RNA Secondary Structure Prediction with Mfold

IGS 350/550 Computer Laboratory

M. Weir / M. Rice / D. Krizanc

(modified from a module developed by Kelly Thayer)


       Mfold program http://frontend.bioinfo.rpi.edu/applications/mfold/cgi-bin/rna-form1.cgi



  • Predict the secondary structure of  tRNAPhe molecule
  • View the secondary structure prediction
  • Compare the standard free energies of several alternate sub-optimal foldings
  • Obtain the standard free energy contributions
  • Generate dot plots of several alternate RNA foldings
  • Compare your prediction to the crystal structure using RasMol to view the structure
  • Make your own RNA structure prediction

RNA molecules play several important roles in the cell, including

  • carrying the genetic messages transcribed from DNA (mRNA)
  • transferring amino acids to the ribosome to incorporate the appropriate amino acid to a growing polypeptide chain (tRNA)
  • structural role in the composition of the ribosome (rRNA) and spliceosome (small nuclear RNA; snRNA)
  • various catalytic activities

RNA secondary structures play important roles in these functions. In class, we will discuss the Nussinov algorithm for predicting RNA structures. In today's lab, we will use Zuker's Mfold algorithm to predict structures of a tRNA. Then the predicted structures will be compared to the known crystal structure.

Step 1. Manual Prediction of Secondary Structure

See if you can predict possible secondary structures of the following portion of tRNAPhe: UCCUGUGUUCGAUCCACAGAA. Predicting by hand the stem-loop structure of this short sequence is fairly easy. However, predicting longer sequences is harder.

Step 2.  Using Mfold to Predict Secondary Structure

To fold a longer RNA sequence, go to http://frontend.bioinfo.rpi.edu/applications/mfold/cgi-bin/rna-form1.cgi and follow these instructions:

  • Enter a name for your sequence (using a helpful, systematic naming convention)
  • Copy the RNA sequence into the sequence entry box. For example, the sequence of tRNAPhe is


  • Select no constraints.
  • Specify that the RNA sequence is linear, not circular.
  • Use default values for the rest of the form, i.e. for percent suboptimality, upper bound, window, maximum distance between paired bases, etc.

For small RNA sequences, you can perform an immediate calculation that displays the outputs in your browser's window. Also, you can modify the output display features as you like, e.g. image resolution, structure format, base numbering frequency, and structure annotation.

At the bottom of the form, click the Fold RNA button.

The output should look like the following with three predicted RNA structures.

Step 3. Evaluating the Folding Results

Scroll down to the list of structures where several files are presented for each structure.   For example, the .ct file contains a listing of the base pairs (columns 1 and 5).

The free energy values (dG) determine how energetically favorable each of the predicted structures are.  The more negative the number, the more favorable.  Positive numbers are not favored, whereas negative ones are.  Which of your three structures is the most stable and by how much?

To look at a predicted structure, click on jpg.   You can adjust the output from there, with the same options that are available on the query page. [The results will appear in a new window which may be on top; otherwise, look for it on the menu bar at the bottom of your screen.]

How do the structures predicted by Mfold compare to your manual predictions (from Step 1)?

Set the option button and then click on the image to redraw the structure with the following options:

  • Magnification "1"
  • Image Width "936"
  • Output "jpg"
  • Annotation "ss-count"
  • Annotation Type "Both"

Clicking on jpg also opened a window called Loop Free Energy Decomposition.  The Helix values are the sum of their stacking interactions.  What is the single most stabilizing interaction?  Which interactions are destabilizing?

You can also compare the dot plots for the structures.  At the bottom of the output page, go to the Dot plot folding comparison for tRNA and select the following options:

  • Dot plot folding comparisons for tRNA:
  • Image format: "jpg"
  • Compare selected foldings: check each of 1 2 3
  • Then click Do the Comparison button.

This shows the dot plot for the RNAs.  The color coding shows which structure the dot pertains to, including overlap possibilities.  (The magnification options should not be needed here because the RNA is small enough that all dots are clearly visible.)

Adjust the percent suboptimality parameter until you find two other structures that are within 3 kcal of the most stable current structure.  Could some of these alternate structures be the functional one (or ones) ?  The secondary structure prediction with the smallest free energy is not necessarily the functional conformation of the RNA.

In our example, some of the tRNA nucleotides are covalently modified (see Alberts et al. and Lodish et al.). How are the predicted structures affected if we use the following unmodified sequence ?


