vidjil.cpp

/*
  This file is part of Vidjil-algo <http://www.vidjil.org>
  Copyright (C) 2011-2018 by Bonsai bioinformatics
  at CRIStAL (UMR CNRS 9189, Université Lille) and Inria Lille
  Contributors: 
      Mathieu Giraud <mathieu.giraud@vidjil.org>
      Mikaël Salson <mikael.salson@vidjil.org>
      Marc Duez <marc.duez@vidjil.org>

  "Vidjil-algo" is free software: you can redistribute it and/or modify
  it under the terms of the GNU General Public License as published by
  the Free Software Foundation, either version 3 of the License, or
  (at your option) any later version.

  "Vidjil-algo" is distributed in the hope that it will be useful,
  but WITHOUT ANY WARRANTY; without even the implied warranty of
  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  GNU General Public License for more details.

  You should have received a copy of the GNU General Public License
  along with "Vidjil-algo". If not, see <http://www.gnu.org/licenses/>
*/

//$$ #include

#include<algorithm>
#include<utility>
#include <cstdlib>
#include <fstream>
#include <iostream>
#include <string>
#include <cstring>
#include <time.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <unistd.h>

#include "core/check-compiler.h"
#include "core/tools.h"
#include "core/json.h"
#include "core/germline.h"
#include "core/kmerstore.h"
#include "core/fasta.h"
#include "core/bioreader.hpp"
#include "core/segment.h"
#include "core/windows.h"
#include "core/cluster-junctions.h"
#include "core/dynprog.h"
#include "core/read_score.h"
#include "core/read_chooser.h"
#include "core/compare-all.h"
#include "core/mkdir.h"
#include "core/labels.h"
#include "core/list_utils.h"
#include "core/windowExtractor.h"

#include "lib/CLI11.hpp"
#include "lib/json.hpp"

#include "vidjil.h"

// RELEASE_TAG may be defined in the "release.h" file.
// If RELEASE_TAG is undefined, the version will be the git hash.
// #define RELEASE_TAG  "2013.04"
#include "release.h"

// GIT_VERSION should be defined in "git-version.h", created by "create-git-version-h.sh", to be used outside of releases
#include "git-version.h"

#define PROGNAME "vidjil-algo"
#define VIDJIL_JSON_VERSION "2016b"

//$$ #define (mainly default options)

#define DEFAULT_MULTI_GERMLINE_PATH "germline/"
#define DEFAULT_MULTI_GERMLINE_FILE "homo-sapiens.g"

#define DEFAULT_READ_HEADER_SEPARATOR " "
#define DEFAULT_READS  "./demo/Stanford_S22.fasta"
#define DEFAULT_MIN_READS_CLONE 5
#define DEFAULT_MAX_REPRESENTATIVES 100
#define DEFAULT_MAX_CLONES 100
#define DEFAULT_RATIO_READS_CLONE 0.0
#define NO_LIMIT "all"

#define COMMAND_WINDOWS "windows"
#define COMMAND_CLONES "clones"
#define COMMAND_SEGMENT "segment"
#define COMMAND_GERMLINES "germlines"
 
enum { CMD_WINDOWS, CMD_CLONES, CMD_SEGMENT, CMD_GERMLINES } ;

#define DEFAULT_OUT_DIR "./out/" 

// Fixed filenames/suffixes
#define CLONES_FILENAME ".vdj.fa"
#define CLONE_FILENAME "clone.fa-"
#define WINDOWS_FILENAME ".windows.fa"
#define SEGMENTED_FILENAME ".segmented.vdj.fa"
#define UNSEGMENTED_FILENAME ".unsegmented.vdj.fa"
#define UNSEGMENTED_DETAIL_FILENAME ".fa"
#define AFFECTS_FILENAME ".affects"
#define EDGES_FILENAME ".edges"
#define COMP_FILENAME "comp.vidjil"
#define JSON_SUFFIX ".vidjil"

#define DEFAULT_K      0
#define DEFAULT_W      50
#define DEFAULT_SEED   DEFAULT_GERMLINE_SEED

#define DEFAULT_MAX_AUDITIONED 2000
#define DEFAULT_RATIO_REPRESENTATIVE 0.5

#define DEFAULT_KMER_THRESHOLD NO_LIMIT_VALUE

#define DEFAULT_EPSILON  0
#define DEFAULT_MINPTS   10

#define DEFAULT_CLUSTER_COST  Cluster
#define DEFAULT_SEGMENT_COST   VDJ

#define DEFAULT_TRIM 0

#define MAX_CLONES_FOR_SIMILARITY 20

// warn
#define WARN_MAX_CLONES 100
#define WARN_PERCENT_SEGMENTED 40
#define WARN_COVERAGE 0.6
#define WARN_NUM_CLONES_SIMILAR 10

