Prediction Models for DNA Transcription Termination Based on SOM Networks

V B Bajic; T H Charn; J X Xu; S K Panda; S P T Krishnan

doi:10.1109/IEMBS.2005.1615543

Prediction Models for DNA Transcription Termination Based on SOM Networks

Conf Proc IEEE Eng Med Biol Soc. 2005:2005:4791-4. doi: 10.1109/IEMBS.2005.1615543.

Authors

V B Bajic¹, T H Charn, J X Xu, S K Panda, S P T Krishnan

Affiliation

¹ Inst. for Infocomm Res.

PMID: 17281313
DOI: 10.1109/IEMBS.2005.1615543

Abstract

This paper presents two efficient models for predicting transcription termination (TT) in human DNA. A neural network, self-organizing map, was used for finding features from a human polyadenylation (polyA) sites dataset. We derived prediction models related to different polyA signals. A program, "Dragon PolyAtt", for predicting TT regions was designed for the two most frequent polyA sites "AAUAAA" and "AUUAAA". In our tests, Dragon PolyAtt predicts TT regions with a sensitivity of 48.4% (13.6%) and specificity of 74% (79.1%) when searching for polyA signal "AAUAAA" ("AUUAAA"). Both tests were done on human chromosome 21. Results of Dragon PolyAtt system are substantially better than those obtained by the well-known "polyadq" program.