vidjil.cpp 60.2 KB
Newer Older
Mikaël Salson's avatar
Mikaël Salson committed
1
/*
2 3
  This file is part of Vidjil-algo <http://www.vidjil.org>
  Copyright (C) 2011-2018 by Bonsai bioinformatics
4 5 6 7 8
  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>
Mikaël Salson's avatar
Mikaël Salson committed
9

10
  "Vidjil-algo" is free software: you can redistribute it and/or modify
Mikaël Salson's avatar
Mikaël Salson committed
11 12 13 14
  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.

15
  "Vidjil-algo" is distributed in the hope that it will be useful,
Mikaël Salson's avatar
Mikaël Salson committed
16 17 18 19 20
  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
21
  along with "Vidjil-algo". If not, see <http://www.gnu.org/licenses/>
Mikaël Salson's avatar
Mikaël Salson committed
22 23
*/

Mathieu Giraud's avatar
Mathieu Giraud committed
24
//$$ #include
Mikaël Salson's avatar
Mikaël Salson committed
25 26 27 28 29 30 31 32 33 34 35 36 37

#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>

38
#include "core/check-compiler.h"
Mikaël Salson's avatar
Mikaël Salson committed
39
#include "core/tools.h"
WebTogz's avatar
WebTogz committed
40
#include "core/json.h"
41
#include "core/germline.h"
Mikaël Salson's avatar
Mikaël Salson committed
42 43
#include "core/kmerstore.h"
#include "core/fasta.h"
44
#include "core/bioreader.hpp"
Mikaël Salson's avatar
Mikaël Salson committed
45
#include "core/segment.h"
Mathieu Giraud's avatar
Mathieu Giraud committed
46
#include "core/windows.h"
Mikaël Salson's avatar
Mikaël Salson committed
47 48 49 50 51 52 53
#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"
Mikaël Salson's avatar
Mikaël Salson committed
54
#include "core/list_utils.h"
Mathieu Giraud's avatar
Mathieu Giraud committed
55
#include "core/windowExtractor.h"
56
#include "core/output.h"
Mikaël Salson's avatar
Mikaël Salson committed
57

Mathieu Giraud's avatar
Mathieu Giraud committed
58
#include "lib/CLI11.hpp"
59 60
#include "lib/json.hpp"

Mikaël Salson's avatar
Mikaël Salson committed
61 62 63 64 65 66 67
#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"

68 69 70
// 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"

71
#define PROGNAME "vidjil-algo"
72
#define VIDJIL_JSON_VERSION "2016b"
Mikaël Salson's avatar
Mikaël Salson committed
73

Mathieu Giraud's avatar
Mathieu Giraud committed
74 75
//$$ #define (mainly default options)

76
#define DEFAULT_MULTI_GERMLINE_PATH "germline/"
77
#define DEFAULT_MULTI_GERMLINE_FILE "homo-sapiens.g"
Mikaël Salson's avatar
Mikaël Salson committed
78

79
#define DEFAULT_READ_HEADER_SEPARATOR " "
80
#define DEFAULT_READS  "./demo/Stanford_S22.fasta"
81
#define DEFAULT_MIN_READS_CLONE 5
82
#define DEFAULT_MAX_REPRESENTATIVES 100
83
#define DEFAULT_MAX_CLONES 100
84
#define DEFAULT_RATIO_READS_CLONE 0.0
85
#define NO_LIMIT "all"
Mikaël Salson's avatar
Mikaël Salson committed
86

Mathieu Giraud's avatar
Mathieu Giraud committed
87
#define COMMAND_WINDOWS "windows"
88
#define COMMAND_CLONES "clones"
Mikaël Salson's avatar
Mikaël Salson committed
89
#define COMMAND_SEGMENT "segment"
90
#define COMMAND_GERMLINES "germlines"
Mikaël Salson's avatar
Mikaël Salson committed
91
 
92
enum { CMD_WINDOWS, CMD_CLONES, CMD_SEGMENT, CMD_GERMLINES } ;
Mikaël Salson's avatar
Mikaël Salson committed
93

94 95 96
#define DEFAULT_OUT_DIR "./out/" 

// Fixed filenames/suffixes
97
#define CLONES_FILENAME ".vdj.fa"
Mikaël Salson's avatar
Mikaël Salson committed
98
#define CLONE_FILENAME "clone.fa-"
99 100
#define WINDOWS_FILENAME ".windows.fa"
#define SEGMENTED_FILENAME ".segmented.vdj.fa"
101
#define UNSEGMENTED_FILENAME ".unsegmented.vdj.fa"
102
#define UNSEGMENTED_DETAIL_FILENAME ".fa"
103
#define AFFECTS_FILENAME ".affects"
104 105 106
#define EDGES_FILENAME ".edges"
#define COMP_FILENAME "comp.vidjil"
#define JSON_SUFFIX ".vidjil"
Mikaël Salson's avatar
Mikaël Salson committed
107