// display
#define WIDTH_NB_READS 7
#define WIDTH_NB_CLONES 3


using namespace std ;
using json = nlohmann::json;

//$$ options: usage

extern char *optarg;

extern int optind, optopt, opterr;

int usage(char *progname, bool advanced)
{
  cout
       << endl 
       << "Examples (see doc/algo.org)" << endl
       << "  " << progname << " -c clones   -g germline/homo-sapiens.g   -2 -3 -r 1  demo/Demo-X5.fa           # (basic usage, detect the locus for each read," << endl
       << "                                                                                               #  cluster reads and report clones starting from the first read (-r 1)," << endl
       << "                                                                                               #  including unexpected recombinations (-2), assign V(D)J genes and try to detect the CDR3s (-3))" << endl
       << "  " << progname << " -c clones   -g germline/homo-sapiens.g:IGH    -3     demo/Stanford_S22.fasta   # (restrict to complete recombinations on the IGH locus)" << endl
       << "  " << progname << " -c clones   -g germline/homo-sapiens.g   -2 -3 -z 20 demo/LIL-L4.fastq.gz      # (basic usage, output detailed V(D)J analysis on the first 20 clones)" << endl
       << "  " << progname << " -c windows  -g germline/homo-sapiens.g   -y 0 -uu -U demo/LIL-L4.fastq.gz      # (splits all the reads into (large) files depending on the detection of V(D)J recombinations)" << endl
       << "  " << progname << " -c segment  -g germline/homo-sapiens.g   -2 -3 -X 50 demo/Stanford_S22.fasta   # (full analysis of each read, only for debug/testing, here on 50 sampled reads)" << endl
       << "  " << progname << " -c germlines -g germline/homo-sapiens.g              demo/Stanford_S22.fasta   # (statistics on the k-mers)" << endl
    ;
  return 1;
}


int atoi_NO_LIMIT(char *optarg)
{
  return strcmp(NO_LIMIT, optarg) ? atoi(optarg) : NO_LIMIT_VALUE ;
}
double atof_NO_LIMIT(char *optarg)
{
  return strcmp(NO_LIMIT, optarg) ? atof(optarg) : NO_LIMIT_VALUE ;
}

string string_NO_LIMIT(string s)
{
  if (!strcmp(NO_LIMIT, s.c_str()))
    return NO_LIMIT_VALUE_STRING ;

  return s;
}



int main (int argc, char **argv)
{
  cout << "# " << PROGNAME << " -- V(D)J recombinations analysis <http://www.vidjil.org/>" << endl
       << "# Copyright (C) 2011-2018 by the Vidjil team" << endl
       << "# Bonsai bioinformatics at CRIStAL (UMR CNRS 9189, Université Lille) and Inria Lille" << endl 
       << endl
       << "# " << PROGNAME << " is free software, and you are welcome to redistribute it" << endl
       << "# under certain conditions -- see http://git.vidjil.org/blob/master/doc/LICENSE" << endl
       << "# No lymphocyte was harmed in the making of this software," << endl
       << "# however this software is for research use only and comes with no warranty." << endl
       << endl
       << "# Please cite http://biomedcentral.com/1471-2164/15/409 if you use " << PROGNAME << "." << endl
       << endl ;

  //////////////////////////////////
  // Display version information or git log

  string soft_version = PROGNAME ;
  soft_version += " " ;
#ifdef RELEASE_TAG
  cout << "# version: " PROGNAME << " " << RELEASE_TAG << endl ;
  soft_version.append(RELEASE_TAG);
#else
  cout << "# development version" << endl ;
#ifdef GIT_VERSION
  cout << "# git: " << GIT_VERSION << endl ;
  soft_version.append("dev ");
  soft_version.append(GIT_VERSION);
#endif
#endif

  CLI::App app{"Vidjil"};

  //$$ options: defaults

  vector <string> v_reps_V ;
  vector <string> v_reps_D ;
  vector <string> v_reps_J ;

  list <pair <string, string>> multi_germline_paths_and_files ;

  string read_header_separator = DEFAULT_READ_HEADER_SEPARATOR ;
  string f_reads = DEFAULT_READS ;
  string seed = DEFAULT_SEED ;
  bool seed_changed = false;
  string f_basename = "";

  string out_dir = DEFAULT_OUT_DIR;
  
  string comp_filename = COMP_FILENAME;

  int wmer_size = DEFAULT_W ;

