RNAsc

RNAsc predicts RNA secondary structure by integrating chemical and enzymatic probing data, including Selective 2′-hydroxyl acylation analyzed by primer extension (SHAPE) and inline-probing, to derive nucleotide-specific pseudo-energy constraints that improve structure prediction.

Key Features:

• Integration of probing data: Incorporates SHAPE and inline-probing reactivity data derived from 2'-hydroxyl footprinting experiments to inform structure prediction.
• RNA soft constraints (RNAsc): Implements pseudo-energy terms assigned to each nucleotide position rather than only to base stacking positions.
• Nucleotide-specific unpaired probabilities: Aligns predicted probabilities of being unpaired for each nucleotide with experimental shape data.
• Self-consistency: Applies constraints in a manner that increases the correlation between predicted secondary structure and input SHAPE data.
• Benchmarking: Evaluated on eight RNAs with available shape data and native structures, matching prior methods in seven cases and yielding a 25% improvement in one case.
• Comparative probing analysis: Enables direct comparison of SHAPE and inline-probing data, exemplified by analysis of yeast aspartyl-tRNA (asp-tRNA) showing greater robustness of SHAPE for that RNA.

Scientific Applications:

• RNA secondary structure prediction: Improves base-pairing prediction accuracy by integrating experimental probing data into thermodynamic models.
• Interpretation of biochemical footprinting: Translates SHAPE and inline-probing reactivities into position-specific energetic constraints for structural inference.
• Method benchmarking and comparison: Provides quantitative comparisons between probing-based prediction methods and native structures across multiple RNAs.
• Case-study analysis: Facilitates analysis of specific RNAs such as yeast aspartyl-tRNA (asp-tRNA) to compare probing techniques.

Methodology:

RNAsc assigns per-nucleotide pseudo-energy terms (RNA soft constraints) across all nucleotide positions and adjusts them to align predicted nucleotide unpaired probabilities with experimental SHAPE/inline-probing data, enforcing self-consistency to improve correlation with input data; the approach was benchmarked against native structures for eight RNAs.