viennarna-structure-prediction

What it does

Predicts RNA secondary structure, MFE folding, base-pair probabilities, and RNA-RNA interactions using ViennaRNA bindings. Provides a pipeline: sequence → MFE → partition function and pair-probability matrix → dot-bracket → duplex. Includes both Python API usage and CLI demonstrations for batch work.

How it works

The skill describes using the ViennaRNA toolkit and Python bindings to perform:

• MFE folding with RNA.fold() or fold_compound.mfe() to obtain a structure and MFE value.
• Partition function computation and retrieval of base-pair probability matrices via fold_compound.pf() and fold_compound.bpp(), with optional rescaling for numerical stability.
• Deriving dot-bracket structures from MFE results and analyzing base-pair probabilities to identify high-confidence pairs.
• RNA-RNA duplex prediction using RNA.cofold() for co-folding between two sequences separated by an ampersand.
• RNA accessibility analysis using a sliding-window approach and per-position unpaired probability from the base-pair probability matrix.
• Command-line batch folding using RNAfold, including parsing its output to extract structure and MFE.
• Constrained folding with hard constraints (hc_add_bp) and enumeration of suboptimal structures via subopt.
• Visualizing the base pair probability matrix as a heatmap and providing sample Python code for batch workflows.

The instructions include concrete code snippets for preparing sequences, creating fold compounds, computing MFE, deriving partition functions, obtaining base-pair probabilities, performing co-folding, and conducting accessibility analyses. It also shows how to execute RNAfold via CLI and parse its output.

Key steps explicitly described include:

• Prepare sequence, normalize (uppercase, T→U).
• Create RNA.fold_compound(sequence) and compute MFE with fc.mfe(); parse structure and mfe.
• Compute partition function with fc.pf() after setting up and optionally fc.exp_params_rescale(mfe).
• Retrieve base-pair probabilities with fc.bpp() and convert to a usable matrix; identify high-confidence pairs (>0.9).
• Build and visualize the base-pair probability matrix with matplotlib.
• Use RNA.cofold(cofold_seq) for duplex prediction and parse the result structure.
• Compute accessibility across a target by deriving per-position unpaired probabilities from bpp and plotting results.
• Run RNAfold from CLI to process sequences in batch and parse outputs.
• Apply constraints with hc_add_bp and enumerate suboptimal structures with RNA.subopt.

When to use it

• When predicting the minimum free-energy (MFE) secondary structure of an RNA sequence.
• When computing base-pair probability matrices to assess structural uncertainty.
• When evaluating siRNA accessibility by folding the target mRNA region for potential exposure.
• When assessing sgRNA targeting efficiency by predicting whether guide RNA forms structures that could impact activity.
• When modeling RNA-RNA interactions via co-folding/duplex prediction.
• When calculating folding energies across multiple sequences for stability comparisons.
• When batch-processing sequences via RNAfold CLI for non-Python workflows.

What it can touch

• Tools explicitly referenced: RNAfold, RNA.fold (via Python bindings), RNA.fold_compound, RNA.cofold, RNA.bpp, RNA.pf, RNA.subopt, RNA.hc_add_bp.
• It demonstrates using Python imports like import RNA and commands such as fc.mfe(), fc.pf(), fc.bpp(), and subprocess-based RNAfold calls for batch processing.

Caveats

• License: MIT for the skill description, but repository license is NOASSERTION per the provided facts.
• It notes dependencies: Python 3.8+, ViennaRNA Python bindings, matplotlib, numpy; conda recommended to manage C library dependencies.
• It emphasizes using Turner's 2004 parameters and advises alternatives (mfold, RNAstructure) for specific algorithms or different param sets.
• It includes sample code that assumes valid input sequences and may require adaptation for large-scale or production use.