108
#define DEFAULT_K      0
109
#define DEFAULT_W      50
Mikaël Salson's avatar
Mikaël Salson committed
110

Mikaël Salson's avatar
Mikaël Salson committed
111
#define DEFAULT_MAX_AUDITIONED 2000
Mathieu Giraud's avatar
Mathieu Giraud committed
112 113
#define DEFAULT_RATIO_REPRESENTATIVE 0.5

114
#define DEFAULT_KMER_THRESHOLD 1
Mathieu Giraud's avatar
Mathieu Giraud committed
115

Mikaël Salson's avatar
Mikaël Salson committed
116 117 118 119 120 121
#define DEFAULT_EPSILON  0
#define DEFAULT_MINPTS   10

#define DEFAULT_CLUSTER_COST  Cluster
#define DEFAULT_SEGMENT_COST   VDJ

122
#define DEFAULT_TRIM 0
Mikaël Salson's avatar
Mikaël Salson committed
123

124
#define MAX_CLONES_FOR_SIMILARITY 20
125

126
// warn
Mathieu Giraud's avatar
Mathieu Giraud committed
127
#define WARN_MAX_CLONES 5000
128
#define WARN_PERCENT_SEGMENTED 40
129
#define WARN_COVERAGE 0.6
130
#define WARN_NUM_CLONES_SIMILAR 10
131

Mikaël Salson's avatar
Mikaël Salson committed
132 133 134
// display
#define WIDTH_NB_READS 7
#define WIDTH_NB_CLONES 3
135

Mikaël Salson's avatar
Mikaël Salson committed
136 137

using namespace std ;
138
using json = nlohmann::json;
Mikaël Salson's avatar
Mikaël Salson committed
139

Mathieu Giraud's avatar
Mathieu Giraud committed
140
//$$ options: usage
Mikaël Salson's avatar
Mikaël Salson committed
141

Mathieu Giraud's avatar
Mathieu Giraud committed
142
extern char *optarg;
Mikaël Salson's avatar
Mikaël Salson committed
143 144 145

extern int optind, optopt, opterr;

146
string usage_examples(char *progname)
Mikaël Salson's avatar
Mikaël Salson committed
147
{
148 149
  stringstream ss;
  ss
Mathieu Giraud's avatar
Mathieu Giraud committed
150
       << "Examples (see doc/algo.org)" << endl
151 152 153
       << "  " << 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
154
       << "  " << progname << " -c clones   -g germline/homo-sapiens.g:IGH    -3     demo/Stanford_S22.fasta   # (restrict to complete recombinations on the IGH locus)" << endl
155 156 157
       << "  " << 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
158
       << "  " << progname << " -c germlines -g germline/homo-sapiens.g              demo/Stanford_S22.fasta   # (statistics on the k-mers)" << endl
Mikaël Salson's avatar
Mikaël Salson committed
159
    ;
160 161

  return ss.str();
Mikaël Salson's avatar
Mikaël Salson committed
162 163
}

Mathieu Giraud's avatar
Mathieu Giraud committed
164

165
int atoi_NO_LIMIT(const char *optarg)
Mathieu Giraud's avatar
Mathieu Giraud committed
166
{
167 168
  return strcmp(NO_LIMIT, optarg) ? atoi(optarg) : NO_LIMIT_VALUE ;
}
169
double atof_NO_LIMIT(const char *optarg)
170 171
{
  return strcmp(NO_LIMIT, optarg) ? atof(optarg) : NO_LIMIT_VALUE ;
Mathieu Giraud's avatar
Mathieu Giraud committed
172 173
}

174 175 176 177 178 179 180 181 182 183
string string_NO_LIMIT(string s)
{
  if (!strcmp(NO_LIMIT, s.c_str()))
    return NO_LIMIT_VALUE_STRING ;

  return s;
}



Mikaël Salson's avatar
Mikaël Salson committed
184 185
int main (int argc, char **argv)
{
186
  cout << "# " << PROGNAME << " -- V(D)J recombinations analysis <http://www.vidjil.org/>" << endl
187
       << "# Copyright (C) 2011-2018 by the Vidjil team" << endl
188
       << "# Bonsai bioinformatics at CRIStAL (UMR CNRS 9189, Université Lille) and Inria Lille" << endl 
Mathieu Giraud's avatar
Mathieu Giraud committed
189
       << endl
190
       << "# " << PROGNAME << " is free software, and you are welcome to redistribute it" << endl
Mathieu Giraud's avatar
Mathieu Giraud committed
191 192 193
       << "# 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
194
       << endl
195
       << "# Please cite http://biomedcentral.com/1471-2164/15/409 if you use " << PROGNAME << "." << endl
Mikaël Salson's avatar
Mikaël Salson committed
196 197
       << endl ;