  IndexTypes indexType = KMER_INDEX;

  int epsilon = DEFAULT_EPSILON ;
  int minPts = DEFAULT_MINPTS ;
  Cost cluster_cost = DEFAULT_CLUSTER_COST ;
  Cost segment_cost = DEFAULT_SEGMENT_COST ;
  bool detect_CDR3 = false;
  
  bool save_comp = false;
  bool load_comp = false;
  
  int verbose = 0 ;
  int command = CMD_CLONES;

  int max_representatives = DEFAULT_MAX_REPRESENTATIVES ;
  int max_clones = DEFAULT_MAX_CLONES ;
  int min_reads_clone = DEFAULT_MIN_READS_CLONE ;
  float ratio_reads_clone = DEFAULT_RATIO_READS_CLONE;
  // int average_deletion = 4;     // Average number of deletion in V or J

  int max_reads_processed = NO_LIMIT_VALUE;
  int max_reads_processed_sample = NO_LIMIT_VALUE;

  float ratio_representative = DEFAULT_RATIO_REPRESENTATIVE;
  unsigned int max_auditionned = DEFAULT_MAX_AUDITIONED;

  int trim_sequences = DEFAULT_TRIM;

  bool trim_sequences_changed = false;
  
  bool output_sequences_by_cluster = false;
  bool output_segmented = false;
  bool output_unsegmented = false;
  bool output_unsegmented_detail = false;
  bool output_unsegmented_detail_full = false;
  bool output_affects = false;
  bool keep_unsegmented_as_clone = false;

  bool several_D = false;

  vector <string> multi_germlines ;
  bool multi_germline = false;
  bool multi_germline_mark = false;
  bool multi_germline_one_unique_index = false;
  bool multi_germline_unexpected_recombinations_12 = false;
  bool multi_germline_unexpected_recombinations_1U = false;

  string forced_edges = "" ;

  map <string, string> windows_labels ;
  string windows_labels_file = "" ;
  bool only_labeled_windows = false ;

  char c ;

  int options_s_k = 0 ;

  double expected_value = THRESHOLD_NB_EXPECTED;
  double expected_value_D = THRESHOLD_NB_EXPECTED_D;

  //json which contains the Levenshtein distances
  json jsonLevenshtein;
  bool jsonLevenshteinComputed = false ;

  int kmer_threshold = DEFAULT_KMER_THRESHOLD;

  //$$ options: definition wiht CLI11
  string group = "";

  // cerr << "Usage: " << progname << " [options] <reads.fa/.fq/.gz>" << endl << endl;

  group = "Files";
  app.add_option("reads", f_reads, "reads.fa/.fq/.gz") -> group(group) -> set_type_name("FILE");


  group = "Command selection";
  string cmd = COMMAND_CLONES;
  app.add_option("-c", cmd, "command"
                 "\n  \t\t" COMMAND_CLONES    "  \t locus detection, window extraction, clone clustering (default command, most efficient, all outputs)"
                 "\n  \t\t" COMMAND_WINDOWS   "  \t locus detection, window extraction"
                 "\n  \t\t" COMMAND_SEGMENT   "  \t detailed V(D)J designation (not recommended)"
                 "\n  \t\t" COMMAND_GERMLINES "  \t statistics on k-mers in different germlines")
    -> group(group) -> set_type_name("COMMAND");

  group = "Input" ;
  app.add_option("--separator", read_header_separator, "separator for headers in the reads file", true) -> group(group) -> level() -> set_type_name("CHAR")  ;

  
  group = "Germline presets (at least one -g or -V/(-D)/-J option must be given for all commands except -c " COMMAND_GERMLINES ")";
  app.add_option("-g", multi_germlines, R"Z(
         -g <.g file>(:filter)
                    multiple locus/germlines, with tuned parameters.
                    Common values are '-g germline/homo-sapiens.g' or '-g germline/mus-musculus.g'
                    The list of locus/recombinations can be restricted, such as in '-g germline/homo-sapiens.g:IGH,IGK,IGL'
         -g <path>
                    multiple locus/germlines, shortcut for '-g <path>/)Z" DEFAULT_MULTI_GERMLINE_FILE R"Z(',
                    processes human TRA, TRB, TRG, TRD, IGH, IGK and IGL locus, possibly with some incomplete/unusal recombination
)Z") -> group(group) -> expected(1) -> set_type_name("GERMLINES");

