Bioinformatics生物信息学理论和实践唐继军jtang@cse.sc.edu.ppt

Analyzing DNA Composition DNA composition varies a lot Stability of a DNA sequence depends on its G+C content (total guanine and cytosine) High G+C makes very stable DNA molecules Online resources are available to measure the GC content of your DNA sequence Also for counting words and internal repeats /emboss/html/ Counting words ATGGCTGACT A, T, G, G, C, T, G, A, C, T AT, TG, GG, GC, CT, TG, GA, AC, CT ATG, TGG, GGC, GCT, CTG, TGA, GAC, ACT www.genomatix.de/cgi-bin/tools/tools.pl EMBOSS servers European Molecular Biology Open Software Suite /emboss/ ORF EMBOSS NCBI /gorf/gorf.html Internal repeats A word repeated in the sequence, long enough to not occur by chance Can be imperfect (regular expression) Dot plot is the best way to spot it /molkit Predicting Genes The most important analysis carried out on DNA sequences is gene prediction Gene prediction requires different methods for eukaryotes and prokaryotes Most gene-prediction methods use hidden Markov Models Predicting Genes in Prokaryotic Genome In prokaryotes, protein-coding genes are uninterrupted No introns Predicting protein-coding genes in prokaryotes is considered a solved problem You can expect 99% accuracy Finding Prokaryotic Genes with GeneMark GeneMark is the state of the art for microbial genomes GeneMark can Find short proteins Resolve overlapping genes Identify the best start codon Use /GeneMark Click the “heutistic models” Predicting Eukaryotic Genes Eukaryotic genes (human, for example) are very hard to predict Precise and accurate eukaryotic gene prediction is still an open problem ENSEMBL contains 21,662 genes for the human genome There may well be more genes than that in the genome, as yet unpredicted You can expect 70% accuracy on the human genome with automatic methods Experimental information is still needed to predict eukaryotic genes Finding Eukaryotic Genes with GenomeScan GenomeScan is the state of the art for eukaryotic genes GenomeScan works best with Long exons Genes with