FFPred_README

#------------------------------------------------------------------------------
# README
# Last modified: November 2019
# Author:    psipred@cs.ucl.ac.uk
# Standalone FFPred Code Version 4.0 for function prediction
# Tested on linux x86_64 machines running Centos 7
#------------------------------------------------------------------------------

  Should you encounter any problems please email:
 		psipred@cs.ucl.ac.uk
  including "Standalone FFPred" in the Subject line.

#------------------------------------------------------------------------------

# IMPORTANT INFORMATION

  Standalone FFPred (Code 4.0) has many dependencies and cannot be run without several
  pieces of third-party software. This code is capable of running analyses with
  Human SVMs v3.0 and Fly SVMs v1.0 (see below)
  Detailed instructions on how to set everything up are reported below. Please
  take some time to read them carefully and in sequential order

#------------------------------------------------------------------------------

# LICENSING TERMS

# FFPRED LICENSING TERMS

THE ENCLOSED SOURCE FILES AND ASSOCIATED DATA FILES ARE COPYRIGHT (c) 2008 A.
LOBLEY, J. WARD, D. BUCHAN, D.COZZETTO, F.MINECCI & D. JONES
UNIVERSITY COLLEGE LONDON.

THIS CODE IS LICENSED FOR USE FREE OF CHARGE TO ACADEMIC USERS ONLY. YOU MAY
MODIFY THE CODE BUT ALL COPYRIGHT NOTICES MUST STAY IN PLACE.

POTENTIAL COMMERCIAL USERS MUST CONTACT UCL BUSINESS (HTTP://WWW.UCLB.COM) TO
DISCUSS LICENSING TERMS BEFORE ATTEMPTING TO DOWNLOAD THE SOFTWARE.
THIS SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND. IF YOU CHOOSE
TO DOWNLOAD THE SOFTWARE, YOU PERSONALLY ASSUME ALL RESPONSIBILITY AND RISK FOR
ITS USE AND ANY RESULTS OBTAINED.

#------------------------------------------------------------------------------

# HARDWARE REQUIREMENTS

  Linux/Unix machine.
  At least 25 GB storage (20+ GB for UniRef90 database).
  At least 3+ GB RAM for PSIBLAST runs against UniRef90 database.

#------------------------------------------------------------------------------

# SOFTWARE REQUIREMENTS

  The following URLs can be used as of February 2013 to obtain needed pieces of
  third-party software.

  Note that pre-downloaded copies of these pieces of software are provided
  within the FFPred4 download, and they can easily be all you need - please see
  the detailed instructions that follow!

  EMBOSS                https://emboss.sourceforge.net/download/
  PSIPRED               http://bioinfadmin.cs.ucl.ac.uk/downloads/psipred/
  DISOPRED              http://bioinfadmin.cs.ucl.ac.uk/downloads/DISOPRED/
  MEMSAT-SVM            http://bioinfadmin.cs.ucl.ac.uk/downloads/memsat-svm/
  COILS                 ftp://ftp.ebi.ac.uk/pub/software/unix/coils-2.2/
  NET-N-GLYC 1.0        https://services.healthtech.dtu.dk/services/NetNGlyc-1.0/
  NET-O-GLYC 3.1        https://services.healthtech.dtu.dk/services/NetOGlyc-4.0/
  NETPHOS 3.1/APE 1.0   https://services.healthtech.dtu.dk/services/NetPhos-3.1/
  SIGNALP 4.1           https://services.healthtech.dtu.dk/services/SignalP-6.0/
  WoLFPSORT             http://bioinfadmin.cs.ucl.ac.uk/downloads/ffpred_feature_suite/wolfpsort_0.2.zip
  PFILT                 http://bioinfadmin.cs.ucl.ac.uk/downloads/pfilt/
  BLAST                 ftp://ftp.ncbi.nlm.nih.gov/blast/executables/legacy.NOTSUPPORTED/2.2.26/
  SVM-Light             http://svmlight.joachims.org/
  Uniref90              ftp://ftp.uniprot.org/pub/databases/uniprot/uniref/uniref90/uniref90.fasta.gz

  Perl 5.6 or later, including core modules Getopt::Long and Digest::MD5

#------------------------------------------------------------------------------

# DOWNLOAD

  Please download FFPred 4.0  GITHUB

    git clone https://github.com/psipred/FFPred.git

  A copy of this README file is included in the tar file - top level directory.