198 199 200
  //////////////////////////////////
  // Display version information or git log

201 202
  string soft_version = PROGNAME ;
  soft_version += " " ;
203
#ifdef RELEASE_TAG
204
  cout << "# version: " PROGNAME << " " << RELEASE_TAG << endl ;
205 206 207 208 209 210 211 212 213 214
  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

215
  CLI::App app{"# vidjil-algo -- V(D)J recombinations analysis", argv[0]};
Mathieu Giraud's avatar
Mathieu Giraud committed
216

Mathieu Giraud's avatar
Mathieu Giraud committed
217
  //$$ options: defaults
Mathieu Giraud's avatar
Mathieu Giraud committed
218
  float ratio_representative = DEFAULT_RATIO_REPRESENTATIVE;
Mikaël Salson's avatar
Mikaël Salson committed
219
  unsigned int max_auditionned = DEFAULT_MAX_AUDITIONED;
220
  // int average_deletion = 4;     // Average number of deletion in V or J
Mathieu Giraud's avatar
Mathieu Giraud committed
221

222
  //$$ options: definition with CLI11
Mathieu Giraud's avatar
Mathieu Giraud committed
223
  string group = "";
Mathieu Giraud's avatar
Mathieu Giraud committed
224

225
  // ----------------------------------------------------------------------------------------------------------------------
226
  string f_reads = DEFAULT_READS ;
227 228 229 230 231 232 233
  app.add_option("reads_file", f_reads, R"Z(reads file, in one of the following formats:
                                  - FASTA (.fa/.fasta, .fa.gz/.fasta.gz)
                                  - FASTQ (.fq/.fastq, .fq.gz/.fastq.gz)
                                  - BAM (.bam)
                              Paired-end reads should be merged before given as an input to vidjil-algo.
                 )Z")
    -> required() -> set_type_name("");
234

235

236
  // ----------------------------------------------------------------------------------------------------------------------
Mathieu Giraud's avatar
Mathieu Giraud committed
237
  group = "Command selection";
238

Mathieu Giraud's avatar
Mathieu Giraud committed
239 240 241 242 243 244 245
  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");
Mikaël Salson's avatar
Mikaël Salson committed
246

247
  // ----------------------------------------------------------------------------------------------------------------------
Mathieu Giraud's avatar
Mathieu Giraud committed
248
  group = "Input" ;
249 250

  string read_header_separator = DEFAULT_READ_HEADER_SEPARATOR ;
251 252
  app.add_option("--header-sep", read_header_separator, "separator for headers in the reads file", false)
    -> group(group) -> level() -> set_type_name("CHAR='" DEFAULT_READ_HEADER_SEPARATOR "'");
Mikaël Salson's avatar
Mikaël Salson committed
253

254 255

  // ----------------------------------------------------------------------------------------------------------------------
Mathieu Giraud's avatar
Mathieu Giraud committed
256
  group = "Germline presets (at least one -g or -V/(-D)/-J option must be given)";
257

258
  vector <string> multi_germlines ;
Mathieu Giraud's avatar
Mathieu Giraud committed
259
  app.add_option("-g", multi_germlines, R"Z(
260
         -g <.g FILE>(:FILTER)
Mathieu Giraud's avatar
Mathieu Giraud committed
261 262 263
                    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'
264 265
         -g PATH
                    multiple locus/germlines, shortcut for '-g PATH/)Z" DEFAULT_MULTI_GERMLINE_FILE R"Z(',
Mathieu Giraud's avatar
Mathieu Giraud committed
266
                    processes human TRA, TRB, TRG, TRD, IGH, IGK and IGL locus, possibly with some incomplete/unusal recombinations)Z")
267 268
    -> group(group) -> set_type_name("GERMLINES");

269 270 271 272
  vector <string> v_reps_V ;
  vector <string> v_reps_D ;
  vector <string> v_reps_J ;
   
273 274 275 276
  app.add_option("-V", v_reps_V,
                 "custom V germline multi-fasta file(s)")
    -> group(group) -> set_type_name("FILE");

277

278
  app.add_option("-D", v_reps_D,
Mathieu Giraud's avatar
Mathieu Giraud committed
279
                 "custom D germline multi-fasta file(s), segment into V(D)J components")
280
    -> group(group) -> set_type_name("FILE");
281

282 283 284
  app.add_option("-J", v_reps_J,
                 "custom V germline multi-fasta file(s)")
    -> group(group) -> set_type_name("FILE");
285

286 287

  // ----------------------------------------------------------------------------------------------------------------------
Mathieu Giraud's avatar
Mathieu Giraud committed
288
  group = "Locus/recombinations";
289 290 291 292

  bool several_D = false;
  bool multi_germline_unexpected_recombinations_12 = false;

