Splice sites identification based on multiclass feature representation

摘要

Accurate splice site identification is a critical component of eukaryotic gene prediction and efficient feature representation plays a key role in the splice sites identification. We propose a splice sites identification algorithm based on multiclass feature representation (denoted as MFR), which uses various features effective for splice sites identification, including position-dependent features, regiondependent motif features, composite-nucleotides features and region-dependent statistical features. The experimental results on HS3D and ?? NN269 acceptor data sets show that our algorithm outperforms current state-of-art algorithms. Besides, our algorithm has lower space complexity and higher time efficiency.