#------------------------------------------------------------------------------

# INSTALL - main FFPred files

        cd FFPred4/

  Now, make the relevant Perl scripts executable:

        chmod a+x FFPred.pl
        chmod a+x featurama/features.pl

  Also, make sure the relevant folders can be written to:

        mkdir bin data featurama jobs out uniref90
        chmod -R u+w bin data featurama SVMs uniref90
        chmod a+w jobs
        chmod a+w out

  Due to their size, SVM model files are distributed seperately.

  wget http://bioinfadmin.cs.ucl.ac.uk/downloads/ffpred/SVMs.tar.gz
  tar -zxvf SVMs.tar.gz

  Check that all is fine with the Perl scripts at this point:

        perl -c FFPred.pl
        perl -c featurama/features.pl

  Finally, edit the included a configuration file, called
  "FFPred4/CONFIG", to set the paths (all of them
  are explained in detail in what follows):

  PFILT		directory containing the executable for pfilt software
  BLAST		directory containing the PSIBLAST executable
  MAKEMAT	directory containing the makemat executable
  SEQ2MTX	directory containing the seq2mtx executable
  PSIPRED	directory containing the "bin/" folder for PSIPRED software
  DISOPRED	directory containing the "bin/" folder for DISOPRED software
  MEMSAT	directory containing the executable for MEMSAT-SVM software
  SIGP		directory containing the executable for SignalP software
  NETPHOS	directory containing the executable for NetPhos software
  NETOGLYC	directory containing the executable for netOglyc software
  NETNGLYC	directory containing the executable for netNglyc software
  PSORT		directory containing the executable for WoLFPSORT software
  LOWC		directory containing the executable for Low Complexity analysis (pfilt)
  COILS		directory containing the executable for COILS software
  SEQFEAT	directory containing the executable for SeqFeat
  PEST		directory containing the executable for EMBOSS/pestefind software
  PATH		directory for temporary FFPred4 files
  DB		  full path to FASTA database (UniRef90)
  MASK_DB	full path to masked version of FASTA database (UniRef90)

  Importantly, as explained in the file header, all paths need to be specified
  either starting inside the main FFPred4 directory or as full paths.
  They are currently set on default values; for all software used for feature
  prediction, these defaults correspond to paths to folders located inside
  "FFPred4/featurama/software/". Such folders still need to be created, see
  below: INSTALL - third-party software (-- FEATURE-PREDICTING SOFTWARE).

  IMPORTANT: You should ensure each of these packages is installed and working
  correctly before you attempt to run FFPred. NETPHOS, NETOGLYC, NETNGLYC will
  all need to be hand edited.
  
  IMPORTANT: You must ensure that the correct blast database created using the
  BLAST formatdb tool

#------------------------------------------------------------------------------

# INSTALL - third-party software


  -- GENERAL INFORMATION

  Recent releases of third-party software, or versions that are more suitable
  for the user's needs/operating system/etc, can be downloaded from appropriate
  sources (see above).
  
  We provide legacy versions of the 3rd party software that were used to train 
  the initial version of FFPred at: 
  http://bioinfadmin.cs.ucl.ac.uk/downloads/ffpred/old/FFPred2.tar.gz
  We note that users may need to get their own licences for various pieces of this
  software themselves. Users can decide whether to use the most recent versions 
  or the legacy versions;

  In case the user needs to re-download some of the software due to different
  flavour of operating system etc (for example for BLAST), an indication of the
  preferred, "default" versions that were used for the training of FFPred4 can
  be found in the appropriate sections below (BLAST, SVM-Light,
  FEATURE-PREDICTING SOFTWARE).

