gene finding in eukaryotesubio.bioinfo.cnio.es/cursos/cursoverano2008/madrid08/gene_finding... ·...

Gene Finding in Eukaryotes

Jan-Jaap [email protected]

Computational and Structural Biology Group, Centro Nacional de InvestigacionesOncologicas

Madrid, July 2008

Jan-Jaap Wesselink [email protected] Gene Finding in Eukaryotes Madrid, July 2008 1 / 24

Outline

1 Gene StructureEukaryotesFind Signals

2 Different Approaches To Gene FindingDifferent InformationSearch by SignalPWMSearch by ContentCoding StatisticsHMMHomologyComparative

3 Accuracy of Predictions

4 References

Outline

4 References

Outline

4 References

Outline

4 References

Outline

4 References

Eurkaryotic Gene Structure

Schematic Gene Structure

GT AG GT AG

Exon Intron Exon Intron Exon

UTR CDS

Gene prediction programs only predict the coding fraction of genes

Signals Exons RegionsStart (ATG) Single Exons

Stops (TGA,TAA,TAG) First IntronsDonor (GT) Internal Intergenic

Acceptor (AG) Terminal 5’ and 3’ UTRs

Signals are difficult to find (1)

ExampleTry reading this sentence:

LOOKITSMUCHEASIERLIKETHIS

Signals are difficult to find (1)

ExampleTry reading this sentence:

Look! It’s much easier like this!

Signals Are Difficult To Find (2)

ExampleGenomic DNA sequence

GTTTCAAGTGATCCTCCCGCCTCAGCCTGCCCAGGTGCTGAGATTACATGTATGAGCCACTGCACCTGGAAAGGAGCCAGAAATGTGAAGTGCTAGCTGAAGGATGAGCAGCAGCTAGCCAGGCAAAGGTAGGGTTTGGGGAAGGAAAGTGCACATTCTCTTCCCATCTGTGTTTCAGGGGGCAATGGCGGCTTCCTGTGTTCTACTGCACACTGGGCAGAAGATGCCTCTGATTGGTCTGGGTACCTGGAAGAGTGAGCCTGGTCAGGTGAGGGATGGGGGAAGAAAAAAGAAACCTCTGCTTCTCTCACCTGGCAGGTAAAAGCAGCTGTTAAGTATGCCCTTAGCGTAGGCTACCGCCACATTGATTGTGCTGCTATCTACGGCAATGAGCCTGAGATTGGGGAGGCCCTGAAGGAGGACGTGGGACCAGGCAAGGTAAGGACTGGGGTTGTAAATAGAGCTGTGGGCCCTGCCCCCTGCACTAG

Only beginning and end of introns are shown

Signals Are Difficult To Find (2)

ExampleGenomic DNA sequence

GTTTCAAGTGATCCTCCCGCCTCAGCCTGCCCAGGTGCTGAGATTACATGTATGAGCCACTGCACCTGGAAAGGAGCCAGAAATGTGAAGTGCTAGCTGAAGGATGAGCAGCAGCTAGCCAGGCAAAGGTAGGGTTTGGGGAAGGAAAGTGCACATTCTCTTCCCATCTGTGTTTCAGGGGGCAATGGCGGCTTCCTGTGTTCTACTGCACACTGGGCAGAAGATGCCTCTGATTGGTCTGGGTACCTGGAAGAGTGAGCCTGGTCAGGTGAGGGATGGGGGAAGAAAAAAGAAACCTCTGCTTCTCTCACCTGGCAGGTAAAAGCAGCTGTTAAGTATGCCCTTAGCGTAGGCTACCGCCACATTGATTGTGCTGCTATCTACGGCAATGAGCCTGAGATTGGGGAGGCCCTGAAGGAGGACGTGGGACCAGGCAAGGTAAGGACTGGGGTTGTAAATAGAGCTGTGGGCCCTGCCCCCTGCACTAG

Only beginning and end of introns are shown

All Signals Predicted by geneid in a Genomic DNA Sequence

All Exons Predicted by geneid in a Genomic DNA Sequence

Outline

4 References

Different Approaches to Gene Finding

Different Types of Information Can be Used:Signals: search for signals of transcription, splicing, translation.Typically, these signals are assigned a score, and the highestscoring signals are combined.Content: here, one tries to discriminate the protein coding fromnon-coding regions. Statistical models of nucleotide frequenciesand dependencies in codons are used here.Homology: significant sequence similarity of a genomic DNAsequence to a known gene, implies that it is likely to share itsfunction. This information may be used in the gene predictionprocess.

