From: In silico methods for predicting functional synonymous variants
Prediction Algorithms | Tool | Input | Output | Special features | Notes | URL | Ref |
---|---|---|---|---|---|---|---|
Free energy minimization | CoFold | Single nucleotide sequence: limit of 50 kb; 1500 nt sequence has a run-time of approximately ~ 15 s | Structure diagram + predicted MFE (visualized as an arc plot and supports many other output formats) | Two different thermodynamic parameter options + scaling choices | Algorithm considers co-transcriptional folding to improve accuracy of predicting structure of longer sequences | [74] | |
remuRNA | Wild-type and mutant sequence (no upper limit) | Structure diagram + predicted MFE; relative entropy plot | Algorithm incorporates relative entropy between Boltzmann ensembles of wild-type and mutant secondary structures | https://github.com/bgruening/galaxytools/tree/master/tools/rna_tools/remurna | [72] | ||
RNAfold | Single nucleotide sequence: 7500 nt limit for partition function calculations; 10,000 nt limit for free energy minimization prediction | Interactive RNA secondary structure plot; mountain plot | Parameter options to avoid isolated base pairs, to use partition function, and/or exclude GU pairs at end of helices | Algorithm employs partition function calculations in addition to free energy minimization | [75] | ||
UNAFold | Single nucleotide sequence or multiple short sequences (no upper limit) | Predicted MFE; circular structure plot; energy dot plot | Parameter constraint options to optimize loop types, base numbering frequencies, regularization angles | Uses free energy minimization program with folding temperature fixed at 37 °C | [71] | ||
Pseudoknots | IPKnot | Single or multiple sequences (FASTA format or multiple sequence alignments) | 2D diagram using VARNA program and structure as an arc plot | Options between multiple scoring models and prediction complexity levels | Uses integer programming to compute the maximum expected accuracy structure (MEA) | [81] | |
Kinefold | Single sequence (no upper limit) | Lowest free energy structure diagram + predicted MFE; folding path movie; helix tracing graph | Stochastic simulation—co-transcriptional folding or renaturation folding | Stochastic folding simulations using folding dynamic algorithms [99] and physical constraint modeling for pseudoknot prediction | [73] | ||
Knotty | Single sequence (no upper limit) | Structure diagram + predicted MFE; provides information on all candidate structures | Predicts complex pseudoknot structures with optimization of run-time through sparsification technique and a CCJ-type algorithm | [82] | |||
LandscapeFold | A list of sequences (up to 2), option to consider intramolecular pseudoknots and define minimum number of nucleotides within each hairpin | Identifies all possible structures and provides indexing/sorting via MFE and equilibrium probabilities | Multiple sequence structural analysis for predicting base interactions with option to assess equilibrium concentrations | Polymer physical model based on entropy calculations of arbitrary pseudoknotted structures | [83] | ||
ProbKnot | Single sequence (no upper limit) | Base pair probability plot | Optimization of iterations and minimum helix length | Predicts for presence of pseudoknots in sequence | https://rna.urmc.rochester.edu/RNAstructureWeb/Servers/ProbKnot/ProbKnot.html | [80] | |
Noncanonical base pairings | CycleFold | Single or multiple sequences (no upper limit), can apply maximum expected accuracy (MEA) or ProbKnot to generate structures | Lowest MFE structure, matrix of pairing probabilities between each nucleotide sequence | TurboFold mode can be engaged to process multiple sequences, considers evolutionary conservation | Uses nucleotide cyclic motifs to predict noncanonical base pairings and minimizes free energy | [85] | |
MC-Fold-DP | No sequence limit, but runtime scales polynomially | Returns all structures within energy band above the ground state | Cannot currently consider pseudoknots | Prediction based on combining small nucleotide cyclic motifs | [84] | ||
Machine-learning | DMfold | Single sequences (no upper limit) | Folded RNA structure and energy model | Folding parameters automatically determined based on deep learning | Deep-learning and improved base pair maximation principles; trained with 3948 known RNA primary sequences [100] | https://github.com/linyuwangPHD/RNA-Secondary-Structure-Database | [86] |
SPOT-RNA | Single sequence (maximum—2000 nts); can run longer sequences or batch sequences locally | 2D plots of structure through VARNA visualization tool, output of secondary structure motifs can be seen through Vienna format | Deep contextual neural network implemented with model training and transfer learning from high quality datasets of > 10,000 RNA structures; trained with bpRNA [101] and PDB [102] databases |