  app.add_option("-V", v_reps_V, "custom V germline multi-fasta file") -> group(group) -> set_type_name("FILE");
  app.add_option("-D", v_reps_D, "custom D germline multi-fasta file (and resets -m and -w options), will segment into V(D)J components") -> group(group) -> set_type_name("FILE");
  app.add_option("-J", v_reps_J, "custom V germline multi-fasta file") -> group(group) -> set_type_name("FILE");

  group = "Locus/recombinations";
  app.add_flag("-d", several_D, "try to detect several D (experimental)") -> group(group);
  app.add_flag("-2",  multi_germline_unexpected_recombinations_12, "try to detect unexpected recombinations (must be used with -g)") -> group(group);


  group = "Experimental options (do not use)";
  app.add_flag("-I", multi_germline_mark, "ignore k-mers common to different germline systems (experimental, must be used with -g, do not use)") -> group(group) -> level();
  app.add_flag("-1", multi_germline_one_unique_index,
               "use a unique index for all germline systems (experimental, must be used with -g, do not use)") -> group(group) -> level();
  app.add_flag("-4", multi_germline_unexpected_recombinations_1U, "try to detect unexpected recombinations with translocations (experimental, must be used with -g, do not use)") -> group(group) -> level();
  app.add_flag("--keep", keep_unsegmented_as_clone, "keep unsegmented reads as clones, taking for junction the complete sequence, to be used on very small datasets (for example --keep -AX 20)") -> group(group) -> level();

  group = "Window prediction";
#ifndef NO_SPACED_SEEDS
  /*
       << "  (use either -s or -k option, but not both)" << endl
           << "  (all these options, except -w, are overriden when using -g)" << endl
  */
#endif

  int kmer_size = 0; // TODO: vierer
  app.add_option("-k", kmer_size, "k-mer size used for the V/J affectation (default: 10, 12, 13, depends on germline)") -> group(group);
  // seed = seed_contiguous(kmer_size);
  //        seed_changed = true;
  //	options_s_k++ ;
    
#ifndef NO_SPACED_SEEDS
  //      << "                (using -k option is equivalent to set with -s a contiguous seed with only '#' characters)" << endl
#endif
  app.add_option("-w", wmer_size,
                 "w-mer size used for the length of the extracted window ('" NO_LIMIT "': use all the read, no window clustering)") -> group(group) -> transform(string_NO_LIMIT);
  app.add_option("-e", expected_value,
                 "maximal e-value for determining if a V-J segmentation can be trusted", true) -> group(group) -> transform(string_NO_LIMIT);
  app.add_option("-t", trim_sequences, // trim_sequences_changed = true
                 "trim V and J genes (resp. 5' and 3' regions) to keep at most <int> nt  (0: no trim)") -> group(group);

  app.add_option("-s",
                 [&](CLI::results_t res) {
                   seed = res[0] ;
                   options_s_k++ ;
                   seed_changed = true;
                   return true;
                 },
                 "spaced seeds used for the V/J affectation (default: depends on germline)"
                 ) -> group(group) -> level();

/*
#ifdef NO_SPACED_SEEDS
        cerr << "To enable the option -s, please compile without NO_SPACED_SEEDS" << endl;
#endif
*/
  



  group = "Labeled sequences (windows related to these sequences will be kept even if -r/-% thresholds are not reached)";
  /*
  app.add_option("-W",
                 // TODO:  windows_labels[string(optarg)] = string("-W");
                 , "label the given sequence") -> group(group);
  */
  app.add_option("-l", windows_labels_file, "label a set of sequences given in <file>") -> group(group) -> set_type_name("FILE");
  app.add_flag("-F", only_labeled_windows, "filter -- keep only the windows related to the labeled sequences") -> group(group);


  //  if (advanced)
  group = "Fine segmentation options (second pass)";


  //       << "  -f <string>   use custom Cost for fine segmenter : format \"match, subst, indels, del_end, homo\" (default "<< DEFAULT_SEGMENT_COST <<" )"<< endl
  //      case 'f''	segment_cost=strToCost(optarg, VDJ);      break;
  app.add_option("-E", expected_value_D, "maximal e-value for determining if a D segment can be trusted", true) -> group(group) -> level();

  app.add_option("-Z", kmer_threshold, "typical number of V genes, selected by k-mer comparison, to compare to the read ('" NO_LIMIT "': all genes, default)", true) -> group(group)
    -> transform(string_NO_LIMIT) -> level() -> set_type_name("NB") ;
  
  group = "Clone analysis (second pass)";
  app.add_flag("-3,--cdr3", detect_CDR3, "CDR3/JUNCTION detection (requires gapped V/J germlines)") -> group(group);

