150 likes | 252 Vues
This study explores enhancing gene structure prediction in plants through genome comparison, focusing on explicit intron length models and splice donor models for GT/AG and GC/AG introns. Performance improvements on single gene sets and genome sets are highlighted, along with a decision tree model for donor sites prediction and results from experiments on Arabidopsis and rice datasets. The analysis includes breakdowns of predictions, data limitations, and improvements on TIGR and GeneBank curated datasets.
E N D
Exploiting Genome Comparison for Gene Structure Prediction in Plants Michael Brent Ping Hu
Performance Improvement on single gene set: Explicit Intron Length Model
Performance Improvement on genome set: Explicit Intron Length Model
Splice Donor Model for both GT/AG and GC/AG introns • GC/AG introns represent: • 252/33350=0.75% in Arabidopsis • 0.7% of total human pre-mRNA introns; • ~0.6% in C. elegans (Nuc Acid Research 30(15) 3360-3368). • 27/2034 = 1.3% in crypto • Old model can not predict the GC/AG intron
Decision Tree Model for GT/GC Donors Donor sites NNNG1T2NNNN NNNG1C2NNNN NNNGTNNG5N NNNGTNNĞ5N NNG-1GTNNGN NNĞ-1GTNNGN NA-2GGTNNGN NĂ-2GGTNNGN NAGGTNNGT6 NAGGTNNGŤ6
Breakdown of Arabidopsis Predictions Total Prediction: 30634/Total annotation: 28581 Identical to ann: 15063 Not Identical to ann: 15588 Overlap with confirmed ann: 3246 Not overlap with confirmed ann: 12342 Not overlap with any ann: 4394 Overlap with other ann: 7948 Same start/Same stop 2358 Diff start/Same stop 2770 Same start/Diff stop 1879 Diff start/Diff stop 941
First Experiment Result 2000 1650 1000 850 650 500 400 300 200 100 M 1 2 3 4 5 6 7 8 9 10 11 12
Rice Annotation Data Set • TIGR data set: • Most annotations were based on FgeneSH • Get manually curate contigs: 3171 genes • May still been influenced by FgeneSH • Gene bank cDNA confirmed data set: • Download Genes with full length cDNA from Genebank • Total: 1084 mRNA and 443 DNA • Filter out the bad genes with stop codon in frame, 341 DNA left • Limitation of this data set: • UTR and Intergenic region very limited • All positive strand, small • Other data sets are all from automatic pipelines
Performance Improvement on Rice GeneBank cDNA-Confirmed Data