This track shows predictions from -the FirstEF +the FirstEF (First Exon Finder) program.
Three types of predictions are displayed: exon, promoter and CpG window. If two consecutive predictions are separated by less than 1000 bp, FirstEF treats them as one cluster of alternative first exons that may belong to same gene. The cluster number is displayed in the parentheses of each item. For example, "exon(405-)" represents the exon prediction in cluster number 405 on the minus strand. The exon, promoter and CpG-window are interconnected by this cluster number. Alternative predictions within the same cluster are denoted by "#N" where "N" is the serial number of an alternative prediction in the cluster.
Each predicted exon is either CpG-related or non-CpG-related, based on a score of the frequency of CpG dinucleotides. @@ -30,25 +30,25 @@ as a decision tree. The probabilistic models are optimized to find potential first donor sites and CpG-related and non-CpG-related promoter regions based on discriminant analysis. For every potential first donor site (GT) and an upstream promoter region, FirstEF decides whether or not the intermediate region can be a potential first exon, based on a set of quadratic discriminant functions. FirstEF calculates the a posteriori probabilities of exon, donor, and promoter for a given GT and an upstream window of length 570 bp.
For a description of the FirstEF program and the underlying classification models, refer to Davuluri et al., 2001.
The predictions for this track are produced by Ramana V. -Davuluri of Ohio State University and Ivo Grosse and +Davuluri of Ohio State University and Ivo Grosse and Michael Q. Zhang of Cold Spring Harbor Lab.
Davuluri RV, Grosse I, Zhang MQ. Computational identification of promoters and first exons in the human genome. Nat Genet. 2001 Dec;29(4):412-7.