CONTACT  |  SITE MAP  |  ABOUT US   
Ask an account
You are here : Home / Home URGI / About us / Publications / Archives / 2011 / Pseudo Agent System for Transposable Element Classification

2011

National,  COM (posters)

CNET, XVIIe edition 4th - 6th july 2011, Lyon France

04 Jul 2011   Pseudo Agent System for Transposable Element Classification
Poster_CNET2011_Arnoux

S. Arnoux, C. Hoede, H. Quesneville

 

Classification of transposable elements (TEs) is an important step for their annotation. Interestingly, this also helps to filter false-positives TEs among the repeats detected by de novo approaches. We develop tools that classify automatically TEs and facilitate their manual curation. A first version of a TE classifier was implemented based on a decision tree. But, in practice, the expert decisions rules do not always follow a decision tree. We rewrote a classifier called PASTEC for Pseudo Agent System for Transposable Element Classification. With this tool, all classifications from Wicker hierarchical system are tested. Each classification is weighted according to the evidences found over those that are expected for the considered classification. Structural features are considered, such as terminal repeats, polyA tail, SSR and ORF. PASTEC can also use homologies with known TEs as in Repbase Update, with HMM profiles, with rDNA and with known host genes. HMM profiles greatly help to classify TEs, as well as potential host gene when a non-TE-specific profile is detected. Its modular multi-agent architecture facilitates the addition of new features (classification or analysis types). Our first benchmarks show significant improvements. We are integrating PASTEC in our TEdenovo pipeline, a tool from the REPET package .

Update: 20 Dec 2013
Creation date: 08 Jul 2011