  
  // if (advanced)
  group = "Additional clustering (experimental)" ;

  app.add_flag_function("-q", [&](size_t n) { indexType = AC_AUTOMATON; });
  app.add_option("-n", epsilon, "minimum required neighbors for automatic clustering. No automatic clusterisation if =0.", true) -> group(group) -> level();

  app.add_option("-N", minPts, "", true) -> group(group) -> level();
  app.add_option("-S", save_comp, "generate and save comparative matrix for clustering") -> group(group) -> level();
  app.add_option("-L", load_comp, "load comparative matrix for clustering") -> group(group) -> level();
  app.add_option("--forced-edges", forced_edges, "manual clustering -- a file used to force some specific edges") -> group(group) -> level();
  // << "  -C <string>   use custom Cost for automatic clustering : format \"match, subst, indels, del_end, homo\" (default "<< DEFAULT_CLUSTER_COST <<" )"<< endl
  // cluster_cost=strToCost(optarg, Cluster);

  group = "Limits to report a clone (or a window)";
  app.add_option("-r", min_reads_clone, "minimal number of reads supporting a clone", true) -> group(group);
  app.add_option("--ratio", ratio_reads_clone, "minimal percentage of reads supporting a clone", true) -> group(group);

  group = "Limits to further analyze some clones";
  app.add_option("-y", max_representatives,
                 "maximal number of clones computed with a consensus sequence ('" NO_LIMIT "': no limit)", true) -> group(group) -> transform(string_NO_LIMIT);
  app.add_option("-z", max_clones, //
                 // [](CLI::results_t res) {
                 //    max_clones = atoi_NO_LIMIT(optarg);
                 //    if ((max_representatives < max_clones) && (max_representatives != NO_LIMIT_VALUE))
                 //    max_representatives = max_clones ;
                 // }
                 "maximal number of clones to be analyzed with a full V(D)J designation ('" NO_LIMIT "': no limit, do not use)", true)-> group(group);

  app.add_flag_function("-A", [&](size_t n) {
      ratio_reads_clone = 0 ;
      min_reads_clone = 1 ;
      max_representatives = NO_LIMIT_VALUE ;
      max_clones = NO_LIMIT_VALUE ;
    },
    "reports and segments all clones (-r 0 -% 0 -y " NO_LIMIT " -z " NO_LIMIT "), to be used only on very small datasets (for example -AX 20)") -> group(group);

  app.add_option("-x", max_reads_processed_sample,
                 "maximal number of reads to process ('" NO_LIMIT "': no limit, default), only first reads") -> group(group) -> transform(string_NO_LIMIT);
  app.add_option("-X", max_reads_processed,
                 "maximal number of reads to process ('" NO_LIMIT "': no limit, default), sampled reads") -> group(group) -> transform(string_NO_LIMIT);

  
  group = "Detailed output per read (generally not recommended, large files, but may be used for filtering, as in -uu -X 1000)";
  app.add_flag("-U", output_segmented, "output segmented reads (in " SEGMENTED_FILENAME " file)") -> group(group);
  app.add_flag_function("-u", [&](size_t n) {
      output_unsegmented = output_unsegmented_detail_full ;       // -uuu
      output_unsegmented_detail_full = output_unsegmented_detail; // -uu
      output_unsegmented_detail = true;                           // -u
    }, R"Z(
        -u          output unsegmented reads, gathered by unsegmentation cause, except for very short and 'too few V/J' reads (in *)Z" UNSEGMENTED_DETAIL_FILENAME R"Z( files)
        -uu         output unsegmented reads, gathered by unsegmentation cause, all reads (in *)Z" UNSEGMENTED_DETAIL_FILENAME R"Z( files) (use only for debug)
        -uuu        output unsegmented reads, all reads, including a )Z" UNSEGMENTED_FILENAME R"Z( file (use only for debug)
)Z") -> group(group);

  app.add_flag("-K", output_affects, "output detailed k-mer affectation on all reads (in " AFFECTS_FILENAME " file) (use only for debug, for example -KX 100)") -> group(group);


  group = "Output";
  app.add_option("-o", out_dir, "output directory", true) -> group(group) -> set_type_name("PATH");
  app.add_option("-b", f_basename, "output basename (by default basename of the input file)") -> group(group) -> set_type_name("FILE");
  app.add_flag("-a", output_sequences_by_cluster, "output all sequences by cluster (" CLONE_FILENAME "*), to be used only on small datasets") -> group(group);
  app.add_flag_function("-v", [&](size_t n) { verbose += n ; }, "verbose mode") -> group(group);


  app.add_flag_function("-H", [&](size_t n) { throw CLI::CallForAdvancedHelp() ; },
                        "help, including advanced and experimental options"
                        "\n                              "
                        "The full help is available in the doc/algo.org file."
                        );



