediction accuracy with respect to the protein product encoded by the predicted gene

15 A Simple Calculation Given x accuracy at exon level, the accuracy of the prediction at the gene level is: P = P (all exons correctly predicted) =x n, where n is the number of exons in the gene. Typically, x<90% and n=5,, then P = 0.9x0.9x0.9x0.9x0.9 = 59%

16 Performance Species-specific setting GC content Gene density Gene/Exon/Intron length distribution Codon usage Benchmark training data set test data set

17 Maize Gene Prediction

18 Gene Finders

19 Accuracy

20 Challenges of Intrinsic Approaches Alternative splicing Nested/overlapped genes Extremely long/short genes Extremely long introns Extremely short exons Non-canonical introns Frame-shift errors Split start codons (that is, the start codon is split by an intron in the genomic sequence) UTR introns Non-ATG triplet as the start codon Polycistronic genes

21 Gene Prediction Approaches Intrinsic (ab( initio) GENSCAN, FGENESH, GeneMark.hmm GlimmerM, Genie; Extrinsic (similarity-based) Spliced alignment: GenomeScan, EuGene,, FGENESH+, FGENESH_C, GeneId+, etc; Genomic comparison: TwinScan,, TWAIN, SLAM, SGP, FGENESH_2, etc; Integrated GeneScope, GeneMachine,, JIGSAW, RiceGAAS, Ensembl,, etc.

22 Similarity-based Gene Prediction EST/cDNA spliced alignment Protein spliced alignment Genomic comparison Intra-genomic Inter-genomic

23 EST/cDNA Spliced Alignment

24 Pros and Cons Pros Cons High accuracy Unavailability or incompleteness of transcript sequence data Extra computation to generate alignments Diverse sequence quality Incomplete full-length cdna Contamination Incorrect sequence orientations

25 Genomic Comparison Microsynteny between M. truncatula and Arabidopsis Hongyan et al, 2003

26 Gene Structure of Syntenic and non-syntenic Homologous Genes Hongyan et al, 2003

27 Comparative Analysis of Cereal Gene Structures

28 Comparative Analysis of Cereal Gene Promoters

29 Pros and Cons Pros Aid to identify low expressed genes Identify genes in multiple species simultaneously Aid to identify transcription factor binding sites Uncover non-protein coding genes Cons Performance will depend on the evolutionary distance between the compared sequences. Exon/intron boundaries may not be conserved

30 Tiling Array

31 ARTADE -ARabidopsis Tiling-Array-based Detection of Exons

32 Gene Prediction Approaches Intrinsic (ab( initio) GENSCAN, FGENESH, GeneMark.hmm GlimmerM, Genie; Extrinsic (similarity-based) Spliced alignment: GenomeScan, EuGene,, FGENESH+, FGENESH_C, GeneId+, etc; Genomic comparison: TwinScan,, TWAIN, SLAM, SGP, FGENESH_2, etc; Integrated GeneScope, GeneMachine,, JIGSAW (combiner), RiceGAAS, Ensembl,, etc.

33 Gene Discovery via Multiple Gene Finders

34 EVM

35 TIGR Rice Genome Annotation Pipeline

36 RiceGAAC

37 Ensembl Gene Prediction Procedure

38 Summary Nothing is perfect Each gene identification approach has its own features and limitations; Genome annotation is an on-going process, and the accuracy is being improved along with the accumulation of the evidence data;trnasnorna trnasnorna

39 Case Study

40 Sorghum-Rice Synteny and EST Read Pair

41 Create a Gene Model

42 Expression Data Data Type EST/FL-cDNA Peptide MPSS SAGE Microarray Tiling array Data Source PASA/Manual curation Koller et al., PNAS, 2002 (6,296 peptides/2,528 fgenesh models) Blake Meyers ( mpss.udel.edu/rice/) 126,663 tags from MGOS ( NSF Rice Oligonucleotide Array project ( ricearrary.org) Deng lab, Yale University

43 Expression Data in Gbrowse

44 GATCGATC I. Library construction Brenner et al., PNAS 97: AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA AAAAAAA mrna MPSS SEQUENCING TECHNOLOGY 1) Cut w/ DpnII 2) Ligate MmeI adapter MmeI TTTTT AAAAA 3) Cut to capture bp signature 4) Add DNA barcode, amplify & capture on beads Each bead contains the amplified product derived from the 3 end of a single transcript. II. Loading the flow cell + NNNN III. Sequencing of tags Brenner et al., Nat. Biotech. 18: NNNX NNXN RS RS CODEX1 CODEX2 2) Sequence by hybridization NXNN RS CODEX3 XNNN RS CODEX4 1) Add adaptors 16 cycles for 4 bp 3) Digest with Type IIS enzyme to uncover next 4 bases, repeat cycle

45 Ovary and mature stigma

46 Refine Gene Structure

47 Have no fear of perfection - you'll never reach it. - Salvador Dalí