Sfold

Sfold predicts RNA secondary structure ensembles and supports rational design of RNA-targeting nucleic acids such as small interfering RNAs (siRNAs), antisense oligonucleotides, and trans‑cleaving ribozymes.

Key Features:

• Statistical Sampling Paradigm: Utilizes statistical sampling to represent an ensemble of probable RNA secondary structures rather than relying solely on minimum free energy predictions.
• Probability Profile Approach: Computes probability profiles that assign probabilities to single-stranded regions to identify accessible antisense target sites and correlate with experimental inhibition of in vitro translation.
• Boltzmann Ensemble Sampling: Samples from the Boltzmann ensemble by computing equilibrium partition functions with updated thermodynamic parameters and generating statistically representative samples via recursive forward/backward sampling.
• Bayesian Inference: Applies a Bayesian algorithm that relaxes fixed energy parameter requirements to provide exact posterior distributions of secondary structures and related variables.
• Module-specific Outputs: Provides module outputs (Sirna, Soligo, Sribo, Srna) that report statistical representations and design metrics for siRNAs, antisense oligonucleotides, ribozymes, and sampled RNA structures.

Scientific Applications:

• Antisense Oligonucleotide Design: Uses probability profiles to select optimal target sites for antisense oligonucleotides and support high-throughput screening and drug target validation.
• siRNA and Ribozyme Design: Predicts RNA target accessibility and evaluates duplex thermodynamic properties to support design of siRNAs and trans‑cleaving ribozymes for gene knock-down studies.
• Functional Genomics and Drug Discovery: Employs ensemble sampling and probability profiling to explore mRNA structure–function relationships and alternative biological structures relevant to functional genomics and drug discovery.

Methodology:

Computational methods include sampling RNA secondary structures from the Boltzmann ensemble by computing equilibrium partition functions with updated thermodynamic parameters and generating representative structures via recursive forward/backward sampling, together with Bayesian inference to obtain posterior distributions and incorporate parameter uncertainty.