  The same is true for the version of the UniRef90 database used for BLAST
  searches, except that in this case the default version is NOT provided within
  the downloaded file due to its size (~20 GB).

  ftp://ftp.uniprot.org/pub/databases/uniprot/previous_releases/
  ftp://ftp.ebi.ac.uk/pub/databases/uniprot/previous_releases/

  Important :
  In all cases, the final path to the directory containing the relevant
  file (or the "bin/" folder in the cases of PSIPRED and DISOPRED programs)
  must be specified in the CONFIG file (see sections below).

  -- Small binaries

  This installs the small scripts that FFPred requires, seq2mtx, features and
  pfilt.

  cd src
  make
  make install

  -- BLAST

  BLAST executables (default version: LEGACY BLAST 2.2.26) should be downloaded 
  from the NCBI. If that is not possible We host a version of legacy BLAST at
  http://bioinfadmin.cs.ucl.ac.uk/downloads/legacy_blast/ncbi.tar.gz
  
  Follow the NCBI instructions for extracting/compiling the material as needed in
  a folder of your choice (e.g. within "FFPred4/material/third_party/"). After
  doing that, note that only few of the files are needed by FFPred4. Firstly,
  copy the following executables into the folder that is specified in the
  CONFIG file next to the 'BLAST' label (the default is simply 'bin', which of
  course indicates the "FFPred4/bin/" folder):

  formatdb
  blastpgp
  makemat

  Also, copy the content of the "data/" folder within the BLAST release
  into a folder called "data/" and sitting in the same directory as that "bin/"
  folder containing the executables (in the default case, that will be
  "FFPred4/data/").


  -- SVM-Light

  SVM-Light version 6.02 must be either downloaded or obtained from
  "FFPred4/material/third_party/svm_light.tar.gz".
  Follow instructions on their website (see above, SOFTWARE REQUIREMENTS) for
  extracting/compiling the material as needed in a folder of your choice (e.g.
  within "FFPred4/material/third_party/"). After doing that, note that only one
  file is needed by FFPred4. Thus, copy the "svm_classify" executable into the
  folder that is specified in the CONFIG file (folder "FFPred4/bin/" in the
  default case).


  -- FEATURE-PREDICTING SOFTWARE

  Twelve pieces of software are used by FFPred4 for feature prediction. With
  the exception of SeqFeat, which is only available as part of the FFPred4
  download, all of them can be obtained elsewhere if needed (see SOFTWARE
  REQUIREMENTS above). Otherwise, default files provided as part of the FFPred4
  release ("FFPred4/material/third_party/") can be used for all of them.
  The default versions are:

  PSIPRED	3.3
  DISOPRED	2.43
  MEMSAT-SVM	1.22
  COILS		2.2
  EMBOSS	6.4.0
  netNglyc	1.0c
  netOglyc	3.1d
  netphos	3.1 (consists of "ape" version 1.0)
  pfilt		1.4 (used for Low Complexity analysis)
  SeqFeat	no version number
  SignalP	4.0e
  WoLFPSORT	0.2

  In all cases, files need to be extracted and compiled/installed locally,
  according to the corresponding (third-party) instructions; this can be
  time consuming and OS-dependent. Note that this must be done in such a way to
  ensure that final files are placed inside the appropriate folder compatibly
  with the content of the CONFIG file ("FFPred4/CONFIG") - default locations
  are folders within "FFPred4/featurama/software/".
  Of course if needed, the CONFIG file can be edited (no change is needed if
  all defaults are used, and if all programs are installed in their default
  locations as in the "Example install" below), so that for each piece of
  software the directory that contains the main executable file (or
  exceptionally, the "bin/" folder in the cases of the PSIPRED and DISOPRED
  programs) is exactly the one reported in the CONFIG file, to the right of the
  corresponding label.
  In detail, for each different program, the path that needs to be specified in
  the CONFIG file is:

