RNAplfold
v2.7.2Calculate local RNA base-pair probabilities.
Calculate local RNA base-pair probabilities.
RNAplfold is a local RNA secondary-structure program in the ViennaRNA Package. Rather than choosing one structure for an entire transcript, it calculates average base-pair and unpaired-region probabilities across overlapping sequence windows. The result is an accessibility profile, useful when a long RNA contains local regulatory elements, antisense target sites, or short regions whose availability matters more than its full-length fold.
Local folding avoids treating distant regions of a large transcript as though they must fold together. A base pair is considered only within the selected maximum span, and its probability is averaged across the windows that contain it. RNAplfold therefore describes a thermodynamic ensemble of local, nested secondary structures. It does not predict pseudoknots, tertiary contacts, protein binding, or cellular RNA remodeling.
The unpaired probabilities are particularly useful for assessing whether a contiguous stretch may be available to bind another RNA, an oligonucleotide, or a protein. For a global minimum-free-energy structure instead, use RNAfold. For a discrete list of locally stable stem-loops, use RNALfold.
ProteinIQ runs RNAplfold on an RNA sequence provided as FASTA, plain text, or an uploaded text file. Select the local window, maximum pairing span, and largest unpaired region of interest, then receive per-position local accessibility values and the native ViennaRNA output files. Optional opening-energy and RNAplex-compatible output modes are available for accessibility-focused workflows.
| Input | Description |
|---|---|
RNA Sequences | RNA in FASTA or plain-sequence form. Paste text or upload .fasta, .fa, or .txt. FASTA headers are retained as sequence identifiers in the results. A submitted FASTA record may contain multiple sequences, which RNAplfold processes independently. |
Each job accepts up to 30,000 nucleotides across at most 2,000 sequences. Split larger inputs across several jobs.
RNA sequences are expected to use standard nucleotide notation. The ViennaRNA engine accepts T by converting it to U under its standard input handling. Each record represents one strand. RNAplfold measures intramolecular local folding, not hybridization between separate strands.
| Setting | Description |
|---|---|
Window size | Width of the overlapping folding windows, from 20 to 200 nt, default 70. Pair and accessibility probabilities are averaged over windows of this size. Choose a window at least as large as the local feature under study. |
Maximum span | Largest permitted separation between paired nucleotides. A pair (i, j) is excluded when j - i exceeds this value. Leave it blank for RNAplfold's default, which equals the window size. A value above the window size is reduced to it. |
Unpaired region length | Largest contiguous unpaired interval to calculate, from 1 to 50 nt, default 31. RNAplfold computes probabilities for regions of every length from 1 nt through this value. |
Cutoff | Minimum average base-pair probability reported in the native dot plot, default 0.01. It does not act as an accessibility threshold for the _lunp table. |
Opening energies | Writes the native opening-energy-oriented profile (_openen) instead of the standard unpaired-probability profile (_lunp). The ViennaRNA manual defines these values as logarithmic probability transforms, not literal mean energies required to unfold a site. |
RNAplex output mode | Creates the additional local-accessibility files used by RNAplex for target-search workflows. Off by default. |
Temperature (°C) | Temperature used to rescale the thermodynamic energy parameters, from 0 to 100 °C, default 37. It should reflect the experimental or biological condition being modeled. |
ProteinIQ passes these values to RNAplfold unchanged. For a sequence shorter than the window, RNAplfold itself reduces the window, and any longer span or unpaired length, to the sequence length and notes this in its log. A 20 nt transcript therefore cannot yield a 31 nt accessibility interval. The table keeps every value at the precision RNAplfold writes, including very small probabilities such as 2.9e-06.