  //$$ options: parsing
  CLI11_PARSE(app, argc, argv);

  //$$ options: post-processing+display

  if (cmd == COMMAND_CLONES)
    command = CMD_CLONES;
  else if (cmd == COMMAND_SEGMENT)
    command = CMD_SEGMENT;
  else if (cmd == COMMAND_WINDOWS)
    command = CMD_WINDOWS;
  else if (cmd == COMMAND_GERMLINES)
    command = CMD_GERMLINES;
  else {
    cerr << "Unknwown command " << optarg << endl;
    usage(argv[0], false);
  }

  list <string> f_reps_V(v_reps_V.begin(), v_reps_V.end());
  list <string> f_reps_D(v_reps_D.begin(), v_reps_D.end());
  list <string> f_reps_J(v_reps_J.begin(), v_reps_J.end());


  for (string arg: multi_germlines)
    {
      multi_germline = true;

      struct stat buffer;
      if (stat(arg.c_str(), &buffer) == 0)
        {
          if( buffer.st_mode & S_IFDIR )
            {
              // argument is a directory
              multi_germline_paths_and_files.push_back(make_pair(arg, DEFAULT_MULTI_GERMLINE_FILE)) ;
              break ;
                }
        }

      // argument is not a directory (and basename can include ':' with a filter)
      multi_germline_paths_and_files.push_back(make_pair(extract_dirname(arg), extract_basename(arg, false)));
    }


  if (!multi_germline && (!f_reps_V.size() || !f_reps_J.size()))
    {
      cerr << ERROR_STRING << "At least one germline must be given with -g or -V/(-D)/-J." << endl ;
      return 1;
    }

  if (options_s_k > 1)
    {
      cerr << ERROR_STRING << "Use at most one -s or -k option." << endl ;
      return 1;
    }

  string out_seqdir = out_dir + "/seq/" ;

  if (verbose)
    cout << "# verbose " << verbose << endl ;

  if (f_reads == DEFAULT_READS)
    {
      cout << "# using default sequence file: " << f_reads << endl ;
    }

  //  else
  //  {
  //    cerr << ERROR_STRING << "Wrong number of arguments." << endl ;
  //    return 1;
  //  }

  size_t min_cover_representative = (size_t) (min_reads_clone < (int) max_auditionned ? min_reads_clone : max_auditionned) ;

  // Default seeds

#ifdef NO_SPACED_SEEDS
  if (! seed_changed)
    {
      cerr << ERROR_STRING << PROGNAME << " was compiled with NO_SPACED_SEEDS: please provide a -k option." << endl;
      return 1;
  }
#endif
	  

#ifndef NO_SPACED_SEEDS
  // Check seed buffer  
  if (seed.size() >= MAX_SEED_SIZE)
    {
      cerr << ERROR_STRING << "Seed size is too large (MAX_SEED_SIZE)." << endl ;
      return 1;
    }
#endif


  if ((wmer_size< 0) && (wmer_size!= NO_LIMIT_VALUE))
    {
      cerr << ERROR_STRING << "Too small -w. The window size should be positive" << endl;
      return 1;
    }

  // Check that out_dir is an existing directory or creates it
  const char *out_cstr = out_dir.c_str();

  if (mkpath(out_cstr, 0755) == -1) {
    cerr << ERROR_STRING << "Directory creation: " << out_dir << endl; perror("");
    return 2;
  }

  const char *outseq_cstr = out_seqdir.c_str();
  if (mkpath(outseq_cstr, 0755) == -1) {
    cerr << ERROR_STRING << "Directory creation: " << out_seqdir << endl; perror("");
    return 2;
  }

  // Compute basename if not given as an option
  if (f_basename == "") {
    f_basename = extract_basename(f_reads);
  }

  out_dir += "/" ;

  /// Load labels ;
  load_into_map(windows_labels, windows_labels_file, "-l");

  switch(command) {
  case CMD_WINDOWS: cout << "Extracting windows" << endl; 
    break;
  case CMD_CLONES: cout << "Analysing clones" << endl; 
    break;
  case CMD_SEGMENT: cout << "Segmenting V(D)J" << endl;
    break;
  case CMD_GERMLINES: cout << "Discovering germlines" << endl;
    break;
  }

  cout << "Command line: ";
  for (int i=0; i < argc; i++) {
    cout << argv[i] << " ";
  }
  cout << endl;

