PconsC and PconsC2

• Maja Gocevska (gocemaja@ulb.ac.be
• Yannick Jadoul (yajadoul@vub.ac.be)
• Inez Van Laer (ivlaer@vub.ac.be)

Python environment

This implementation of the PconsC and PconsC2 methods is written in Python, version 2.7. It needs the scipy, numpy, scikit-learn and matplotlib libraries to be installed.

Data

The data is assumed to be ordered in the following way, within the data folder:

data/
  sequence_names
  sequences/
    <sequence_name>.fa
  folds/
    set<i>
  contacts/
    <sequence_name>.CB
  <method>/
    <sequence_name>.pred
  alignments/
    <sequence_name>.fa
  psipred/
    <sequence_name>.pred
  netsurfp/
    <sequence_name>.pred

The methods are assumed to be formed as alignment_E-value_prediction-method, see constants.py.

Running

Both pconsc.py and pconsc2.py can be called with the following parameters:

• --data: the data folder
• --intermediate: a folder to save intermediate data
• --results: the folder where results are saved
• --cores: the number of cores that can be used
• --fold: the fold to leave out when training and to use as test data

Other important parameters of the methods can be changed in the constants.py source file.

When running pconsc.py, the folder pconsc should exist in the results folder. For pconsc2.py, pconsc2_layer_ has to exist for all layers for 0 to the number of extra layers.