Applied Bioinformatics Group

A   A   A
Home > Software > FRED > Examples

Skip to content. | Skip to navigation


Here we shortly describe the examples included with the FRED package. The examples can be found in FRED/examples. All required input files can also be found in this directory.

    In this example we demonstrate the different ways to obtain protein sequences:
    1. Sequences from a list of strings
    2. Sequences from a fasta file
    3. Sequences from Swiss-Prot
    4. Sequences from RefSeq

    In this example we demonstrate how to predict conserved epitopes.


    In this example we demonstrate how to make consensus predictions.


    In this example we demonstrate how FRED can deals with polymorphic protein sequences. The polymorphic sequence is given in a special polymorphic fasta format. The polymorphic positions and the observed amino acids are specified in the header. The sequence is given as reference sequence.

    > sequence | [5:A,G;10:R,M,V]

    codes for

    FRED can treat the polymorphic sequence as non-polymorphic sequence. Then, the reference sequence (the one specified in the fasta file) is used as normal protein sequence, the polymorphisms specified in the fast header are ignored.

    Alternatively FRED can consider all polymorphic peptides resulting from the polymorphic sequence. Both alternatives are shown in the example.


    In this example we show how FRED can be used to assess the performance of prediction methods.

    1. Assess the performance w.r.t. binary binding information (0 for non binding peptide, 1 for binding peptides).The binary binding information is read in from a file (tab delimited: peptide sequence -tab- binding[0|1]).
      Predictions can be made for the peptides using different prediction models. These predictions can then be compared to the original values. This performance analysis is threshold dependent! The predictions results have to be filtered using one of the threshold methods first.

      The following performance measures are available:
      • Matthews Correlation Coefficient (MCC)
      • Accuracy (ACC)
      • Sensitivity (SE)
      • Specificity (SP)
      • Area under the ROC curve (AUC)

      Additionally, the threshold that would yield the highest MCC on this dataset is given, along with the corresponding ACC, SE, and SP.

    2. Compute the correlation between predictions.

      FRED can be used to compute the correlation between prediction methods or between experimental values and prediction values. The reference prediction values can be read in from a file (see input_example4_scores.txt for details). The pairwise correlation (Pearson's Correlation Coefficient or Rank correlation) between these reference values and predictions can be computed.


    In this example we show how FRED can be used to solve a typical problem in vaccine design against highly variable viruses. We search for epitopes that are highly conserved in a set of HCV core protein sequences.
    This application of FRED is based on the paper by Toussaint et al. (2008, PLoS Comput Biol, 4(12), e1000246).