Mathieu Giraud's avatar
Mathieu Giraud committed
293
  app.add_flag("-d", several_D, "try to detect several D (experimental)") -> group(group);
294
  app.add_flag("-2", multi_germline_unexpected_recombinations_12, "try to detect unexpected recombinations (must be used with -g)") -> group(group);
295

296

297 298
  // ----------------------------------------------------------------------------------------------------------------------
  group = "Recombination detection (\"window\" prediction, first pass)";
299 300 301
  group += "\n    (use either -s or -k option, but not both)";
  group += "\n    (using -k option is equivalent to set with -s a contiguous seed with only '#' characters)" ;
  group += "\n    (all these options, except -w, are overriden when using -g)";
302 303 304 305

  int options_s_k = 0 ;

  IndexTypes indexType = KMER_INDEX;
306
  app.add_flag_function("-q",
307
                        [&](size_t n) { UNUSED(n); indexType = AC_AUTOMATON; },
308 309
                        "use Aho-Corasick-like automaton (experimental)")
    -> group(group) -> level();
310

311 312
  string seed = DEFAULT_SEED ;
  bool seed_changed = false;
313 314 315 316 317 318 319 320 321 322 323
  app.add_option("-k",
                 [&](CLI::results_t res) {
                   int kmer_size ;
                   bool worked = CLI::detail::lexical_cast(res[0], kmer_size);
                   if (worked) {
                     seed = seed_contiguous(kmer_size);
                     seed_changed = true;
                     options_s_k++ ;
                   }
                   return worked;
                 },
324
                 "k-mer size used for the V/J affectation (default: 10, 12, 13, depends on germline)")
325
    -> group(group) -> level() -> set_type_name("INT");
326

327
  int wmer_size = DEFAULT_W ;
328
  app.add_option("-w", wmer_size,
329
                 "w-mer size used for the length of the extracted window ('" NO_LIMIT "': use all the read, no window clustering)")
330
    -> group(group) -> level() -> transform(string_NO_LIMIT);
331

332 333

  double expected_value = THRESHOLD_NB_EXPECTED;
334
  app.add_option("-e", expected_value,
335
                 "maximal e-value for determining if a V-J segmentation can be trusted", true)
336
    -> group(group) -> level() -> transform(string_NO_LIMIT);
337

338 339
  int trim_sequences = DEFAULT_TRIM;
  bool trim_sequences_changed = false;
340 341
  app.add_option("-t",
                 [&](CLI::results_t res) {
Mikaël Salson's avatar
Mikaël Salson committed
342
                   CLI::detail::lexical_cast(res[0], trim_sequences);
343 344 345 346
                   trim_sequences_changed = true;
                   return true;
                 },
                 // trim_sequences,
347
                 "trim V and J genes (resp. 5' and 3' regions) to keep at most <INT> nt  (0: no trim)")
348
    -> group(group) -> level() ->  set_type_name("INT");
Mikaël Salson's avatar
Mikaël Salson committed
349

350 351 352 353 354 355 356
  app.add_option("-s",
                 [&](CLI::results_t res) {
                   seed = res[0] ;
                   options_s_k++ ;
                   seed_changed = true;
                   return true;
                 },
Mathieu Giraud's avatar
Mathieu Giraud committed
357
                 "seed, possibly spaced, used for the V/J affectation (default: depends on germline), given either explicitely or by an alias"
358 359 360
                 "\n                             " + string_of_map(seedMap, " ")
                 )
    -> group(group) -> level() -> set_type_name("SEED=" DEFAULT_SEED);
361

Mikaël Salson's avatar
Mikaël Salson committed
362

363 364
  // ----------------------------------------------------------------------------------------------------------------------
  group = "Recombination detection, experimental options (do not use)";
365

366 367 368 369
  bool multi_germline_mark = false;
  bool multi_germline_one_unique_index = false;
  bool multi_germline_unexpected_recombinations_1U = false;

370 371 372 373
  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();

374
  app.add_flag("-1", multi_germline_one_unique_index,
375 376 377 378 379 380 381
               "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();

382
  bool keep_unsegmented_as_clone = false;
383 384 385
  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();
386

387

388
  // ----------------------------------------------------------------------------------------------------------------------
389
  group = "Labeled sequences (windows related to these sequences will be kept even if -r/--ratio thresholds are not reached)";
390 391 392 393

  vector <string> windows_labels_explicit ;
  string windows_labels_file = "" ;

394 395
  app.add_option("-W", windows_labels_explicit, "label the given sequence(s)") -> group(group) -> level() -> set_type_name("SEQUENCE");
  app.add_option("-l", windows_labels_file, "label a set of sequences given in <file>") -> group(group) -> level() -> set_type_name("FILE");
396 397

  bool only_labeled_windows = false ;
398
  app.add_flag("-F", only_labeled_windows, "filter -- keep only the windows related to the labeled sequences") -> group(group) -> level();
Mikaël Salson's avatar
Mikaël Salson committed
399 400


