Mise à jour terminée. Pour connaître les apports de la version 13.8.4 par rapport à notre ancienne version vous pouvez lire les "Release Notes" suivantes :
https://about.gitlab.com/releases/2021/02/11/security-release-gitlab-13-8-4-released/
https://about.gitlab.com/releases/2021/02/05/gitlab-13-8-3-released/

segment.h 12 KB
Newer Older
Mikaël Salson's avatar
Mikaël Salson committed
1 2 3 4
#ifndef SEGMENT_H
#define SEGMENT_H

#include <string>
5
#include <fstream>
Mikaël Salson's avatar
Mikaël Salson committed
6
#include <iostream>
7
#include "bioreader.hpp"
Mikaël Salson's avatar
Mikaël Salson committed
8 9
#include "dynprog.h"
#include "tools.h"
10
#include "germline.h"
Mikaël Salson's avatar
Mikaël Salson committed
11 12
#include "kmerstore.h"
#include "kmeraffect.h"
Mathieu Giraud's avatar
Mathieu Giraud committed
13
#include "affectanalyser.h"
Marc Duez's avatar
Marc Duez committed
14
#include "../lib/json.hpp"
Mikaël Salson's avatar
Mikaël Salson committed
15

16 17
// #define DEBUG_EVALUE

Mikaël Salson's avatar
Mikaël Salson committed
18
#define EXTEND_D_ZONE 5
19

20
#define RATIO_STRAND 2          /* The ratio between the affectations in one
Mikaël Salson's avatar
Mikaël Salson committed
21 22
                                   strand and the other, to safely attribute a
                                   segment to a given strand */
Mikaël Salson's avatar
Mikaël Salson committed
23

24 25
#define DETECT_THRESHOLD_STRAND 5   /* If the number of total affectations
                                       is above this threshold, then a sequence with no clearly attributed
26
                                       strand will be marked as STRAND_NOT_CONSISTENT */
27

Mathieu Giraud's avatar
Mathieu Giraud committed
28 29
#define JSON_REMEMBER_BEST  4   /* The number of V/D/J predictions to keep  */

30
#define BAD_EVALUE  1e10
31

32
#define THRESHOLD_NB_EXPECTED 1.0 /* Threshold of the accepted expected value for number of found k-mers */
33
#define THRESHOLD_NB_EXPECTED_D  .05 /* e-value threshold, D-REGION */
34

35 36 37 38
#define BOTTOM_TRIANGLE_SHIFT  20   /* Should equal to (max allowed 'k-band') + (max allowed number of V/J deletions) - (min size to recognize facing J/V)
                                       As we need ~10 bp to recognize the facing V/J, this value should be large enough to handle V/J deletions until ~30 bp,
                                       (and even larger V/J deletions if there is a large facing J or V in the read). */

39 40 41
#define DEFAULT_WINDOW_SHIFT 5  /* Number of nucleotide to try shifting or
                                   reducing the window when it doesn't fit at
                                   its place */
42
#define MINIMAL_WINDOW_LENGTH 30 /* As we now dynamically adapt the window
43 44
                                    length we need to specify a minimum
                                    otherwise we could go as low as
45 46 47
                                    2*DEFAULT_WINDOW_SHIFT. Of course this can
                                    be overriden by the command line by
                                    providing a shorter -w*/
48

49
#define DEFAULT_MINIMIZE_WIDTH       12   /* Number of nucleotides on which the minimization integer is computed */
50

51 52 53 54 55
#define DEFAULT_MINIMIZE_ONE_MARGIN  20   /* Number of nucleotides on ends of the reads excluded from the minimization in SEG_METHOD_ONE.
                                             - A read smaller than (DEFAULT_MINIMIZE_ONE_MARGIN*2 + max(k, DEFAULT_MINIMIZE_WIDTH)), that is 52,
                                               will trigger UNSEG_TOO_SHORT_FOR_WINDOW.
                                             - Margins above the half of the window length may produce shortened or shifted windows
                                          */
56 57


Mikaël Salson's avatar
Mikaël Salson committed
58
using namespace std;
Marc Duez's avatar
Marc Duez committed
59
using json = nlohmann::json;
Mikaël Salson's avatar
Mikaël Salson committed
60