  //////////////////////////////////
  // Display time and date
  time_t rawtime;
  struct tm *timeinfo;
  char time_buffer[80];

  time (&rawtime );
  timeinfo = localtime (&rawtime);

  strftime (time_buffer, 80,"%F %T", timeinfo);

  cout << "# " << time_buffer << endl ;


  //////////////////////////////////
  // Warning for non-optimal use

  if (max_clones == NO_LIMIT_VALUE || max_clones > WARN_MAX_CLONES)
    {
      cout << endl
	   << "* WARNING: " << PROGNAME << " was run with '-A' option or with a large '-z' option" << endl ;
    }
  
  if (command == CMD_SEGMENT)
    {
      cout << endl
	   << "* WARNING: " << PROGNAME << " was run with '-c segment' option" << endl ;
    }
  
  if (max_clones == NO_LIMIT_VALUE || max_clones > WARN_MAX_CLONES || command == CMD_SEGMENT)
    {
      cout << "* " << PROGNAME << " efficientl extracts windows overlapping the CDR3" << endl
           << "* to cluster reads into clones ('-c clones')." << endl
           << "* Computing accurate V(D)J designations for many sequences ('-c segment' or large '-z' values)" << endl
           << "* is slow and should be done only on small datasets or for testing purposes." << endl
	   << "* More information is provided in the 'doc/algo.org' file." << endl 
	   << endl ;
    }


  /////////////////////////////////////////
  //            JSON OUTPUT              //
  /////////////////////////////////////////

  string f_json = out_dir + f_basename + JSON_SUFFIX ;

  ostringstream stream_cmdline;
  for (int i=0; i < argc; i++) stream_cmdline << argv[i] << " ";

  json j = {
    {"vidjil_json_version", VIDJIL_JSON_VERSION},
    {"samples", {
        {"number", 1},
        {"original_names", {f_reads}},
        {"run_timestamp", {time_buffer}},
        {"producer", {soft_version}},
        {"commandline", {stream_cmdline.str()}}
      }}
  };


  /////////////////////////////////////////
  //            LOAD GERMLINES           //
  /////////////////////////////////////////

  if (command == CMD_GERMLINES)
    {
      multi_germline = true ;
      multi_germline_one_unique_index = true ;
    }

  MultiGermline *multigermline = new MultiGermline(indexType, !multi_germline_one_unique_index);

    {
      cout << "Load germlines and build Kmer indexes" << endl ;
    
      if (multi_germline)
	{
          for (pair <string, string> path_file: multi_germline_paths_and_files)
            {
              try {
                multigermline->build_from_json(path_file.first, path_file.second, GERMLINES_REGULAR,
                                               FIRST_IF_UNCHANGED("", seed, seed_changed),
                                               FIRST_IF_UNCHANGED(0, trim_sequences, trim_sequences_changed), (kmer_threshold != NO_LIMIT_VALUE));
              } catch (std::exception& e) {
                cerr << ERROR_STRING << PROGNAME << " cannot properly read " << path_file.first << "/" << path_file.second << ": " << e.what() << endl;
                delete multigermline;
                return 1;
              }
            }
	}
      else
	{
	  // Custom germline
	  Germline *germline;
	  germline = new Germline("custom", 'X',
                                  f_reps_V, f_reps_D, f_reps_J,
                                  seed, trim_sequences, (kmer_threshold != NO_LIMIT_VALUE));

          germline->new_index(indexType);

	  multigermline->insert(germline);
	}
    }

    cout << endl ;

    if (multi_germline_one_unique_index) {
      multigermline->build_with_one_index(seed, true);
    }

      if (multi_germline_unexpected_recombinations_12 || multi_germline_unexpected_recombinations_1U) {
        if (!multigermline->index) {
          multigermline->build_with_one_index(seed, false);
        }
      }

      if (multi_germline_unexpected_recombinations_12) {
        Germline *pseudo = new Germline(PSEUDO_UNEXPECTED, PSEUDO_UNEXPECTED_CODE, "", trim_sequences, (kmer_threshold != NO_LIMIT_VALUE));
        pseudo->seg_method = SEG_METHOD_MAX12 ;
        pseudo->set_index(multigermline->index);
        multigermline->germlines.push_back(pseudo);
      }

      if (multi_germline_unexpected_recombinations_1U) {
        Germline *pseudo_u = new Germline(PSEUDO_UNEXPECTED, PSEUDO_UNEXPECTED_CODE, "", trim_sequences, (kmer_threshold != NO_LIMIT_VALUE));
        pseudo_u->seg_method = SEG_METHOD_MAX1U ;
        // TODO: there should be more up/downstream regions for the PSEUDO_UNEXPECTED germline. And/or smaller seeds ?
        pseudo_u->set_index(multigermline->index);
        multigermline->germlines.push_back(pseudo_u);
    }

