In the Sali Lab¶
If you are working in the Sali lab, you don’t need to build and install
CryptoSite - it is already set up for you as a module. Just run
module load cryptosite to load it.
All dependencies listed below are expected to be found in standard
system paths. This may require setting
LD_LIBRARY_PATH environment variables. Note that Linux is the only platform
for which all these dependencies are available, and so is the only platform
on which CryptoSite currently functions.
- Python 2.6 or later (Python 3 should also be OK).
- MODELLER plus the
soap_protein_od.hdf5file needs to be placed into MODELLER’s
- DSSP. It is expected that the
mkdssp binary is in the
- fpocket (version 2).
- NCBI BLAST+
plus a local copy of the UniProt database for it to search against.
To make this local database,
uniprot_sprot.fasta.gzfiles available from the UniProt website and combine them into a single
- USEARCH version 8.1 or later.
- NumPy and SciPy.
- scikit-learn. Note that precisely version 0.12 is needed - other versions won’t work.
- AllosMod is needed to run part of the protocol.
- nose is also needed to run the test suite (recommended but not essential).
In the Sali lab, running
module load modeller muscle dssp fpocket patch_dock imp blast+ usearch
will get all of these dependencies.
make PYTHON=python3 or
make PYTHON=python2 to build the library
(depending on which version of Python you want to use).
make test to test the library, and
make install to install it.
In most cases you will need to tell
make where to install (if running on
a Linux cluster, CryptoSite will need to be installed on a network-accessible
filesystem) and where your local copy of UniProt is, with something like
make PREFIX=/shared/cryptosite UNIPROT=/database/uniprot install. See
Makefile.include for all make variables that can be configured.