350 likes | 362 Views
CSE182-L11. Gene Finding. HMM for detecting CpG Islands. A. 1.0. B. A. G. A. 0.2. 0.1. end. G. 1.0. 1.0. start. 0.2. end. C. start. 0.4. T. C. T. 0.2. 0.1. In the best parse of a genomic sequence, each base is assigned a state from the sets A, and B.
E N D
CSE182-L11 Gene Finding
HMM for detecting CpG Islands A 1.0 B A G A 0.2 0.1 end G 1.0 1.0 start 0.2 end C start 0.4 T C T 0.2 0.1 • In the best parse of a genomic sequence, each base is assigned a state from the sets A, and B. • Any substring with multiple states coming from B can be described as a CpG island. • We move from genomic to CpG by goint to the G state , and from CpG to genomic only by going to the C state.
Optimum parse Seq: C T C T C T C G C G C G C G C T C T C T Gen: 4 4 4 2 1 1 1 1 2 4 4 4 CpG: 2 2 2 2 2 2 2 2 2 2 2 2 Opt: 4 4 4 1 2 2 2 2 1 4 4 4 Optπ:A A A A A A B B B B B B B B A A A A A A Win: 0 0 0 0 1 2 2 3 3 4 4 • All missing columns have identical values and can be ignored in comparison • The denominator is the same and can be ignored • Pr(Gen) =~ 46 22= 214 • Pr(CpG) =~ 262224 = 212 • Pr(Opt) =~ 4624=216 AND precise boundaries • A windowed count will give the same answer, but without boundaries
HMM: Summary • HMMs are a natural technique for modeling many biological domains. • They can capture position dependent, and also compositional properties. • HMMs have been very useful in an important Bioinformatics application: gene finding.
Gene Finding: The 1st generation • Given genomic DNA, does it contain a gene (or not)? • Key idea: The distributions of nucleotides is different in coding (translated exons) and non-coding regions. • Therefore, a statistical test can be used to discriminate between coding and non-coding regions.
Coding versus non-coding • Fickett and Tung (1992) compared various measures • Measures that preserve the triplet frame are the most successful. • Genscan: 5th order Markov Model • Conservation across species
Coding vs. non-coding regions Compute average coding score (per base) of exons and introns, and take the difference. If the measure is good, the difference must be biased away from 0.
Other Signals ATG AG GT Coding
Coding region can be detected • Plot the coding score using a sliding window of fixed length. • The (large) exons will show up reliably. • Not enough to predict gene boundaries reliably Coding
Other Signals • Signals at exon boundaries are precise but not specific. Coding signals are specific but not precise. • When combined they can be effective ATG AG GT Coding
The second generation of Gene finding • Ex: Grail II. Used statistical techniques to combine various signals into a coherent gene structure. • It was not easy to train on many parameters. Guigo & Bursett test revealed that accuracy was still very low. • Problem with multiple genes in a genomic region
HMMs and gene finding • HMMs allow for a systematic approach to merging many signals. • They can model multiple genes, partial genes in a genomic region, as also genes on both strands.
HMMs and gene finding • The Viterbi algorithm (and backtracking) allows us to parse a string through the states of an HMM • Can we describe Eukaryotic gene structure by the states of an HMM? • This could be a solution to the GF problem.
Generalized HMMs, and other refinements • A probabilistic model for each of the states (ex: Exon, Splice site) needs to be described • In standard HMMs, there is an exponential distribution on the duration of time spent in a state. • This is violated by many states of the gene structure HMM. Solution is to model these using generalized HMMs.
Generalized HMM for gene finding • Each state also emits a ‘duration’ for which it will cycle in the same state. The time is generated according to a random process that depends on the state.
Forward algorithm for gene finding qk j i Duration Prob.: Probability that you stayed in state qk for j-i+1 steps Emission Prob.: Probability that you emitted Xi..Xj in state qk (given by the 5th order markov model) Forward Prob: Probability that you emitted I symbols and ended up in state qk
HMMs and Gene finding • Generalized HMMs are an attractive model for computational gene finding • Allow incorporation of various signals • Quality of gene finding depends upon quality of signals.
DNA Signals • Coding versus non-coding • Splice Signals • Translation start
Splice signals • GT is a Donor signal, and AG is the acceptor signal GT AG
PWMs 321123456 AAGGTGAGT CCGGTAAGT GAGGTGAGG TAGGTAAGG • Fixed length for the splice signal. • Each position is generated independently according to a distribution • Figure shows data from > 1200 donor sites
MDD • PWMs do not capture correlations between positions • Many position pairs in the Donor signal are correlated
Choose the position which has the highest correlation score. • Split sequences into two: those which have the consensus at position I, and the remaining. • Recurse until <Terminating conditions>
De novo Gene prediction: Sumary • Various signals distinguish coding regions from non-coding • HMMs are a reasonable model for Gene structures, and provide a uniform method for combining various signals. • Further improvement may come from improved signal detection
How many genes do we have? Nature Science
Comparative methods • Gene prediction is harder with alternative splicing. • One approach might be to use comparative methods to detect genes • Given a similar mRNA/protein (from another species, perhaps?), can you find the best parse of a genomic sequence that matches that target sequence • Yes, with a variant on alignment algorithms that penalize separately for introns, versus other gaps.
Comparative gene finding tools • Procrustes/Sim4: mRNA vs. genomic • Genewise: proteins versus genomic • CEM: genomic versus genomic • Twinscan: Combines comparative and de novo approach.