Gene Expression: Details (Eukaryotes) Pre-mRNA Secondary Structure Prediction Aids Splice Site Recognition

Transcription

1 re-mrna rediction Aids plice ite Recognition Gene Expression: Details (Eukaryotes) DNA pre-mrna mrna rotein nucleus gene rotein Donald J. atterson, Ken Yasuhara, Walter L. Ruzzo January 3-7, 2002 DNA (chromosome) acific ymposium on Biocomputing University of Washington Computational Molecular Biology Group 1 cell premrna mrna 3 Architecture of a Gene Characteristics of human genes (Nature, 2/2001, Table 21) pre-mrna s transcribed from most genes contain introns, which must be spliced out to form useful mrnas Internal exon Exon number Introns Median 122 bp 7 1,023 bp Mean 145 bp 8.8 3,365 bp ample (size) Refeq alignments to draft genome sequence, with confirmed intron boundaries (43,317 exons) Refeq alignments to finished sequence (3,501 genes) Refeq alignments to finished sequence (27,238 introns) Exons: ' UTR 5' UTR 400 bp 240 bp 770 bp 300 bp Confirmed by mrna or ET on chromo 22 (689) Confirmed by mrna or ET on chromo 22 (463) Introns: a b c re-mrna Coding seq (CD) 1,100 bp 367 aa 1340bp 447 aa elected Refeq entries (1,804)* mrna Genomic extent 14 kb 27 kb elected Refeq entries (1,804)* Encodes a protein 7 * 1,804 selected Refeq entries were those with fulllength unambiguous alignment to finished sequence 8 1

2 Relevance of plice rediction Cell, Vol. 92, , February 6, 1998 Mechanical Devices of the pliceosome: Motors, Clocks, prings, and Things Jonathan. taley and Christine Guthrie 9 plice site prediction is critical to eukaryotic gene prediction. Average human gene has 8.8 exons Genes with over 175 exons known Current primary sequence models do not display the same discriminatory power that cells exhibit in vivo mall per-site error rate compounds 10 ossible acceptor splice sites re-mrna sequences Hypothesis rediction structure contains information useful for predicting splice site location. This information is in addition to primary sequence information. pecific instances of secondary structure variation affecting the splicing process. rimary redictions

3 ossible acceptor splice sites re-mrna sequences rimary rediction redictions Data et Drawn from 462 unrelated, annotated, multiexon human genes with standard splicing. (Reese 97) 1,980 acceptor splice sites (3 end of intron) 1,980 non-sites selected randomly Aligned to an AG consensus Located within 100 bases of an annotated acceptor splice site ossible acceptor splice sites re-mrna sequences rimary rediction What's in the rimary equence? exon 5 redictions intron 15 exon 16 3

4 What's in the rimary equence? A C G T intron acceptor splice site exon Weight Matrix Model (0 th order Markov Model) 17 equence-based Metric 1 st order Weight Array Matrix / Markov Model i (N i ={A,C,G,U} N i-1 ={A,C,G,U} ) Training Generate two conditional probability tables for positions ( 21,+3), one from positive examples and one from negative examples. Testing For each sequence, x, calculate its likelihood ratio: + & ( )# WAM x log 10 $ ' % ( x)!! WAM " 18 ossible acceptor splice sites re-mrna sequences Acceptor plice ite rimary equence rediction Model redictions

5 ossible acceptor splice sites re-mrna sequences rimary rediction redictions ptimal Folding Energy Max Helix score Neighbor airing Correlation Model ptimal Folding Energy...CUGCUUUCUCCCCUCUCAGGGACUUACAGUUUGAGAUGC... equence rediction 2. Max Helix What is the highest probability that a helix will form nearby? Calculate Calculate Htart, x HEnd, x Free Energy kcal/mole Free Energy kcal/mole Free Energy -2.0 kcal/mole 23 Helix MaxHelix i = max ( Htart, x, HEnd, x ) x" ( i! 5, i+ 5) 24 5

6 3. Neighbor airing Correlation Model 3. Neighbor airing Correlation Model Change the premrna alphabet from nucleotides to structural symbols Unpaired base aired base aired and stacked base Change the premrna alphabet from nucleotides to structural symbols Unpaired base aired base aired and stacked base Neighbor airing Correlation Model 2 nd order Markov Model i (N i ={,,} N i-1 ={,,} ^ N i-2 ={,,} ) Training Generate two conditional probability tables for positions ( 50,+3), one from positive examples and one from negative examples. Testing For each sequence, x, calculate its log likelihood ratio: rimary ossible acceptor splice sites re-mrna sequences rediction redictions & log 10 $ % + NCM ' NCM ( x) # ( x)!! "

7 ossible acceptor splice sites re-mrna sequences s Decision Trees Quinlan s C4.5 upport Vector Machines Noble s svm 1.1 Radial Basis Kernel degree 2 Both take a vector of statistics and produce a yes/no binary classifier. 29 rimary rediction redictions 31 Features WAM (baseline) WAM,FE Results (Decision Trees) Mean Accuracy (%) % Error Reduction 5.5 p LLR of Base airing 25% more likely for acceptor splice sites to pair at position -2 WAM,FE,NCM WAM,FE,MH WAM,FE,NCM,MH WAM = Weight Array Matrix (rimary equence Method) FE = ptimal Free Energy MH = Max Helix NCM = Neighbor airing Correlation Matrix Wilcoxon p-value under 10-fold cross-validation

8 LLR of Helix Initiation 35% more likely for acceptor splice sites to initiate a helix at position 2 and -1 LLR of Helix Results Continuation 45% more likely for acceptor splice sites to continue a helix through the splice site

9 Helix Formed at plice ite Acceptor Non-Acceptor r(no Helix) r(helix) r(folds Left) r(folds Right) Conclusions structure statistics correlate with splice site location. ur models (Max Helix, NCM) can represent some of the relevant secondary structure. These models capture correlations that current primary sequence models don t capture Future Work ther organisms ryza sativa (rice) in progress Donor splice sites ther features? More structure models tochastic Context Free Grammars? Acknowledgements Don aterson Ken Yasuhara Jeff toner Kevin Chu More Info 40 UW CE Computational Biology Group 41 9