401 402
  // ----------------------------------------------------------------------------------------------------------------------
  group = "Limits to report a clone (or a window)";
403
  int max_clones_id = NO_LIMIT_VALUE ;
404 405 406
  int min_reads_clone = DEFAULT_MIN_READS_CLONE ;
  float ratio_reads_clone = DEFAULT_RATIO_READS_CLONE;

407
  app.add_option("--max-clones", max_clones_id, "maximal number of output clones ('" NO_LIMIT "': no maximum, default)", false) -> group(group);
408 409 410 411 412 413
  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 (second pass)";
414

415 416 417
  int max_clones = DEFAULT_MAX_CLONES ;
  int max_representatives = DEFAULT_MAX_REPRESENTATIVES ;

418
  app.add_option("-y", max_representatives,
419 420
                 "maximal number of clones computed with a consensus sequence ('" NO_LIMIT "': no limit)", true)
    -> group(group) -> transform(string_NO_LIMIT);
421 422 423 424 425 426 427 428 429

  app.add_option("-z",
                 [&max_clones, &max_representatives](CLI::results_t res) {
                   max_clones = atoi_NO_LIMIT(res[0].c_str());
                   if ((max_representatives < max_clones) && (max_representatives != NO_LIMIT_VALUE))
                     max_representatives = max_clones ;
                   return true;
                   // TODO: return false on bad input
                 },
430 431
                 "maximal number of clones to be analyzed with a full V(D)J designation ('" NO_LIMIT "': no limit, do not use)")
    -> group(group) -> set_type_name("INT=" + string_of_int(max_clones));
432 433

  app.add_flag_function("-A", [&](size_t n) {
434
      UNUSED(n);
435 436 437 438 439
      ratio_reads_clone = 0 ;
      min_reads_clone = 1 ;
      max_representatives = NO_LIMIT_VALUE ;
      max_clones = NO_LIMIT_VALUE ;
    },
440 441
    "reports and segments all clones (-r 0 --ratio 0 -y " NO_LIMIT " -z " NO_LIMIT "), to be used only on very small datasets (for example -AX 20)")
    -> group(group);
442

443 444 445
  int max_reads_processed = NO_LIMIT_VALUE;
  int max_reads_processed_sample = NO_LIMIT_VALUE;

446
  app.add_option("-x", max_reads_processed,
447 448 449
                 "maximal number of reads to process ('" NO_LIMIT "': no limit, default), only first reads")
    -> group(group) -> transform(string_NO_LIMIT);

450
  app.add_option("-X", max_reads_processed_sample,
451 452
                 "maximal number of reads to process ('" NO_LIMIT "': no limit, default), sampled reads")
    -> group(group) -> transform(string_NO_LIMIT);
453 454 455 456


  // ----------------------------------------------------------------------------------------------------------------------
  group = "Clone analysis (second pass)";
457

458
  Cost segment_cost = DEFAULT_SEGMENT_COST ;
459 460 461 462 463
  app.add_option("-f",
                 [&segment_cost](CLI::results_t res) {
                   segment_cost = strToCost(res[0].c_str(), VDJ); 
                   return true;
                 },
464 465
                 "use custom Cost for fine segmenter : format \"match, subst, indels, del_end, homo\" (default " + string_of_cost(DEFAULT_SEGMENT_COST) + ")")
    -> group(group) -> level() -> set_type_name("COST");
Mathieu Giraud's avatar
Mathieu Giraud committed
466

467
  double expected_value_D = THRESHOLD_NB_EXPECTED_D;
468 469 470
  app.add_option("-E", expected_value_D,
                 "maximal e-value for determining if a D segment can be trusted", true)
    -> group(group) -> level();
471

472
  int kmer_threshold = DEFAULT_KMER_THRESHOLD;
473
  app.add_option("-Z", kmer_threshold,
474
                 "typical number of V genes, selected by k-mer comparison, to compare to the read ('" NO_LIMIT "': all genes)", true)
475
    -> group(group) -> transform(string_NO_LIMIT) -> level();
476

477
  bool detect_CDR3 = false;
478 479
  app.add_flag("-3,--cdr3", detect_CDR3, "CDR3/JUNCTION detection (requires gapped V/J germlines)")
    -> group(group);
480

