CPAT

CPAT assesses the coding potential of RNA sequences by classifying transcripts as protein-coding or noncoding using a logistic regression model and sequence-derived linguistic features.

Key Features:

• Logistic regression classifier: Employs a logistic regression model to assign coding versus noncoding labels to RNA sequences.
• Sequence-derived linguistic features: Uses pure linguistic features extracted from RNA nucleotide sequences as input predictors.
• Input types: Analyzes nucleotide sequences or genomic coordinates of RNAs.
• Output scores: Produces coding probability scores ranging from 0 to 1 for each RNA sequence.
• Prebuilt species models: Includes prebuilt logistic models for Homo sapiens, Mus musculus, Danio rerio, and Drosophila melanogaster.
• Model training: Provides procedures to train new logistic models for additional genomes.
• Performance characteristics: Designed for high accuracy and speed in coding-potential classification.

Scientific Applications:

• Coding potential classification: Differentiates protein-coding and noncoding RNAs based on computed probabilities.
• Cross-species analysis: Applies prebuilt models to analyze RNAs from human, mouse, zebrafish, and fruit fly datasets.
• Extension to other genomes: Enables generation of species-specific models to study coding potential in non-model organisms.

Methodology:

CPAT computes sequence-derived linguistic features from nucleotide sequences or genomic coordinates and applies a logistic regression classifier to produce coding-probability scores (0–1); it includes prebuilt logistic models for specified model organisms and supports training new models.