  PSIPRED	path to directory containing PSIPRED's "bin/" folder (that contains the psipred executable)
  DISOPRED	path to directory containing DISOPRED's "bin/" folder (that contains the disopred executable)
  MEMSAT	path to directory containing the "run_memsat-svm.pl" executable
  COILS		path to directory containing the "ncoils" executable
  PEST		path to directory containing the "epestfind" executable within the EMBOSS software
  NETNGLYC	path to directory containing the "netNglyc" executable
  NETOGLYC	path to directory containing the "netOglyc" executable
  NETPHOS	path to directory containing the "ape" executable within the ape software
  LOWC		path to directory containing the "pfilt" executable (i.e. "FFPred4/bin": see _NOTE_A_ below)
  SEQFEAT	path to directory containing the "features" executable
  SIGP		path to directory containing the "signalp" executable
  PSORT		path to directory containing the "runWolfPsortSummary" executable within WoLFPSORT

  Example install:

  The following instructions install the EMBOSS program, needed to compute the
  PEST sequence features by means of its "epestfind" executable, respecting all
  default paths within the FFPred4 release. EMBOSS is the largest third-party
  program, with the most time consuming compilation/installation process; the
  same can then be done easily for all other programs.
  As usual, note that a different version can be downloaded if needed for you
  system/platform - these instructions refer to the default case.
  From the main FFPred4 directory run:

        tar -xzvf EMBOSS-6.4.0.tar.gz -C featurama/software/

  This command used the version of EMBOSS that is provided with the FFPred4
  release, but it can easily be adapted using any EMBOSS tar file as the first
  argument.
  The command places files into a "FFPred4/featurama/software/EMBOSS-6.4.0/"
  folder.
  It is then sufficient to follow the third-party instructions found in file
  "FFPred4/featurama/software/EMBOSS-6.4.0/INSTALL". For example, run the
  following commands (importantly, note that, as suggested in that file, the
  "configure" command will need the prefix option to specify the local FULL
  path to the newly created "EMBOSS-6.4.0/" directory; such full path is simply
  indicated here as "/full/path/to/FFPred4/featurama/software/EMBOSS-6.4.0"):

        cd featurama/software/EMBOSS-6.4.0/
        ./configure --prefix="/full/path/to/FFPred4/featurama/software/EMBOSS-6.4.0"
        make
        make check
        make install
        make installcheck

  In the case of EMBOSS, this may require some time, and will also print out
  several messages.
  Note that the defaults in the CONFIG file are respected (see tables above),
  since the executable "epestfind" is found inside the folder with path
  "featurama/software/EMBOSS-6.4.0/bin".

  Similarly, in the default case when the versions provided here are used, all
  other third-party programs can be decompressed using for example the
  following commands, which are of course run again from the main FFPred4
  directory (note the different flags for different compression methods):

        tar -xzvf psipred3.3.tar.gz -C featurama/software/
        tar -xzvf disopred2.43.tar.gz -C featurama/software/
        tar -xzvf memsat-svm1.2.tar.gz -C featurama/software/
        tar -xZvf signalp-4.0e.Linux.tar.Z -C featurama/software/
        tar -xZvf netphos-3.1.Linux.tar.Z -C featurama/software/
        tar -xZvf netOglyc-3.1d.Linux.tar.Z -C featurama/software/
        tar -xZvf netNglyc-1.0c.Linux.tar.Z -C featurama/software/
        tar -xzvf WoLFPSORT_package_v0.2.tar.gz -C featurama/software/
        tar -xzvf ncoils.tar.gz -C featurama/software/
        tar -xzvf seqfeat.tar.gz  -C featurama/software/

