Method for predicting RNA coding potential

Abstract

The invention belongs to the field of gene annotation and in particular relates to a method for predicting RNA coding potential. The method (named as CPPred) comprises the following steps: by integrating multiple sequence characteristics, particularly describing global distribution of RNA by using CTD; taking redundancy and relevance among candidate characteristics as standards, and combining a characteristic increasing selection method to select an optimum characteristic set to serve as a characteristic vector; establishing a prediction model by a support vector machine (SVM); finally, acquiring the prediction result according to a to-be-predicted RNA sequence characteristic vector. The prediction method provided by the invention is equivalent to a current existing method (having accuracyreaching 90% or higher) while predicting a long RNA sequence, while the method is obviously better than the current existing method while predicting a short RNA sequence.

Description

technical field

[0001] The invention belongs to the field of gene annotation, and more specifically relates to a method for predicting RNA coding potential. Background technique

[0002] In recent years, next-generation sequencing technologies have generated tens of thousands of new transcripts, so quickly and accurately distinguishing coding RNAs from non-coding RNAs (ncRNAs) has become the key to analyzing these data. In organisms, although ncRNAs cannot encode proteins, they also have important biological functions, such as gene regulation, gene silencing, RNA modification and processing.

[0003] In the field of prediction of coding potential, a coding potential assessment tool CPAT using a matchless logistic regression model has been disclosed. It uses 4 sequence features: length of open reading frame, coverage of open reading frame, Fickett score and hexamer score. In this field of prediction, CPC2 is also disclosed, which also uses only 4 sequence features: the leng...

Images