      // Should come after the initialization of regular (and possibly pseudo) germlines
    {
      for (pair <string, string> path_file: multi_germline_paths_and_files)
        multigermline->build_from_json(path_file.first, path_file.second, GERMLINES_INCOMPLETE,
                                       FIRST_IF_UNCHANGED("", seed, seed_changed),
                                       FIRST_IF_UNCHANGED(0, trim_sequences, trim_sequences_changed), (kmer_threshold != NO_LIMIT_VALUE));
      if ((! multigermline->one_index_per_germline) && (command != CMD_GERMLINES)) {
        multigermline->insert_in_one_index(multigermline->index, true);
      }
    }

    if (multi_germline_mark)
      multigermline->mark_cross_germlines_as_ambiguous();

    multigermline->finish();
    cout << "Germlines loaded: " ;
    cout << *multigermline ;
    cout << endl ;

    // Number of reads for e-value computation
    unsigned long long nb_reads_for_evalue = (expected_value > NO_LIMIT_VALUE) ? nb_sequences_in_file(f_reads, true) : 1 ;

    
  //////////////////////////////////
  //$$ Read sequence files

    int only_nth_read = 1 ;
    if (max_reads_processed_sample != NO_LIMIT_VALUE)
      {
        only_nth_read = nb_sequences_in_file(f_reads) / max_reads_processed_sample;
        if (only_nth_read == 0)
          only_nth_read = 1 ;

        max_reads_processed = max_reads_processed_sample ;

        if (only_nth_read > 1)
          cout << "Processing every " << only_nth_read
               << (only_nth_read == 2 ? "nd" : (only_nth_read == 3 ? "rd" : "th"))
               << " read" << endl ;
      }

  OnlineBioReader *reads;

  try {
    reads = OnlineBioReaderFactory::create(f_reads, 1, read_header_separator, max_reads_processed, only_nth_read);
  } catch (const invalid_argument e) {
    cerr << ERROR_STRING << PROGNAME << " cannot open reads file " << f_reads << ": " << e.what() << endl;
    return 1;
  }

  out_dir += "/";


  //////////////////////////////://////////
  //         DISCOVER GERMLINES          //
  /////////////////////////////////////////
  if (command == CMD_GERMLINES)
    {
      map <char, int> stats_kmer, stats_max;
      IKmerStore<KmerAffect> *index = multigermline->index ;

      // Initialize statistics, with two additional categories
      index->labels.push_back(make_pair(KmerAffect::getAmbiguous(), BIOREADER_AMBIGUOUS));
      index->labels.push_back(make_pair(KmerAffect::getUnknown(), BIOREADER_UNKNOWN));
      
      for (list< pair <KmerAffect, BioReader> >::const_iterator it = index->labels.begin(); it != index->labels.end(); ++it)
	{
	  char key = affect_char(it->first.affect) ;
	  stats_kmer[key] = 0 ;
	  stats_max[key] = 0 ;
	}
      
      // init forbidden for .max()
      set<KmerAffect> forbidden;
      forbidden.insert(KmerAffect::getAmbiguous());
      forbidden.insert(KmerAffect::getUnknown());
      
      // Loop through all reads

      int nb_reads = 0 ;
      int total_length = 0 ;
      int s = index->getS();

      int kmer_size = seed_weight(seed);

      while (reads->hasNext())
	{
	  reads->next();
	  nb_reads++;
	  string seq = reads->getSequence().sequence;
	  total_length += seq.length() - s + 1;

	  KmerAffectAnalyser *kaa = new KmerAffectAnalyser(*index, seq);

	  for (int i = 0; i < kaa->count(); i++) 
	    { 
	      KmerAffect ksa = kaa->getAffectation(i);
	      stats_kmer[affect_char(ksa.affect)]++ ;
	    }

          delete kaa;

	  CountKmerAffectAnalyser ckaa(*index, seq);
	  ckaa.setAllowedOverlap(kmer_size-1);

	  stats_max[affect_char(ckaa.max(forbidden).affect)]++ ;

	}