61 62 63
enum SEGMENTED { NOT_PROCESSED,
		 TOTAL_SEG_AND_WINDOW,
                 SEG_PLUS, SEG_MINUS,
64
                 SEG_CHANGED_WINDOW,
65
                 UNSEG_TOO_SHORT, UNSEG_STRAND_NOT_CONSISTENT,
66
		 UNSEG_TOO_FEW_ZERO,  UNSEG_ONLY_V, UNSEG_ONLY_J,
67
                 UNSEG_BAD_DELTA_MIN, UNSEG_AMBIGUOUS,
68
		 UNSEG_TOO_SHORT_FOR_WINDOW,
69

Mikaël Salson's avatar
Mikaël Salson committed
70
		 STATS_SIZE } ;
71 72 73

#define STATS_FIRST_UNSEG UNSEG_TOO_SHORT

74 75 76
const char* const segmented_mesg[] = { "?",
                                       "SEG",
                                       "SEG_+", "SEG_-",
77
                                       "SEG changed w",
78
                                       "UNSEG too short", "UNSEG strand",
79
				       "UNSEG too few V/J", "UNSEG only V/5'", "UNSEG only J/3'",
80
				       "UNSEG < delta_min", "UNSEG ambiguous",
81
                                       "UNSEG too short w",
Mikaël Salson's avatar
Mikaël Salson committed
82
                                      } ;
Mikaël Salson's avatar
Mikaël Salson committed
83

84 85


86 87 88 89
/**
 * An alignment box (AlignBox) gather all parameters for a recombined gene segment (V, D, J, other D...)
 **/

90 91 92
class AlignBox
{
 public:
93
  string key;
94
  string color;
95

96 97 98 99 100
  int del_left;
  int start;
  int end;
  int del_right;

101
  AlignBox(string key = "", string color="");
102
  string getSequence(string sequence);
103
  void addToJson(json &seg);
104

105 106 107 108 109
  /**
   * Returns 'V', 'D', 'J', or possibly '5', '4', '3', '?', depending on the ref_label and on the key
   */
  char getInitial();

110 111 112 113 114
  /**
   * Returns the length
   */
  int getLength();

115 116 117 118 119 120 121 122 123 124 125 126
  /**
   * Returns the position in the reference string corresponding to the position in the read
   * Preliminary implementation, only works for the start of V and J boxes
   */
  int posInRef(int i);

  /**
   * Format the reference string in a given range position, possibly adding the ANSI colors
   * where there is the alignment
   */
  string refToString(int from, int to);

127
  /* Identifier, label and sequence of the reference sequence (the best one) */
128 129 130 131
  int ref_nb;
  string ref_label;
  string ref;

132 133 134
  /* Marked position, for Cys104 and Phe118/Trp118 */
  int marked_pos;

135 136
  /* Identifiers and scores of other possible reference sequence */
  vector<pair<int, int> > score;
137 138
};

139
ostream &operator<<(ostream &out, const AlignBox &box);
140

141
/**
142
 * Check whether there is an overlap between two boxes,
143 144 145
 * If this is the case, fix this overlap (finding the best split point), and update segmentation accordingly
 * @param seq:                             the read
 * @param seq_begin, seq_end:              the positions to consider on 'seq' for the two sequences that may overlap
146
 * @param *box_left, *box_right            the two boxes
147 148
 * @param segment_cost:                    the cost used by the dynamic programing
 *
149 150
 * @post  box_left->del_left and box_right->del_right are set to the best number of nucleotides to trim in order to remove the overlap.
 *        box_left->end and box_right->start are shifted by the good number of nucleotides
151 152 153 154 155
 *
 * @return                                 the N segment
 */

string check_and_resolve_overlap(string seq, int seq_begin, int seq_end,
156 157
                                 AlignBox *box_left, AlignBox *box_right,
                                 Cost segment_cost);
158

Mikaël Salson's avatar
Mikaël Salson committed
159 160 161
class Segmenter {
protected:
  string sequence;
162
  string sequence_or_rc;
163

164
  // JUNCTIONstart/end and CDR3start/end are 1-based
165
  int JUNCTIONstart, JUNCTIONend;
166
  string JUNCTIONaa;
167
  bool JUNCTIONproductive;
168

Marc Duez's avatar
Marc Duez committed
169
  int CDR3start, CDR3end;
170 171
  string CDR3nuc;
  string CDR3aa;
172

Mikaël Salson's avatar
Mikaël Salson committed
173
  bool reversed, segmented, dSegmented;
174
  bool junctionChanged;
175
  int because;
Mikaël Salson's avatar
Mikaël Salson committed
176

