Please see Develop for details on the download and installation of GTfold source code. This page contains instructions for running GTfold-related programs.


The RNAStructProfiling code (known as 'profiling') uses GTBoltzmann to sample 1,000 structures, which it then processes into helix classes and profiles. Such a combinatorial view of the 1,000 structures allows for structural signal to emerge from the noisy Boltzmann sample.

If installing profiling on a local machine is not needed, profiling for small RNAs is also available through the webserver.

Downloads and Installation

Go to the Github page. Click on the "Download ZIP" button, and unzip. Navigate to the gtfold-RNAStructProfiling-xxx folder and type

In order to generate samples from gtfold, a library made from the gtfold code and labelled 'libgtfold.a' must be in the profiling directory . A sample gtfold Linux library (libgtfold.a) and a OSX one (libgtfold_osx.a) are included in the zip file. If your operating system is not compatible with these libraries, you may have to make your own. First, download and install gtfold. After running 'make' on gtfold, archive all the *.o files:
ar -cvq libgtfold.a *.o
Copy this newly created libgtfold.a file to the profiling directory.


After installing the source code, users should navigate to the directory containing the RNAprofile executable. This executable can be called from the command line using:


Profiling accepts as input a sequence file in FASTA format, along with various options. Profiling outputs to the command line either a condensed (default) or a verbose (-v option) report concerning helix classes, profiles, and the summary profile graph. It also produces

  • an 'output.samples' file, which is the gtboltzmann output of 1,000 sampled structures.
  • a '' (default, unless alternate name specified) file, which contains the summary profile graph in graphviz dot format
  • and a 'structure.out' file, which records the extended profile (composed of all helix classes) of each structure, and its translation to a profile (composed of features)


option description
-v produces verbose output
-g disables graph making portion of profiling; used when only interested in helix class or profile stats, especially with longer sequences
-e FILE when using a samples file in gtboltzmann format as input to profiling
-sfold FILE when using a samples file in sfold format as input to profiling
-o FILE name of graph file to be produced in graphviz dot format; default is ''
-h PERCENT choose all helix classes as features that occur in at least [h] percent of the input sample structures