NmSEER V2.0

NmSEER V2.0 predicts 2'-O-methylation (2'-O-me or Nm) sites within mRNAs and various non-coding RNAs to identify post-transcriptional modification locations for studies of RNA function.

Key Features:

• Target modifications: Predicts 2'-O-methylation (2'-O-me or Nm) sites in mRNAs and non-coding RNAs.
• Training data: Uses a refined dataset derived from Nm-seq high-resolution profiling that maps Nm sites with single-nucleotide precision and enhanced sensitivity.
• Encoding strategy: Combines one-hot encoding, position-specific dinucleotide sequence profiles, and K-nucleotide frequency encoding to represent sequence features.
• Algorithm: Employs a random forest classifier identified as the most robust method.
• Model evaluation: Assessed by rigorous 5-fold cross-validation and independent evaluations.
• Performance metric: Reports an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.862.

Scientific Applications:

• Identification of candidate Nm sites: Predicts potential 2'-O-methylation locations for downstream experimental or computational analysis.
• RNA biology studies: Supports exploration of the functional implications of 2'-O-methylation in RNA molecules.
• Gene expression regulation research: Facilitates investigation of relationships between Nm modifications and regulation of gene expression and other cellular processes.

Methodology:

Modeling used Nm-seq–derived datasets with feature encodings (one-hot, position-specific dinucleotide sequence profiles, K-nucleotide frequency); a random forest classifier was selected and evaluated by 5-fold cross-validation and independent testing, yielding AUROC=0.862.