PCirc: random forest-based plant circRNA identification software.

Area of Science:

• Plant molecular biology
• Bioinformatics
• Genomics

Background:

• Circular RNAs (circRNAs) are novel RNA molecules with a closed-loop structure involved in gene regulation.
• Identifying plant circRNAs from RNA-seq data is crucial but challenging due to unique sequence characteristics like splicing signals and transposable elements.
• Existing animal circRNA identification methods are not directly applicable to plants.

Purpose of the Study:

• To develop a machine learning-based software for accurate identification of plant circRNAs.
• To address the limitations of traditional methods and the unique features of plant circRNA sequences.
• To provide a convenient tool for plant circRNA researchers.

Main Methods:

• Extracted features from rice circRNA and lncRNA data, including open reading frames, k-mers, and splicing junction sequences.
• Trained a machine learning model using the random forest algorithm with tenfold cross-validation.
• Developed a localized software package, Pcirc, integrating the trained model and programming scripts.

Main Results:

• The machine learning model achieved high performance with accuracy, precision, and F1 scores above 0.99 on test data.
• The model demonstrated good performance on other plant test datasets, with accuracy scores exceeding 0.8.
• The developed software, Pcirc, is available for local use.

Conclusions:

• A machine learning model for plant circRNA recognition was successfully constructed using rice data.
• The Pcirc software, based on this model, can be applied to various plant species like Arabidopsis thaliana and maize.
• Pcirc offers a user-friendly solution for plant circRNA identification.