177 178
  /**
   * Compares evalue_left, evalue_right and evalue against the provided threshold
179
   * @post some evalue is above the threshold ==> because is set to UNSEG_TOO_FEW_ZERO, UNSEG_ONLY_J or UNSEG_ONLY_V
180 181 182
   */
  void checkLeftRightEvaluesThreshold(double threshold, int strand);

Mikaël Salson's avatar
Mikaël Salson committed
183 184 185 186 187
  string removeChevauchement();
  bool finishSegmentation();
  bool finishSegmentationD();

 public:
188
  Germline *segmented_germline;
189
  string label;
Mikaël Salson's avatar
Mikaël Salson committed
190
  string code;
191 192
  string info;        // .vdj.fa header, fixed fields
  string info_extra;  // .vdj.fa header, other information, at the end of the header
193
  string seg_V, seg_N, seg_J, system;
194

195 196
  AlignBox *box_V, *box_D, *box_J;

197
  double evalue;
198 199
  double evalue_left;
  double evalue_right;
Mikaël Salson's avatar
Mikaël Salson committed
200 201 202
  string seg_N1, seg_D, seg_N2;
  Cost segment_cost;

Mathieu Giraud's avatar
Mathieu Giraud committed
203
  virtual ~Segmenter();
Mikaël Salson's avatar
Mikaël Salson committed
204

Mathieu Giraud's avatar
Mathieu Giraud committed
205
  /* Queries */
Mikaël Salson's avatar
Mikaël Salson committed
206 207 208 209 210 211 212 213 214 215 216 217 218



  Sequence getSequence() const ;

  /**
   * @param l: length around the junction
   * @return the string centered on the junction (ie. at position
   *         (getLeft() + getRight())/2).
   *         The string has length l unless the original string 
   *         is not long enough.
   *         The junction is revcomp-ed if the original string comes from reverse
   *         strand.
219
   * @post If the size or position of the window had to be dynamically adapted to fit
220
   *       in the read, isJunctionChanged() will return true
Mikaël Salson's avatar
Mikaël Salson committed
221
   */
222
  string getJunction(int l, int shift=0);
Mikaël Salson's avatar
Mikaël Salson committed
223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258

  /**
   * @return the left position (on forward strand) of the segmentation.
   */
  int getLeft() const;
  
  /**
   * @return the right position (on forward strand) of the segmentation
   */
  int getRight() const;
  
  /**
   * @return the left position (on forward strand) of the D segmentation.
   */
  int getLeftD() const;
  
  /**
   * @return the right position (on forward strand) of the D segmentation
   */
  int getRightD() const;

  /**
   * @return true iff the string comes from reverse strand
   */
  bool isReverse() const;

  /**
   * @return true iff the sequence has been successfully segmented
   */
  bool isSegmented() const;
  
  /**
   * @return true if a D gene was found in the N region
   */
  bool isDSegmented() const;

259 260 261 262
  /**
   * @return true iff the junction has been shifted or shortened
   *              dynamically to fit into the sequence
   */
263
  bool isJunctionChanged() const;
264

265 266 267 268 269 270 271 272 273
  /**
   * @return the status of the segmentation. Tells if the Sequence has been segmented
   *         of if it has not, what the reason is.
   * @assert getSegmentationStatus() == SEG_PLUS || getSegmentationStatus() == SEG_MINUS
   *         <==> isSegmented()
   */
  int getSegmentationStatus() const;

  string getInfoLine() const;
Mikaël Salson's avatar
Mikaël Salson committed
274

275 276 277 278 279
  /**
   * @post status == SEG_PLUS || status == SEG_MINUS <==> isSegmented()
   */
  void setSegmentationStatus(int status);

Mikaël Salson's avatar
Mikaël Salson committed
280 281 282 283 284 285 286 287 288 289 290
  friend ostream &operator<<(ostream &out, const Segmenter &s);
};



ostream &operator<<(ostream &out, const Segmenter &s);



class KmerSegmenter : public Segmenter
{
Mathieu Giraud's avatar
Mathieu Giraud committed
291
 private:
292
  KmerAffectAnalyser *kaa;
Mikaël Salson's avatar
Mikaël Salson committed
293 294 295 296
 protected:
  string affects;

 public:
297
  int score;
298 299
  int pvalue_left;
  int pvalue_right;
300
  KmerAffect before, after;
301 302

