RNAparametric

RNAparametric computes Boltzmann partition functions of RNA secondary structures under explicit global structural constraints using four algorithms (RNAhairpin, RNAmloopNum, RNAmloopOrder, RNAmloopHP) to quantify ensemble properties such as counts of hairpins and multiloops.

Key Features:

• Partition Function Calculation: Implements RNAhairpin (computes partition functions Z(k) for structures with exactly k hairpins, 0 ≤ k ≤ K, with time complexity O(K(2)^n(3)) and space complexity O(Kn^2)), RNAmloopNum (calculates partition functions for structures with exactly k multiloops using similar resources), RNAmloopOrder (determines partition functions based on the maximum order/depth of multiloops), and RNAmloopHP (computes partition functions Z(m, h) for structures with m multiloops and h hairpins with runtime O(M^2H^2n^3) and space O(MHn^2)).
• Sampling from Low-Energy Ensembles: RNAhairpin and RNAmloopHP can sample structures from low-energy ensembles to characterize structural diversity under the specified constraints.
• Optimization via Fast Fourier Transform (FFT): RNAhairpin and RNAmloopNum include FFT-based optimizations achieving time complexity O(n^4) and space complexity O(n^2), while this FFT optimization is not applicable to RNAmloopOrder.

Scientific Applications:

• Improved Structural Predictions: Structures sampled from RNAmloopHP show higher accuracy than minimum free-energy structures for many Rfam RNA families, with reported sensitivity improvements such as 24% for transfer RNA and 5% for certain ribozyme families.
• Feature Discrimination for Machine Learning: Probabilities p(k) = Z(k)/Z for forming k hairpins or multiloops serve as features for binary classifiers (e.g., support vector machines, relevance vector machines) to aid classification of Rfam RNA families.
• Insights into Structural Features: Analysis indicates that multiloop order does not substantially improve discriminatory power over hairpin and multiloop counts, informing feature selection for noncoding RNA gene finders.

Methodology:

RNAparametric is implemented in C/C++.