Search by Signal

Signals are usually represented as patterns

Example (patterns)Strings: P = GCCACCTAGGConsensus sequences: subsitutions occur at certain positionsRegular expressions: describe set of strings generated by aregular language.Decision treesPosition Weight Matrices

Search by Signal

Patterns: examples

Example (patterns 2)Consensus sequence:

p1 = CTAAAAATAA

p2 = TTAAAAATAA

p3 = TTTAAAATAA

p4 = CTATAAATAA

p5 = TTATAAATAA

p6 = CTTAAAATAG

p7 = TTTAAAATAG

P = YTWWAAATAR

Y = pyrimidine (C or T), W = A or T, R = purine ( A or G)Regular expression: Prosite pattern:P = G − [GN]− [SGA]−G − x − R − x − [SGA]− C − x(2)− [IV ]

Patterns: examples

Example (patterns 2)Consensus sequence:

p1 = CTAAAAATAA

p2 = TTAAAAATAA

p3 = TTTAAAATAA

p4 = CTATAAATAA

p5 = TTATAAATAA

p6 = CTTAAAATAG

p7 = TTTAAAATAG

P = YTWWAAATAR

Y = pyrimidine (C or T), W = A or T, R = purine ( A or G)Regular expression: Prosite pattern:P = G − [GN]− [SGA]−G − x − R − x − [SGA]− C − x(2)− [IV ]

Position Weight Matrices

Construction of PWM for splice sites (...GT.... or ...AG....)1 align sequences for true splice sites2 calculate relative frequencies3 same for false donor sites4 calculate log odds ratio true/false frequencies

Mi,j = ln(ftrue

ffalse)

5 score of a given sequence is sum of matrix coefficients

Mi,j = ln(ftrue

ffalse)

Mi,j = ln(ftrue

ffalse)

Mi,j = ln(ftrue

ffalse)

Mi,j = ln(ftrue

ffalse)

Mi,j = ln(ftrue

ffalse)

Example: Donor Sites

Search by Content

Coding statisticsHidden Markov Models

Coding Statisticsthere are 64 codons for 22 amino acidsdifferent codons are used in exons than in intronscompute codon usage from coding DNA → frequenciesfor a sequence, S, the higher the number of codons in S thatoccur frequently in coding sequences, the higher the probabilitythat S is coding for a protein

Search by Content

Coding Statistics

The probability of a sequence S being protein coding:

p(S) =p(C1) · p(C2) . . . p(Cn)

164 ·

164 . . . 1

p(S) ≈ f (C1) · f (C2) . . . f (Cn)1

64 ·1

64 . . . 164

Coding Statistics

The probability of a sequence S being protein coding:

p(S) =p(C1) · p(C2) . . . p(Cn)

164 ·

164 . . . 1

p(S) ≈ f (C1) · f (C2) . . . f (Cn)1

64 ·1

64 . . . 164

Hidden Markov Models

can have HMM’s for entire genescan have HMM’s for coding/non-coding regions

Search by Homology

Various Uses of Homology in Gene Predictionone can compare a genomic DNA sequence with a database of ESTs(using e.g. blastn)genomic DNA sequences can be compared to a database proteinsequences (using blastx, to identify coding regionscomparison of predicted peptides with a protein sequence data base canbe used to assign putative functionsthe genome of one species can be compared to the genome of another,closely related species: conserved regions often correspond toconserved functions (e.g. exons, parts of promoters)

Search by Homology

Comparative Gene Prediction

Comparing the human FOS gene with:

(a) Mouse (b) Chicken (c) Pufferfish

using tblastx

Outline

4 References

Gene Prediction Accuracy

measured in annotated sequencescan measure at nucleotide, exon and gene level

Sn =TP

TP + FNSP =

TPTP + FP

Gene Prediction Accuracy

measured in annotated sequencescan measure at nucleotide, exon and gene level

Sn =TP

TP + FNSP =

TPTP + FP

Outline

4 References

References

1 http://genome.imim.es/courses/BioinformaticaUPF/

2 http://genome.imim.es/∼jjw/oeiras05/3 Guigo, R. (1999) DNA Composition, Codon Usage and Exon Prediction.

In Bisshop, M., ed. Genetic Databases, Academic Press.

4 Eddy, S. (2004) What is a Hidden Markov Model? Nature 22:1315–6.

5 Burset, M. and Guigo, R. (1996). Evaluation of gene structure predictionprograms. Genomics, 34:353–7.

6 Brent M.R. and Guigo, R. (2004). Recent advances in gene structureprediction. Curr. Opin. Struct. Biol. 14:264–72.

7 Guigo, R. and Reese, M.G. (2005). EGASP: collaboration throughcompetition to find human genes. Nature Methods, 2:575–6.

gene finding in eukaryotesubio.bioinfo.cnio.es/cursos/cursoverano2008/madrid08/gene_finding... ·...

Documents