The default table is parsed from the native _lunp file. It contains one row per nucleotide position, with each value describing the probability that an interval ending at that position is entirely unpaired.
| Column | Meaning |
|---|---|
Sequence ID | FASTA identifier or generated sequence name. |
Position | One-based rightmost position of the reported interval. For a 5 nt value at position 20, the interval is positions 16 through 20. |
P(unpaired 1nt) | Probability that the nucleotide at Position is unpaired in the local folding ensemble. |
P(unpaired 5nt), P(unpaired 10nt), P(unpaired 15nt), P(unpaired 20nt), P(unpaired 31nt) and other lengths | Probability that the interval of that length ending at Position is entirely unpaired. One column is written for every length from 1 nt up to Unpaired region length; lengths without a configured label get an automatic column label. |
When Opening energies is enabled, the table uses the corresponding Opening energy (n nt) columns parsed from _openen. These transformed values are most useful when a downstream workflow expects ViennaRNA accessibility output. They should not be interpreted as independently measured unfolding free energies.
An interval cannot end before it is long enough: the 20 nt interval ending at position 10 does not exist, so RNAplfold writes NA and the cell is left empty in the first rows of each sequence. In the _openen output, NA also replaces a probability of exactly 0, for which the logarithm is undefined.
The Files tab holds two files for the whole job. rnaplfold_files.zip keeps the files RNAplfold writes for each sequence, unchanged and named after its FASTA header:
_lunp: plain-text matrix of unpaired-region probabilities in the standard mode._openen: opening-energy-oriented accessibility matrix when that mode is selected._dp.ps: PostScript dot plot of local base-pair probabilities above the selected cutoff.RNAplex output mode is selected.table/rnaplfold_rows.json holds every row of the results table as JSON. If very long FASTA headers make the table too large to display in full, the table shows the leading rows with a notice, and this file keeps the rest.
RNAplfold uses a partition-function calculation with the ViennaRNA nearest-neighbor energy model. Instead of retaining a single lowest-energy structure, the partition function sums the Boltzmann weights of the possible local secondary structures, producing a probability for each permitted base pair and for each unpaired interval.
The calculation is repeated conceptually across overlapping windows. A pair or unpaired segment can appear in more than one window, so RNAplfold averages its probability across the windows that contain it. With sequence length and window size , the algorithm uses memory and time. Keeping modest makes it suitable for scanning long transcripts while retaining sensitivity to local structures.
Changing Window size and Maximum span changes the question being asked. A 70 nt window asks which interactions are supported by nearby sequence context; a larger setting permits more distant local partners. It does not make RNAplfold a global-folding method. If long-range structure is the hypothesis, compare the relevant region with a global RNA secondary-structure prediction.
High P(unpaired n nt) identifies a region that is frequently unpaired across the local thermodynamic ensemble. It is a stronger condition as n increases: a site may have a high one-nucleotide accessibility yet a low probability that all 20 nucleotides around it are simultaneously open. Match Unpaired region length to the footprint that needs to bind, such as the length of an oligonucleotide seed or an RNA interaction segment.
Low accessibility can arise because the local ensemble supports a stable stem, but it is not direct evidence that a protein or another RNA cannot bind. Ionic conditions, ligands, RNA-binding proteins, chemical modifications, co-transcriptional folding, and tertiary structure are outside the standard calculation. For an interaction score that explicitly combines hybridization with the cost of exposing an interaction site, use RNAup.
RNAplfold probabilities are comparable only when the temperature, window, span, and energy model are held constant. Comparing raw probabilities across settings can mistake a change in modeling assumptions for a biological change in accessibility.
Take a 3,000 nt mRNA and a 20 nt antisense oligonucleotide. Unpaired region length must be at least 20, because the 20 nt column is only written when that length is requested. The default 31 covers footprints up to 31 nt, and a lower value gives smaller tables. Keep it at or below Window size, since a window cannot hold a longer unpaired stretch.
Window size and Maximum span set how far apart two nucleotides may pair. The ViennaRNA defaults are a window of 70 nt and a span equal to the window. Lange et al. (2012) compared settings against transcriptome-wide enzymatic probing data and curated cis-regulatory elements.
They found local folding more accurate than the classic global approach, a span of 150 a reasonable balance between correct and spurious long-range pairs, and fewer errors at artificial sequence ends when the window is clearly larger than the span.
