SRAMP

SRAMP predicts N6-methyladenosine (m6A) modification sites in mammalian RNA sequences to enable identification of m6A sites for studies of RNA regulatory processes including degradation, splicing, subcellular localization, and structural changes.

Key Features:

• Predictive capability: Predicts m6A modification sites on RNA sequences with models tailored for mammalian transcriptomes.
• Machine learning architecture: Employs three distinct random forest classifiers that leverage different sequence-context representations.
• Positional nucleotide sequence pattern: Uses the specific nucleotide arrangement around candidate m6A sites as an input feature.
• K-nearest neighbor information: Incorporates information from neighboring sequences to inform predictions.
• Position-independent nucleotide pair spectrum: Considers nucleotide pair interactions without positional constraints as features.
• Input types: Accepts both genomic and cDNA sequences as input.
• Performance validation: Performance has been evaluated using cross-validation and independent benchmarking.
• Feature analysis: Highlights nucleotide usage preferences at proximal and distal positions around m6A sites as important discriminative features.

Scientific Applications:

• m6A site mapping: Enables identification of m6A modification sites across mammalian transcriptomes.
• RNA regulation studies: Supports investigation of how m6A influences RNA degradation, splicing, subcellular localization, and RNA structure.
• Disease and functional research: Facilitates exploration of m6A roles in biological processes and diseases and aids identification of potential therapeutic targets.

Methodology:

Applies three random forest classifiers built on positional nucleotide sequence patterns, k-nearest neighbor-derived features, and position-independent nucleotide pair spectrum features, with performance assessed by cross-validation and independent benchmarking.