Deep forest ensemble learning for classification of alignments of non-coding RNA sequences based on multi-view...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Non-coding RNAs (ncRNAs) are vital in biological processes, but their functions remain largely uncharacterized.
• Accurate classification of ncRNAs is essential for understanding their diverse roles.
• Computational methods are increasingly important for automated and precise ncRNA classification.

Purpose of the Study:

• To develop a novel deep fusion learning framework, GcForest fusion method (GCFM), for classifying ncRNA sequence alignments.
• To improve the accuracy of ncRNA classification and clustering.
• To enhance the prediction of RNA interactions and phylogenetic relationships.

Main Methods:

• Integration of a convolutional neural network and a multi-grained cascade forest (GcForest) algorithm.
• Development of a multi-view structure feature representation including sequence-structure alignment, structure image, and shape alignment encoding.
• Application of GCFM for pairwise ncRNA sequence classification, ncRNA family clustering, phylogenetic tree construction, and RNA interaction prediction.

Main Results:

• GCFM achieved a 6% F-value improvement over existing alignment-based methods for pairwise ncRNA sequence classification.
• GCFM demonstrated a 20% accuracy improvement in ncRNA family clustering compared to RNAclust, Ensembleclust, and CNNclust.
• GCFM achieved 90.63% accuracy in predicting RNA interactions and correctly positioned most ncRNAs in a constructed phylogenetic tree.

Conclusions:

• GCFM offers a powerful and accurate deep fusion learning approach for ncRNA classification and clustering.
• The framework enhances the understanding of ncRNA functions, relationships, and interactions.
• A publicly available web server and source code facilitate the accessibility and application of GCFM in biological research.