  KmerSegmenter();
Mikaël Salson's avatar
Mikaël Salson committed
303 304 305
  /**
   * Build a segmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
306
   * @param germline: the germline
Mikaël Salson's avatar
Mikaël Salson committed
307
   */
308
  KmerSegmenter(Sequence seq, Germline *germline, double threshold = THRESHOLD_NB_EXPECTED, double multiplier=1.0);
Mathieu Giraud's avatar
Mathieu Giraud committed
309

310 311
  KmerSegmenter(const KmerSegmenter &seg);

Mathieu Giraud's avatar
Mathieu Giraud committed
312 313 314 315 316
  ~KmerSegmenter();

  /**
   * @return the KmerAffectAnalyser of the current sequence.
   */
317
  KmerAffectAnalyser *getKmerAffectAnalyser() const;
Mathieu Giraud's avatar
Mathieu Giraud committed
318

319
  string getInfoLineWithAffects() const;
Marc Duez's avatar
Marc Duez committed
320
  json toJson();
321

Mathieu Giraud's avatar
Mathieu Giraud committed
322
 private:
323
  void computeSegmentation(int strand, KmerAffect left, KmerAffect right,
324
                           double threshold, double multiplier);
Mikaël Salson's avatar
Mikaël Salson committed
325 326
};

327 328 329

class KmerMultiSegmenter
{
330 331
 private:
  double threshold_nb_expected;
332 333 334
 public:
  /**
   * @param seq: An object read from a FASTA/FASTQ file
335 336 337 338
   * @param multigermline: the multigerm
   * @param threshold: threshold of randomly expected segmentation
   */
  KmerMultiSegmenter(Sequence seq, MultiGermline *multigermline, ostream *out_unsegmented,
339
                     double threshold = THRESHOLD_NB_EXPECTED, int nb_reads_for_evalue = 1);
340

341 342
  ~KmerMultiSegmenter();

343
  KmerSegmenter *the_kseg;
344
  MultiGermline *multi_germline;
345 346 347
};


Mikaël Salson's avatar
Mikaël Salson committed
348 349 350
class FineSegmenter : public Segmenter
{
 public:
Mikaël Salson's avatar
Mikaël Salson committed
351 352 353
   vector<pair<int, int> > score_V;
   vector<pair<int, int> > score_D;
   vector<pair<int, int> > score_J;
354 355

   vector <AlignBox*> boxes ;
Mikaël Salson's avatar
Mikaël Salson committed
356
   
Mikaël Salson's avatar
Mikaël Salson committed
357 358 359
   /**
   * Build a fineSegmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
360
   * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
361
   */
362
   FineSegmenter(Sequence seq, Germline *germline, Cost segment_cost,
363
                 double threshold = THRESHOLD_NB_EXPECTED, double multiplier=1.0);
364 365 366

   ~FineSegmenter();

Mikaël Salson's avatar
Mikaël Salson committed
367 368
  /**
  * extend segmentation from VJ to VDJ
369
  * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
370
  */
371
  void FineSegmentD(Germline *germline, bool several_D,
372
                    double threshold = THRESHOLD_NB_EXPECTED_D, double multiplier=1.0);
373 374

  bool FineSegmentD(Germline *germline,
375
                    AlignBox *box_Y, AlignBox *box_DD, AlignBox *box_Z,
376
                    int forbidden_id,
377
                    int extend_DD_on_Y, int extend_DD_on_Z,
378
                    double threshold = THRESHOLD_NB_EXPECTED_D, double multiplier=1.0);
379

Mathieu Giraud's avatar
Mathieu Giraud committed
380 381 382 383
  /**
   * find JUNCTION/CDR3, by using marked Cys104 and Phe118/Trp118 positions
   * in the germline V and J genes and the backtrack of the DP matrix
   */
Marc Duez's avatar
Marc Duez committed
384
  void findCDR3();
Mathieu Giraud's avatar
Mathieu Giraud committed
385

Marc Duez's avatar
Marc Duez committed
386
  json toJson();
Mikaël Salson's avatar
Mikaël Salson committed
387
  
Mikaël Salson's avatar
Mikaël Salson committed
388 389 390
};


391
void align_against_collection(string &read, BioReader &rep, int forbidden_rep_id,
392 393
                              bool reverse_ref, bool reverse_both, bool local,
                              AlignBox *box, Cost segment_cost);
Mikaël Salson's avatar
Mikaël Salson committed
394 395

#endif