Predict RNA secondary structure with ViennaRNA, reporting the Boltzmann ensemble (partition function, base-pair probabilities, centroid, MEA, per-base confidence, stochastic samples) rather than a single minimum free energy (MFE) fold. Covers consensus folding from alignments, SHAPE-constrained folding, RNA-RNA interaction, local and linear-time methods for long RNA, and the cases where pseudoknot-aware or deep-learning tools are needed.
# ViennaRNA (RNAfold, RNAalifold, RNAcofold, RNAduplex, RNAup, RNAplfold, RNALfold, Python API)
conda install -c bioconda viennarna
# Optional: linear-time folding for long sequences
conda install -c bioconda linearfold linearpartition
# Optional: pseudoknot-aware folding
conda install -c bioconda ipknot
# Python dependencies
pip install matplotlib numpy
Tell your AI agent what you want to do: - "Fold this RNA and report the ensemble, not just the MFE" - "How well-defined is this structure? Show me the per-base confidence" - "Build a consensus structure from my alignment of homologs" - "This RNA is 8 kb, fold it sensibly" - "My RNA probably has a pseudoknot, predict it" - "Use my SHAPE reactivities to constrain folding"
"Fold this RNA and report the MFE, centroid, and MEA structures, and tell me which to trust."
"Predict the structure of my 5' UTR and tell me whether the hairpin is well-defined or ambiguous."
"Compute the base-pair probability dot plot and the positional entropy for this RNA."
"Is this RNA likely to adopt a single structure or switch between conformations?"
"I have a Stockholm alignment of my RNA family. Predict the consensus structure with covariation, and tell me whether the covariation is statistically significant."
"Fold this 10 kb mRNA and give me local accessibility, not one global structure."
"Predict the binding between my sRNA and its target mRNA, accounting for whether the target site is accessible."
"Fold my RNA but force positions 15-20 unpaired."
"My RNA is a frameshift element with a pseudoknot; standard folding misses it."
"Use my SHAPE reactivity data to guide secondary structure prediction."
-p) so ensemble quantities are available, not just the MFE-p; the base-pair probability matrix and positional entropy say which parts of the fold are trustworthy. The MFE alone hides the uncertainty and is often not the functional structure.*_dp.ps/*_ss.ps to the working directory; pass --noPS.