481
  int alternative_genes = 0;
482
  app.add_option("--alternative-genes", alternative_genes, "number of alternative V(D)J genes to show beyond the most similar one", true)
483
    -> group(group) -> level();
484 485
  // ----------------------------------------------------------------------------------------------------------------------
  group = "Additional clustering (third pass, experimental)" ;
486 487 488

  int epsilon = DEFAULT_EPSILON ;
  int minPts = DEFAULT_MINPTS ;
489
  app.add_option("-n", epsilon, "minimum required neighbors for automatic clustering. No automatic clusterisation if =0.", true) -> group(group) -> level();
490
  app.add_option("-N", minPts, "minimum required neighbors for automatic clustering", true) -> group(group) -> level();
491 492 493

  bool save_comp = false;
  bool load_comp = false;
494 495
  app.add_flag("-S", save_comp, "generate and save comparative matrix for clustering") -> group(group) -> level();
  app.add_flag("-L", load_comp, "load comparative matrix for clustering") -> group(group) -> level();
496 497

  string forced_edges = "" ;
498
  app.add_option("--forced-edges", forced_edges, "manual clustering -- a file used to force some specific edges") -> group(group) -> level() -> set_type_name("FILE");
499

500
  Cost cluster_cost = DEFAULT_CLUSTER_COST ;
501 502 503 504 505
  app.add_option("-C",
                 [&cluster_cost](CLI::results_t res) {
                   cluster_cost = strToCost(res[0].c_str(), Cluster);
                   return true;
                 },
506 507
                 "use custom Cost for automatic clustering : format \"match, subst, indels, del_end, homo\" (default " + string_of_cost(DEFAULT_CLUSTER_COST) + ")")
    -> group(group) -> level() -> set_type_name("COST");
Mathieu Giraud's avatar
Mathieu Giraud committed
508 509

  
510
  // ----------------------------------------------------------------------------------------------------------------------
Mathieu Giraud's avatar
Mathieu Giraud committed
511
  group = "Detailed output per read (generally not recommended, large files, but may be used for filtering, as in -uu -X 1000)";
512 513

  bool output_segmented = false;
514 515 516 517
  app.add_flag("-U", output_segmented,
               "output segmented reads (in " SEGMENTED_FILENAME " file)")
    -> group(group);

518 519 520 521
  bool output_unsegmented = false;
  bool output_unsegmented_detail = false;
  bool output_unsegmented_detail_full = false;

Mathieu Giraud's avatar
Mathieu Giraud committed
522
  app.add_flag_function("-u", [&](size_t n) {
523 524 525
      output_unsegmented = (n >= 3);             // -uuu
      output_unsegmented_detail_full = (n >= 2); // -uu
      output_unsegmented_detail = (n >= 1);      // -u
Mathieu Giraud's avatar
Mathieu Giraud committed
526 527 528
    }, 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)
529 530
        -uuu        output unsegmented reads, all reads, including a )Z" UNSEGMENTED_FILENAME R"Z( file (use only for debug))Z")
    -> group(group);
Mathieu Giraud's avatar
Mathieu Giraud committed
531

532
  bool output_affects = false;
533 534 535
  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);
536

537

538
  // ----------------------------------------------------------------------------------------------------------------------
Mathieu Giraud's avatar
Mathieu Giraud committed
539
  group = "Output";
540 541 542 543

  string out_dir = DEFAULT_OUT_DIR;
  string f_basename = "";

544
  app.add_option("-o", out_dir, "output directory", true) -> group(group) -> set_type_name("PATH");
545
  app.add_option("-b", f_basename, "output basename (by default basename of the input file)") -> group(group) -> set_type_name("STRING");
546 547

  bool output_sequences_by_cluster = false;
Mathieu Giraud's avatar
Mathieu Giraud committed
548
  app.add_flag("-a", output_sequences_by_cluster, "output all sequences by cluster (" CLONE_FILENAME "*), to be used only on small datasets") -> group(group);
549 550

  int verbose = 0 ;
Mathieu Giraud's avatar
Mathieu Giraud committed
551
  app.add_flag_function("-v", [&](size_t n) { verbose += n ; }, "verbose mode") -> group(group);
552

553

554 555 556 557 558
  // ----------------------------------------------------------------------------------------------------------------------
  group = "Help";
  app.set_help_flag("-h", "help")
    -> group(group);

559
  app.add_flag_function("-H", [&](size_t n) { UNUSED(n); throw CLI::CallForAdvancedHelp() ; },
560 561
                        "help, including advanced and experimental options"
                        "\n                              "
562 563
                        "The full help is available in the doc/algo.org file.")
    -> group(group);
Mikaël Salson's avatar
Mikaël Salson committed
564 565


566
  // ----------------------------------------------------------------------------------------------------------------------
567 568
  app.set_footer(usage_examples(argv[0]));

Mathieu Giraud's avatar
Mathieu Giraud committed
569 570
  //$$ options: parsing
  CLI11_PARSE(app, argc, argv);
Mathieu Giraud's avatar
Mathieu Giraud committed
571

Mathieu Giraud's avatar
Mathieu Giraud committed
572
  //$$ options: post-processing+display
Mikaël Salson's avatar
Mikaël Salson committed
573

574
  int command = CMD_CLONES;
Mathieu Giraud's avatar
Mathieu Giraud committed
575 576 577 578 579 580 581 582 583 584
  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;
585
    throw CLI::CallForHelp();
Mathieu Giraud's avatar
Mathieu Giraud committed
586
  }
587