  If using different versions, make sure to extract files to those locations
  (see the "-C" flag) when using the default CONFIG file. After that, compiling
  and installing them appropriately in those locations will respect all default
  path values.
  Note that most programs will need the main scripts to be edited and few
  useful paths and variables, usually at the top of file, to be set
  appropriately according to local configurations.
  Also, due to compatibility issues and problems with FFPred legacy code, the
  following need to be observed:
  _NOTE_A_) the "pfilt" program (if using the legacy set obtain the source files from
  "FFPred4/material/third_party/pfilt1.4.tar.gz" or at the webpage found above
  under "SOFTWARE REQUIREMENTS") needs to be used later to prepare the database
  for use by FFPred4, and also it actually constitutes the (only) file needed
  for one of the feature analyses (the Low Complexity analysis): to make sure
  one copy of the executable only is used, FFPred4 is configured so that the
  "pfilt" executable obtained after compiling needs to be placed inside
  "FFPred4/bin/", as per the default in the CONFIG file.
  Thus, just extract/compile the pfilt program as needed in a folder of your
  choice (e.g. within "FFPred4/material/third_party/") and then simply place
  the "pfilt" executable inside the "FFPred4/bin/" folder.
  This also means that there will be no "LowComplexity" folder inside the
  "FFPred4/featurama/software/" directory, as no more files are needed for this
  analysis;
  _NOTE_B_) once PSIPRED has been installed, the "seq2mtx" executable found
  inside PSIPRED's "bin/" folder needs to be copied into the "FFPred4/bin/"
  directory as well;
  _NOTE_C_) for the "ape" program, the README file has been included as
  "FFPred4/material/third_party/netphos-3.1.readme", as it is not included in
  the tar file. Also, note that a new symbolic link may need to be created
  inside folder "ape-1.0/bin/" that points to the appropriate executable, e.g.
  "nnhowplayer6_Linux.i386", and has a name that is appropriate for the local
  OS; for example, if command 'uname -s' gives 'Linux' and 'uname -m' gives
  'x86_64' for your system, the command to be run inside the "ape-1.0/bin/"
  folder is 'ln -s nnhowplayer6_Linux.i386 nnhowplayer6_Linux.x86_64';
  _NOTE_D_) for BOTH the "netNglyc" and "netOglyc" programs, while configuring
  variables at the top of the main script (the one with the same name as the
  program's), the 'setenv SIGNALP ...' command needs to be commented out to
  become '#setenv SIGNALP ...'; this is because the programs are currently not
  compatible with the latest version of SignalP, and anyway the interaction of
  both programs with SignalP is not used by FFPred4;
  _NOTE_E_) for the "coils" program, we suggest to re-compile it (see the
  third-party INSTRUCTIONS file) and call the output executable "ncoils", so
  that it does not clash with existing files - this means that the the
  parameter passed to the "-o ncoils-osf" flag in the command for compiling
  needs to be modified to "-o ncoils"; this will be compatible with the default
  in the CONFIG file. Also, note that both source C files used in that command
  need an extra line on top: "#include <string.h>". Importantly, note that the
  environment variable $COILSDIR mentioned in the INSTRUCTIONS file does not
  need to be set: this is taken care of by the FFPred4 code.

  Lastly, note that at this point the "FFPred4/bin/" directory should have
  exactly the following content:

  blastpgp
  formatdb
  makemat
  pfilt
  seq2mtx
  svm_classify

  -- DATABASE

  The database to be used for BLAST searches by some of the programs needs to
  be obtained and installed.
  These instructions explain how to do so in the sample case when the default
  version of the database is used, and placed in the default location - proceed
  in a similar way for the database of your choice. In particular, any final
  location can be used, as long as it is specified correctly in the CONFIG file
  ("FFPred4/CONFIG").

