ORFhunteR

ORFhunteR identifies open reading frames (ORFs) in human mRNA by applying nucleotide sequence vectorization and random forest classification to assess coding potential across large transcriptomic datasets.

Key Features:

• Machine learning (random forest): Uses a random forest classifier to distinguish true ORFs from multiple candidate ORFs within long human mRNA sequences.
• Sequence vectorization: Transforms nucleotide sequences into numerical feature vectors for machine-learning-based ORF prediction.
• Validation against reference databases: Performance was validated using datasets from NCBI RefSeq and Ensembl, achieving nearly 95% accuracy in identifying true ORFs.
• Scalability: Designed for large-scale, transcriptome-wide ORF analysis rather than single-molecule-only processing.
• Implementation: Implemented as an R and C++ library.

Scientific Applications:

• Transcriptome-wide ORF annotation: Identification of ORFs in whole transcriptome sequencing datasets to annotate coding regions in mRNA.
• Coding potential assessment: Determination of the coding potential of RNA molecules to support studies of gene function and regulation.
• Support for integrative omics: Provides ORF calls useful for downstream genomics, proteomics, and systems biology analyses.

Methodology:

Input nucleotide sequences are vectorized into feature sets; a random forest model is trained on validated datasets (NCBI RefSeq and Ensembl) and the pre-trained model is applied to new mRNA datasets with predictions validated against reference data.