Mathieu Giraud's avatar
Mathieu Giraud committed
588 589 590
  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());
591

592

593 594 595
  list <pair <string, string>> multi_germline_paths_and_files ;
  bool multi_germline = false;

Mathieu Giraud's avatar
Mathieu Giraud committed
596 597 598 599 600 601 602 603 604 605 606
  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)) ;
607 608
              continue ;
            }
Mathieu Giraud's avatar
Mathieu Giraud committed
609 610 611 612 613 614 615
        }

      // 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)));
    }


616
  if (!multi_germline && (!f_reps_V.size() || !f_reps_J.size()))
617
    {
618
      cerr << ERROR_STRING << "At least one germline must be given with -g or -V/(-D)/-J." << endl ;
619
      return 1;
620 621
    }

622 623
  if (options_s_k > 1)
    {
624
      cerr << ERROR_STRING << "Use at most one -s or -k option." << endl ;
625
      return 1;
626 627
    }

628 629
  map <string, string> windows_labels ;

630 631 632
  for(string lab : windows_labels_explicit)
    windows_labels[lab] = string("-W");
  
Mikaël Salson's avatar
Mikaël Salson committed
633 634
  string out_seqdir = out_dir + "/seq/" ;

Mikaël Salson's avatar
Mikaël Salson committed
635 636 637
  if (verbose)
    cout << "# verbose " << verbose << endl ;

638
  if (f_reads == DEFAULT_READS)
Mikaël Salson's avatar
Mikaël Salson committed
639 640 641
    {
      cout << "# using default sequence file: " << f_reads << endl ;
    }
Mathieu Giraud's avatar
Mathieu Giraud committed
642

643 644 645 646 647
  //  else
  //  {
  //    cerr << ERROR_STRING << "Wrong number of arguments." << endl ;
  //    return 1;
  //  }
Mathieu Giraud's avatar
Mathieu Giraud committed
648

649
  size_t min_cover_representative = (size_t) (min_reads_clone < (int) max_auditionned ? min_reads_clone : max_auditionned) ;
650

Mathieu Giraud's avatar
Mathieu Giraud committed
651 652 653
  // Check seed buffer  
  if (seed.size() >= MAX_SEED_SIZE)
    {
654
      cerr << ERROR_STRING << "Seed size is too large (MAX_SEED_SIZE)." << endl ;
655
      return 1;
Mathieu Giraud's avatar
Mathieu Giraud committed
656
    }
657

Mathieu Giraud's avatar
Mathieu Giraud committed
658
  if ((wmer_size< 0) && (wmer_size!= NO_LIMIT_VALUE))
659
    {
660
      cerr << ERROR_STRING << "Too small -w. The window size should be positive" << endl;
661
      return 1;
662 663
    }

Mikaël Salson's avatar
Mikaël Salson committed
664 665 666 667
  // 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) {
668
    cerr << ERROR_STRING << "Directory creation: " << out_dir << endl; perror("");
669
    return 2;
Mikaël Salson's avatar
Mikaël Salson committed
670 671
  }

Mikaël Salson's avatar
Mikaël Salson committed
672 673
  const char *outseq_cstr = out_seqdir.c_str();
  if (mkpath(outseq_cstr, 0755) == -1) {
674
    cerr << ERROR_STRING << "Directory creation: " << out_seqdir << endl; perror("");
675
    return 2;
Mikaël Salson's avatar
Mikaël Salson committed
676 677
  }

678 679 680 681 682
  // Compute basename if not given as an option
  if (f_basename == "") {
    f_basename = extract_basename(f_reads);
  }

Mikaël Salson's avatar
Mikaël Salson committed
683 684 685
  out_dir += "/" ;

  /// Load labels ;
686
  load_into_map(windows_labels, windows_labels_file, "-l");
Mikaël Salson's avatar
Mikaël Salson committed
687 688

  switch(command) {
Mathieu Giraud's avatar
Mathieu Giraud committed
689
  case CMD_WINDOWS: cout << "Extracting windows" << endl; 
Mikaël Salson's avatar
Mikaël Salson committed
690
    break;
691
  case CMD_CLONES: cout << "Analysing clones" << endl; 
Mikaël Salson's avatar
Mikaël Salson committed
692 693 694
    break;
  case CMD_SEGMENT: cout << "Segmenting V(D)J" << endl;
    break;
695 696
  case CMD_GERMLINES: cout << "Discovering germlines" << endl;
    break;
Mikaël Salson's avatar
Mikaël Salson committed
697 698
  }

