Dragon Gene Start Finder: An Advanced System for Finding Approximate Locations of the Start of Gene Transcriptional Units

Abstract

We present an advanced system for recognition of gene starts in mammalian genomes. The system makes predictions of gene start location by combining information about CpG islands, transcription start sites (TSSs), and signals downstream of the predicted TSSs. The system aims at predicting a region that contains the gene start or is in its proximity. Evaluation on human chromosomes 4, 21, and 22 resulted in Se of over 65% and in a ppv of ∼78%. The system makes on average one prediction per 177,000 nucleotides on the human genome, as judged by the results on chromosome 21. Comparison of abilities to predict TSS with the two other systems on human chromosomes 4, 21, and 22 reveals that our system has superior accuracy and overall provides the most confident predictions.