Usage:
  ======
  # SecNet, Protein secondary structure prediction software
  # Version 1.1.4

  usage: secnet.py3 [-h] -i INPUT -o OUTPUT -l {3,4,5,8,all} [-r {1,2,both}]
                    [-q] [-c CPU]

  required named arguments:
    -i INPUT, --input INPUT
                          for example, -i abcd.fasta or --input
                          /home/user/abcd.seq or -i "./input_dir/*.fasta" or -i
                          "/data/*.seq" or -i /data/dir_with_sequences (for a
                          directory, processes all *.seq and *.fasta). Do not
                          forget to include "..." quotes for * or ? wildcard
                          matching.
    -o OUTPUT, --output OUTPUT
                          for example, -o /home/user/output or --output
                          ./project/
    -l {3,4,5,8,all}, --label {3,4,5,8,all}
                          for example, -l 8 or --label 3 --rule 1 or -l all

  optional arguments:
    -h, --help            show this help message and exit
    -r {1,2,both}, --rule {1,2,both}
                          for example, -r 1 or --rule 2 or -r both
    -q, --quiet           limits standard output to reporting processed
                          sequences with -q or --quiet
    -c CPU, --cpu CPU     by default secnet uses all available CPU cores, for
                          example override it with -c 2 or --cpu 4 or -c 8

  Labels and rules for 3 labels:
  ==============================
  (1) 8 labels: H, E, C, T, G, S, B, I
      8 original DSSP labels are unchanged (C, S, B, T, I, G, H, E)
      Argument: --label 8 or -l 3

  (2) 5 labels: H, E, C, T, G
      (C, S, B) -> C, (H, I) -> H, (E) -> E, (T) -> T, (G) -> G
      Argument: --label 5 or -l 5

  (3) 4 labels: H, E, C, T
      (C, S, B, G) -> C, (H, I) -> H, (E) -> E, (T) -> T
      Argument: --label 4 or -l 4

  (4) 3 harder to predict labels of Rule #1: H, E, C
      (C, S, T) -> C, (H, I, G) -> H, (E, B) -> E
      Argument: --label 3 or -l 3 or --label 3 --rule 1 or -l 3 -r 1
      Note: The rule option is only applicable to 3 labels. If no option is specified, Rule #1 is assumed by default.

  (5) 3 easier to predict labels of Rule #2: H, E, C
      (C, S, B, T, I, G) -> C, (H) -> H, (E) -> E
      Argument: --label 3 --rule 2 or -l 3 -r 2

  (6) 3 labels with both Rule #1 and Rule #2
      Argument: --label 3 --rule both or -l 3 -r both

  (7) all sets of the above will be predicted
      Argument: --label all or -l all

  Input:
  ======
  (1) Single file with FASTA-formatted sequence
      --input /home/user/secnet/input/example.fasta or -i ~/project/favorite.seq or --input ABCD4.sequence

  (2) Directory with all sequence files matching *.seq and *.fasta extensions
      --input /data/myproject/dir_with_seqs -i ./all_seq/

  (3) Wildcard matching, please use ' or " quotes
      otherwise a shell may expand all matching files into a set of program arguments and the program will fail
      --input '/pdb/2A*.fasta' or -i "~/project_jan/*.seq" or -i "ABCD?.seq"

  Output directory:
  =================
  Program will predict requested sets of labels with .ss8 or .ss5, .ss4, .ss3rule1 or .ss3rule2 extensions.
  The base name of generated files will be taken from corresponding input files by removing their extensions,
  for example example.fasta will lead to example.ss8 or ABCD4.seq will produce ABCD4.ss3rule1
      --output ./output or -o /mydata/output/

  CPU:
  ====
  By default, SecNet will detect a number of available CPU cores and will use all of them to run 3rd-party software
  to generate input features and to process neural networks to make predictions.
  A user can limit a number of used CPU cores to make the system more responsive during the prediction for
  other user tasks. It will take longer time to make the predictions.
      --cpu 4 or -c 2 or --cpu 7

  Quiet:
  ======
  A user may reduce amount of standard output with the following optional flag. If no such flag is provided,
  the output includes all information about running 3rd-party software for input feature generation and
  additional information from neural networks such as 10 predictions from 10 neural networks before taking
  the major vote. If you wish suppress any output during the execution, you may add "> /dev/null 2>&1"
  at the end of your command.
      --quiet

  Help:
  =====
  Prints required and optional arguments as well as this help.
      --help or -h

  Kit directory and file structure:
  =================================

  3rd_databases      -- extracted downloaded sequence databases for psiblast and hmm feature generation
  3rd_software       -- third-party software (psiblast and hmm)
                        with binaries for Unix and MacOS (supports Catalina with 64-bit support only)
  anaconda3          -- locally installed Anaconda Python 3 with installed required Python libraries
  bin                -- scripts with secnet subroutines
  download           -- directory for temporary files during installation
  EXAMPLE            -- directory with an example sequence and expected output with properly working software
  features           -- saves generated input feature files, speeds up nnet computation for subsequent calls;
                        the content of the following subdirectories may be deleted later if needed
  features/psiblast  -- saved input psiblast features with PSSM profiles from the 1st and 2nd rounds
                        (example.mtx.1 and exammple.mtx.2)
  features/hhm       -- saved input hhm features with HHM parameters (example.hhm1)
  features/temp      -- includes undeleted temporary files to track down errors in failed generation of input features
  input              -- includes example.fasta and 4 other sample sequences in FASTA format
  models             -- trained neural networks for 5 sets of labels each with 10 cross-validation trainings
  nnpython3          -- symbolic link to locally installed Anaconda Python 3
  output             -- empty directory for output convenience
  QUICK_START        -- quick start instructions with examples
  README             -- a readme file with this content
  secnet.py3         -- SecNet executable which should be executed as ./secnet.py3 or /home/user/secnet_kit/secnet.py3
  Test2018           -- files related to Test2018 data set from our paper
  Test2018/input     -- 149 FASTA-formatted sequence files from Test2018
  Test2018/expected  -- expected secondary-structure predictions for all 5 label alphabets of 149 Test2018 sequences
  Test2018/test.bash -- a script that tests your SecNet installation on Test2018 and saves results to Test2018/output.
                        You may compare your re-generated results with the expected ones stored in Test2018/expected.
                        It is a normal situation when due to different hardware and software your predictions and
                        expected predictions slightly vary by one or few labels. Your overall accurarcy will be close
                        within 0.01-0.03% to the one reported in our paper.
  Test2018/output    -- an empty directory for output from Test2018/test.bash