It becomes a hot spot to design various disease specific target penetrating peptide in the process of drug development.Cell penetrating peptides (CPPs) is a small molecule polypeptide (5~30 animo acids) with a cell membrane penetrating ability that can be used as a carrier for efficient delivery of a variety of biologically active substances,including proteins,peptides,nucleic acids,lipids and small molecules,etc.So use of machine learning methods to find potential new CPPs will enable more rapid screening for applications such as drug delivery.However,the short sequences made it difficult to display the intrinsic properties of CPPs.Although various features such as amino acid composition and dipeptide composition have been widely employed in CPPs prediction algorithms,the bad prediction performances imply it is urgent to propose novel approaches for short seqences analysis.In this work,a 40-D weighted vector was derived from a novel cylindrical representation,which are employed as numerical descriptors for CPPs prediction.The results showed that the 40-D vector based CPPs prediction algorithm got better prediction results than traditional feature parameters.Thus,this work can provide helpful tool for CPPs predictions and short sequences analysis.