LinearSampling

LinearSampling samples RNA secondary structures to capture alternative equilibrium foldings for long RNA sequences such as the 29,903-nucleotide SARS-CoV-2 genome.

Key Features:

• LazySampling (on-demand caching): Uses on-demand caching to avoid redundant calculations during sampling.
• Linear Time Complexity: Implements an end-to-end linear-time sampling algorithm that scales linearly with sequence length.
• Empirical Efficiency: Demonstrates a 428× speedup over RNAsubopt when sampling the full 29,903-nt SARS-CoV-2 genome (72 seconds versus 8.6 hours).
• Accuracy Benchmarking: Sampled structures from nine diverse RNA families correlate more closely with established secondary structures than those produced by Vienna RNAsubopt and RNAplfold.

Scientific Applications:

• Genomic analysis of SARS-CoV-2: Enables sampling of full viral genomes to study structural ensembles across long RNA sequences.
• Experimental correlation: Produces sampled structures that align with experimentally guided secondary-structure models and identified alternative conformations.
• Identification of therapeutic targets: Detects regions of high accessibility, exemplified by 23 regions of 15 nucleotides each in the SARS-CoV-2 genome proposed as diagnostic or therapeutic targets.

Methodology:

Combines LazySampling's on-demand caching to eliminate redundant computations with an end-to-end linear-time sampling algorithm (LinearSampling).