  Download file uniref2012_06.tar.gz from one of the ftp sites suggested for
  UniRef90 under GENERAL INFORMATION above (after getting to the main ftp site,
  follow "release-2012_06/uniref/") and extract the UniRef90 from its content
  using e.g. 'tar -xzvf' and then 'tar -xvf' and 'gunzip' as appropriate.
  All this can be achieved for example running the following commands from the
  main FFPred4 directory:

        wget -P uniref90/ ftp://ftp.ebi.ac.uk/pub/databases/uniprot/previous_releases/release-2012_06/uniref/uniref2012_06.tar.gz
        tar -xzvf uniref90/uniref2012_06.tar.gz -C uniref90/
        tar -xvf uniref90/uniref90.tar -C uniref90/
        gunzip uniref90/uniref90.xml.gz

  Check that the UniRef90 files are part of the correct release (2012_06,
  13-Jun-2012) by inspecting file "uniref90/uniref90.release_note".
  For old releases, the FASTA version of UniRef90 is not provided - only the
  XML version can be found in the tar file. To convert the XML version to FASTA
  after unzipping it, use the Perl script found in "FFPred4/material/uniref/".
  For example if the unzipped XML file has path "uniref90/uniref90.xml", as in
  this sample case, from the main FFPred4 directory simply run:

        material/uniref/UniRef_xml_to_fasta.pl uniref90/uniref90.xml > uniref90/uniref90.fasta

  Once such final FASTA version of UniRef90 is obtained, all other files
  obtained from the UniProt download should be discarded.

  Note that as explained above, in any case the final UniRef90 file needs to be
  found inside the location specified under "DB" in the CONFIG file (default:
  "FFPred4/uniref90/"), as indeed it is in this sample case.

  Importantly, the following manipulations of the database need to be performed
  from within its own folder - if not using the default location, please adapt
  all following commands accordingly.

  Change working directory to where the database is:

        cd uniref90/

  As usual, the database needs to be formatted for later use with BLAST:

        ../bin/formatdb -i uniref90.fasta

  Also, a masked version needs to be created for use with PSIPRED:

        ../bin/pfilt uniref90.fasta > uniref90.masked.fasta

  Again, the final location of this file needs to be the same as in the CONFIG
  file, under "DB_MASK", and the masked database needs to be formatted too:

        ../bin/formatdb -i uniref90.masked.fasta

#------------------------------------------------------------------------------

# USEFUL TIPS

  The next section details how to run the FFPred4 standalone; however, here is
  some important advice to bear in mind before doing so.

  A) Importantly, BLAST running times can be noticeably speeded up by using
  more than one processor if available. In order to do so, FFPred4 passes the
  appropriate flag (-a) to the blastpgp executable - the default is set to 1
  processor. To change this default to the desired value, simply edit file
  "FFPred4/featurama/lib/PsiBlast.pm" and place the number of available
  processors after the "-a" flag, on line 42.

  B) When running, the FFPred4 main script outputs useful/debugging information
  to standard output or to standard error. Most of that (and in particular, ALL
  of the standard output) is produced by third-party programs; however, some
  messages in the standard error stream are produced specifically by FFPred4.
  It can be useful to redirect those (especially the standard error) and save
  them to file, for later inspection, especially when having problems or when
  looking for detailed information about the prediction proces (including md5
  code associated to each query sequence).

  C) Several sections of commented out text are available at the top of the
  main FFPred4 script ("FFPred4/FFPred.pl") to concisely summarise the most
  important instructions to run the script. Apart from what is explained there
  and what is found in this README file, nothing else should need to be
  modified for a correct running of FFPred4.

  D) Temporary files used by both third-party programs and FFPred4 are deleted
  by default. These too may be useful - if needed, consult the section below
  for instructions on how to retain a copy of all of them.

#------------------------------------------------------------------------------

# RUNNING FFPred4, and EXAMPLE OUTPUT

  As explained in comments to code at the top of the main FFPred4 executable
  ("FFPred4/FFPred.pl"), FFPred4 runs on a single FASTA file containing one or
  more sequences each with a FASTA header, for example placed inside the
  "FFPred4/in/" input folder.

  Temporary files are saved inside the "FFPred4/jobs/" folder. For a given
  sequence, all files are saved in a subfolder named using the first 5
  characters of the sequence's md5 code, or simple variations in case more than
  one sequence have the same md5 code.
  In the default case, such sub-folders are deleted as the job terminates. In
  case it is needed to inspect those files, the "FFPred4/FFPred.pl" executable
  needs to be edited and line 240 needs to be commented out, so that it reads:
    #Clean_temp($job);
  Also, the standard error stream needs to be redirected and saved to file in
  this case, since it contains information useful to associate a sequence's
  name to its md5 code.