Mathieu Giraud's avatar
Mathieu Giraud committed
699
  cout << "Command line: ";
Mikaël Salson's avatar
Mikaël Salson committed
700
  for (int i=0; i < argc; i++) {
Mathieu Giraud's avatar
Mathieu Giraud committed
701
    cout << argv[i] << " ";
Mikaël Salson's avatar
Mikaël Salson committed
702
  }
Mathieu Giraud's avatar
Mathieu Giraud committed
703
  cout << endl;
Mikaël Salson's avatar
Mikaël Salson committed
704 705 706 707 708 709 710 711 712 713 714 715

  //////////////////////////////////
  // 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);

Mathieu Giraud's avatar
Mathieu Giraud committed
716
  cout << "# " << time_buffer << endl ;
Mikaël Salson's avatar
Mikaël Salson committed
717 718


719 720
  //////////////////////////////////
  // Warning for non-optimal use
721

722
  if (max_clones == NO_LIMIT_VALUE || max_clones > WARN_MAX_CLONES)
723 724
    {
      cout << endl
725
	   << "* WARNING: " << PROGNAME << " was run with '-A' option or with a large '-z' option" << endl ;
726 727 728 729 730
    }
  
  if (command == CMD_SEGMENT)
    {
      cout << endl
731
	   << "* WARNING: " << PROGNAME << " was run with '-c segment' option" << endl ;
732 733
    }
  
734
  if (max_clones == NO_LIMIT_VALUE || max_clones > WARN_MAX_CLONES || command == CMD_SEGMENT)
735
    {
736
      cout << "* " << PROGNAME << " efficientl extracts windows overlapping the CDR3" << endl
737
           << "* to cluster reads into clones ('-c clones')." << endl
738 739
           << "* 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
Mathieu Giraud's avatar
Mathieu Giraud committed
740
	   << "* More information is provided in the 'doc/vidjil-algo.md' file." << endl 
741 742
	   << endl ;
    }
743

744

745 746 747 748 749 750 751
  //

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


752 753 754 755 756 757 758 759 760
  /////////////////////////////////////////
  //            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] << " ";

761
  SampleOutput *output = new SampleOutput({
762 763 764 765 766 767 768 769
    {"vidjil_json_version", VIDJIL_JSON_VERSION},
    {"samples", {
        {"number", 1},
        {"original_names", {f_reads}},
        {"run_timestamp", {time_buffer}},
        {"producer", {soft_version}},
        {"commandline", {stream_cmdline.str()}}
      }}
770
  });
771 772


773 774 775
  /////////////////////////////////////////
  //            LOAD GERMLINES           //
  /////////////////////////////////////////
776

777 778 779
  if (command == CMD_GERMLINES)
    {
      multi_germline = true ;
780
      multi_germline_one_unique_index = true ;
781 782
    }

783
  MultiGermline *multigermline = new MultiGermline(indexType, !multi_germline_one_unique_index);
784

785
    {
786
      cout << "Load germlines and build Kmer indexes" << endl ;
787 788 789
    
      if (multi_germline)
	{
790
          for (pair <string, string> path_file: multi_germline_paths_and_files)
791 792
            {
              try {
793
                multigermline->build_from_json(path_file.first, path_file.second, GERMLINES_REGULAR,
794
                                               FIRST_IF_UNCHANGED("", seed, seed_changed),
795
                                               FIRST_IF_UNCHANGED(0, trim_sequences, trim_sequences_changed), (kmer_threshold != NO_LIMIT_VALUE));
796
              } catch (std::exception& e) {
797
                cerr << ERROR_STRING << PROGNAME << " cannot properly read " << path_file.first << "/" << path_file.second << ": " << e.what() << endl;
798
                delete multigermline;
799
                return 1;
800 801
              }
            }
802 803 804 805 806
	}
      else
	{
	  // Custom germline
	  Germline *germline;
807
	  germline = new Germline("custom", 'X',
808 809
                                  f_reps_V, f_reps_D, f_reps_J,
                                  seed, trim_sequences, (kmer_threshold != NO_LIMIT_VALUE));
810

811
          germline->new_index(indexType);
812

813 814 815
	  multigermline->insert(germline);
	}
    }
816

817
    cout << endl ;
818

819
    if (multi_germline_one_unique_index) {
820
      multigermline->build_with_one_index(seed, true);
821
    }
822

823
      if (multi_germline_unexpected_recombinations_12 || multi_germline_unexpected_recombinations_1U) {
824 825 826
        if (!multigermline->index) {
          multigermline->build_with_one_index(seed, false);
        }
827
      }
828

829
      if (multi_germline_unexpected_recombinations_12) {
830
        Germline *pseudo = new Germline(PSEUDO_UNEXPECTED, PSEUDO_UNEXPECTED_CODE, "", trim_sequences, (kmer_threshold != NO_LIMIT_VALUE));
831
        pseudo->seg_method = SEG_METHOD_MAX12 ;
832
        pseudo->set_index(multigermline->index);
833
        multigermline->germlines.push_back(pseudo);
834 835 836
      }

      if (multi_germline_unexpected_recombinations_1U) {