Long non-coding RNAs (lncRNAs,longer than 200 bp) play important biological roles in dosage compensation,genomic imprinting,and cell differentiation,and have been implicated in human disease such as cancer.Facilitated by high throughput sequencing technology,RNA sequencing has revealed numerous novel noncoding RNAs,the majority of which are long noncoding RNAs.To comprehensively annotate newly discovered transcripts,the first step is to distinguish lncRNAs from protein-coding transcripts[1].Herein we present a novel tool named PLADL for lncRNA identification,which has several advantages over existing tools.First,PLADL incorporates the intrinsic features of transcript sequences,such as ORF length,ORF ratio,and entropy density profiles[2],to build the mathematical model of lncRNAs.Second,PLADL employs deep learning to construct a classification algorithm.Compared with existing tools,CPC,CPAT,CNCI,PLEK,lncRNA-MFDL and lncRScan-SVM,PLADL achieves totally the best accuracy on human training data with 10-fold cross validation (98.2%) as well as on the test set (Table 1).Moreover,PLADL outperforms other tools remarkably on cross-species when we predicted lncRNAs for mouse and zebrafish (Table 2).