  Output files are saved into the "FFPred4/out/" folder. For a given sequence,
  all files are saved in a subfolder named "FFPred_XXX/" where XXX is a short
  identifier taken from the sequence's FASTA header (e.g. if the FASTA header
  consists of a UniProtKB AC number, XXX will be the AC number).

  Example input-output files are provided inside the "FFPred4/test/" folder.
  The "FFPred4/test/in/" folder contains an input FASTA file including 4 test
  sequences. Folders "FFPred4/test/out/jobs/" and "FFPred4/test/out/outputs/"
  contain the corresponding temporary and output folders.
  In order to test the standalone FFPred4 program, copy the input file into the
  input folder and run FFPred4 (note that in the default case, the temporary
  files will be deleted). From the main FFPred4 folder, simply run:

        cp test/in/TEST.fsa in/
        ./FFPred.pl -i in/TEST.fsa

  The running time of FFPred4 on different sequences may vary between several
  minutes to hours, depending on the sequence and hardware specifications.
  In general, FFPred4 can be run in a similar way:

        ./FFPred.pl -i <input_FASTA_file>

  or:

        ./FFPred.pl -i <input_FASTA_file> -o <output_folder>

  Note that if the output folder is omitted (first case) it defaults to the
  "FFPred4/out/" folder.

  Results can be compared to those inside "FFPred4/test/out/".
  For each sequence, a folder containing 5 files is created:
  *.results		values for all features - not of direct interest
  *.features		rescaled final features - not of direct interest
  *.lax_raw		predicted GO terms
  *.lax_formatted	predicted GO terms, better formatted version
  *.lax_all		posterior probabilities for all GO terms
  The last 2 files in the above list are those that can be downloaded using the
  web server version of FFPred4 (however note that values can be different, as
  explained below, since the web server updates its UniRef database very
  often). GO terms with a "low reliability level" are those shown against a red
  background on the web server output (see the relevant publications).
  Note that your outputs and the test outputs should be identical only when
  exactly the same set of third-party programs and default database have been
  installed. Also, small difference can arise due to BLAST running on different
  systems/platforms, even when the same database is used.
  If for example the same versions of feature-predicting programs are used, but
  the most recent version of the UniRef90 database is installed instead of the
  default one, results should still be very similar to the test outputs, with
  all GO terms getting very similar values of posterior probability (see the
  output *.lax_all files).

  -- RUNNING 'ARRAY' JOBS ON A CLUSTER

  Finally, FFPred4 can be run to submit batch jobs to a cluster/HPC facility
  via e.g. SGE or other Grid Engine systems.
  Note that in that case, there also exists the possibility to submit 'array'
  jobs using "qsub -t..." or similar commands. In order to do that, users need
  to use a special syntax when running the FFPred.pl script (note the new flag,
  and the fact that "/full/path/to/FFPred4/in" is now the FULL path to the
  folder containing input FASTA files, rather than a single input file):

        ./FFPred.pl -a -i /full/path/to/FFPred4/in

  Importantly, they also need to set variable $grid_engine_task_ID_variable (on
  line 39) to the name of the environment variable that the system creates to
  store the (progressive) index number of the array job task. For example for
  the Sun Grid Engine, this environment variable is called 'SGE_TASK_ID', which
  is the default value in the FFPred.pl script. Users need to consult the
  manual page for their submission command (e.g. 'man qsub' when using SGE) to
  find out the appropriate environment variable name.

  -- Other Options

  -s the species SVMs to use, current options are fly or human
  -m the svm class, current options strict or full

  -- Drawing Images

  Provide the path including file prefix to the directory containing the
  .featcfg file

  getSchematic.pl out/FFPred_test/test
  getTmIMAGE.pl out/FFPred_test/test

#------------------------------------------------------------------------------