In ProteinIQ a blank Maximum span equals the window, so following that finding means setting both explicitly, for example Window size 200 and Maximum span 150. That study scored structure prediction in mRNAs, not siRNA or oligonucleotide efficacy, so treat 150 as a starting point and keep the same values when comparing candidate sites.
A P(unpaired 20nt) of 0.9 at Position 350 means that 90% of the local ensemble weight has nucleotides 331 to 350 all unpaired. A value of 0.1 means only 10% does. Both are averages over the overlapping windows that contain the interval. The cost of opening the site is the same quantity on an energy scale, in kcal/mol, which is what the Opening energies option writes (read from the ViennaRNA 2.7.2 source).
P(unpaired) | Opening energy at 37 °C (kcal/mol) |
|---|---|
| 0.99 | 0.01 |
| 0.9 | 0.07 |
| 0.5 | 0.43 |
| 0.1 | 1.42 |
| 0.01 | 2.84 |
| 0.001 | 4.26 |
Kertesz et al. (2007) scored microRNA targets as the duplex free energy minus the energetic cost of unpairing the target site, so each kcal/mol of opening cost removes a kcal/mol of binding gain. On that scale 0.9 and 0.1 differ by about 1.4 kcal/mol. No universal cutoff exists: rank sites within the same transcript and settings rather than applying a fixed number.
As a rule, a longer interval is less accessible than a shorter one ending at the same position, because every nucleotide in it must be open at once. A transcript with many 0.8 values at 1 nt may therefore have no 20 nt stretch above 0.2.
Pick the column that matches the footprint, then rank Position by that column in the results table or in the downloaded _lunp file. Position is the last nucleotide of the interval, so a site of n nucleotides starts at Position minus n plus 1. Prefer a plateau of neighbouring positions over a single spike, because adjacent intervals share most of their nucleotides.
For siRNA and microRNA sites, accessibility adds to complementarity rather than replacing it. Tafer et al. (2008) combined target-site accessibility with known siRNA design criteria in the RNAxs tool, and Kertesz et al. (2007) found that mutations lowering accessibility reduced microRNA-mediated repression about as much as mutations that disrupt complementarity.
Check the hybridization energy of candidate sites with RNAduplex, RNAup or, for long targets, RNAplex using the RNAplex output mode files.
For a ribosome binding site, accessibility of the Shine-Dalgarno sequence is described as essential for translation to start in prokaryotes (Bernhart et al., 2011). Include the 5' UTR and a stretch of coding sequence, choose n to span from the Shine-Dalgarno motif through the start codon, and read the row at the last nucleotide of the start codon. Compare UTR variants with identical settings.
RNAplfold (Bernhart et al., 2006), distributed in the ViennaRNA Package (Lorenz et al., 2011), restricts pairing to a local span. On a 3,000 nt mRNA, a global fold can pair a 5' UTR nucleotide with one 2,000 nt downstream, and every such partner competes in the ensemble. With a span of 70 or 150, RNAplfold cannot form that pair, so a region that looks buried in the global structure can look open locally.
RNAfold with the partition function reports whole-molecule ensemble metrics, not unpaired probabilities for every interval. RNAplfold values are averages over windows and do not describe a single ensemble of the whole molecule.
Position as the start or the centre of the interval. It is the rightmost nucleotide.P(unpaired 1nt) as the accessibility of a multi-nucleotide site. Use the column for the full footprint.NA as zero accessibility. It means the interval does not exist or, in _openen, the probability was 0.P(unpaired)._dp.ps dot plot as accessibility. It shows averaged base-pair probabilities above Cutoff, not unpaired intervals.Calculate RNA density-of-states energy distributions.
Evaluate RNA secondary-structure free energy.
Predict RNA secondary structures with ViennaRNA. Get MFE structures and plots, or enable the partition function for ensemble results. Sign in to run your sequences.
Find locally stable RNA structures in long sequences.
Enumerate suboptimal RNA secondary structures.
Predict and analyze RNA secondary structures.
Predict RNA-RNA dimer secondary structures.
Automated atomic model building for cryo-EM maps
Compare RNA secondary structures by base-pair and tree distance.
Predict RNA duplex hybridization between two sequences.