NmSEER V2.0: a prediction tool for 2'-O-methylation sites based on random forest and multi-encoding combination

Area of Science:

• RNA biology
• Post-transcriptional modification
• Computational biology

Background:

• 2'-O-methylation (Nm) is a crucial RNA modification found in mRNA and non-coding RNAs, impacting biological processes.
• Nm-seq technology enables high-resolution profiling of Nm sites.
• Previous computational methods for Nm site prediction exist, but require refinement with new data.

Purpose of the Study:

• To develop a more robust in silico model for predicting RNA 2 -O-methylation (Nm) sites.
• To leverage a newly available, high-confidence dataset generated by refined Nm-seq.
• To improve the accuracy and reliability of Nm site prediction.

Main Methods:

• Redesigned the prediction model using machine learning algorithms and multi-encoding schemes.
• Evaluated and optimized models through 5-fold cross-validation and independent testing.
• Selected Random Forest as the most robust algorithm, combined with one-hot encoding, position-specific dinucleotide sequence profile, and K-nucleotide frequency encoding.

Main Results:

• The Random Forest algorithm demonstrated superior performance for Nm site prediction.
• A combination of one-hot encoding, position-specific dinucleotide sequence profile, and K-nucleotide frequency encoding yielded the best predictive model.
• The updated predictor achieved an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.862.

Conclusions:

• The updated predictor, NmSEER V2.0, provides accurate prediction of RNA Nm sites.
• NmSEER V2.0 is available as a free online server for researchers.
• This tool facilitates further investigation into the biological roles of RNA 2 -O-methylation.