Step 4. Comparison of 2D prediction to a Crystal Structure

We will now look for the 3D structure in the Nucleic Acid Database at Rutgers and compare the secondary and tertiary structures.  From the NDB homepage at http://ndbserver.rutgers.edu/, search for the NDB ID TRNA06 -- this is the ID for PHE tRNA in yeast. In the "Coordinates + Structure Data" section, click on "Biological Unit coordinates (PDB format)". This retrieves the information about the crystal structure in a PDB format. Paste the text into a Notepad file and save it with the name 1TRA.pdb.

[You can alternatively obtain the pdb file using Protein Explorer or from NCBI
For example, you can search under "Structure" at NCBI using the pdb code "1TRA" (the pdb file code for yeast PHE tRNA) Run the Protein Explorer application at proteinexplorer.org. The pdb file can be run using RasMol.]

The following instructions show how to use this file to view the molecule in RasMol. (The RasMol is available on campus, or you can download your own copy from http://www.umass.edu/microbio/rasmol/.) [Also, with appropriate MS Explorer configuration, you can use the Protein Explorer web interface to view structures instead of RasMol.  Protein Explorer may not be as widely available on campus, but it can be run from the Protein Explorer web server.]

Using RasMol

Run the RasMol application and open your file by dragging its icon over the black display box.  There is one difference in the best structure of your tRNA and the crystal structure - do you notice any base pair that Mfold missed?  Check what was observed in the crystal structure.  In RasMol, you have to open the command line by clicking on the menu bar at the bottom of your screen the icon called RasMol command line.  Then you can type some commands in conjunction with using the pull-down menu to adjust your view. 

To obtain a labeled view of the structure in RasMol, select the Display option "Wireframe", the Colours option "Group", and the Options "Labels". You can see how the clover leaf structure folds on itself -- two of the stem loops fold back towards the main structure.

In addition, here are some useful commands in Rasmol that you can use in comparing the known tertiary (3D) structure of the tRNA and the predicted secondary (2D) structures. Equivalent operations are also possible in Protein Explorer.

       Select different display modes for the whole structure

       Go to the RasMol window, and using the pull-down menus, do edit>select all.  Then go to Display and select a new display.

       Select and Highlight residue 3

The counting starts from the 5'end.  In the command line, type "select residue 5" (or "select 5").  Now color it yellow by typing "color yellow".  You can change how only that residue is displayed by going to the pull-down menu and clicking Display and make your selection.  Commands in RasMol apply to whatever is selected.

       Show only part of the molecule

Use the restrict command in the command line.  If you want to see just residues 1 to 8 and 66 to 76, separate the arguments (things the command should execute on), type restrict 1-8,66-76.  The rest of the molecule is not visible.  To get it back, from the pull down menu, edit>select all, then pick a way to display it.  To add something to the restricted part, select it as above and display it using the pull-down menu.

The preceding figure demonstrates what you can do with a combination of these commands.

By selectively coloring the end points of predicted stems, compare the predicted base pairing with the actual base pairing. How do the actual and predicted structures differ?

[From Alberts et al. Molecular Biology of the Cell]

After looking at the crystal structure, can you speculate why the structure predicted by Mfold differs from the crystal structure ?

Step 5. Use Another RNA Sequence

You can use the list of RNA databases at the beginning of this lab to search for other RNA molecule sequences.   Find a molecule of interest to you and answer the following questions:

  • What is the name of the molecule and where did you find its sequence?
  • What is the standard free energy of this molecule?
  • How many alternate structures can Mfold find within 3 kcal of  this structure?
  • Which structure do you think is the functional one?  Save a picture of this molecule.
  • Using the dot plot, would you say that the molecules found within 3kcal are similar to each other in secondary structural elements or not?

For example, you could examine yeast 5S ribosomal RNA


Compare your predicted structure with the following accepted 2D structure from structural studies (http://www.rna.ccbb.utexas.edu/).

Step 6. Review of Lab Objectives

  • Predict the secondary structure of  tRNAPhe molecule
  • View the secondary structure prediction
  • Compare the standard free energies of several alternate sub-optimal foldings
  • Obtain the standard free energy contributions
  • Generate dot plots of several alternate RNA foldings
  • Compare your prediction to the crystal structure using RasMol
  • Make your own RNA structure prediction


  1. Print the picture with the best prediction for tRNAPhe and label the picture with energy values listed for helices and loops.
  2. Draw your optimal predicted secondary structure for tRNAPhe above using the (i) circle and (ii) parens representations.
  3. Do the modeling algorithms always predict structures correctly? If not, what is the value of these algorithms? Can you imagine ways to improve the algorithms -- what additional information might you incorporate?

Copyright Wesleyan University 2008