Utilizing sequence intrinsic composition to classify protein-coding and long non-coding transcripts