PPfold

PPfold predicts consensus RNA secondary structure from multiple-sequence alignments by combining the pfold stochastic context-free grammar with phylogenetic analysis.

Key Features:

• Parallelization and Multithreading: Distributes phylogenetic calculations and the inside-outside algorithm across multiple cores to reduce runtime for large RNA alignments.
• Extended-Exponent Datatype: Employs an extended-exponent datatype to avoid floating-point underflow and enable accurate computation on large-scale alignments.
• Pfold Model Implementation: Implements the pfold model that integrates stochastic context-free grammar with phylogenetic models for comparative secondary-structure prediction.
• Integration of Structure-Probing Data: PPfold 3.0 incorporates flexible probabilistic models to integrate auxiliary data from structure probing experiments into single-sequence and alignment predictions.
• Scalability to Large Datasets: Capable of predicting consensus structures for large alignments, including complete viral genomes and long genomic transcripts.
• Comparable Accuracy: Reports improved single-sequence and alignment prediction accuracy competitive with RNAstructure.

Scientific Applications:

• Comparative RNA secondary-structure prediction: Infers consensus secondary structures from multiple-sequence alignments for evolutionary and structural analyses.
• Analysis of viral genomes and long transcripts: Predicts secondary-structure features across complete viral genomes and long genomic transcripts.
• Integration with experimental probing data: Refines structure predictions by incorporating auxiliary data from structure probing experiments.
• Identification of structural elements: Detects conserved known structural elements and novel RNA features within alignments.

Methodology:

Combines stochastic context-free grammar (pfold model) with phylogenetic analysis, applies the inside-outside algorithm, parallelizes phylogenetic calculations and the inside-outside algorithm across multiple cores, employs an extended-exponent datatype to prevent floating-point underflow, and uses flexible probabilistic models to integrate auxiliary data from structure probing experiments.