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 10.5 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 7 8 9
#include <iostream>
#include "fasta.h"
#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

#define EXTEND_D_ZONE 5
17

18
#define RATIO_STRAND 2          /* The ratio between the affectations in one
Mikaël Salson's avatar
Mikaël Salson committed
19 20
                                   strand and the other, to safely attribute a
                                   segment to a given strand */
Mikaël Salson's avatar
Mikaël Salson committed
21

22 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
                                       stranf will be marked as STRAND_NOT_CONSISTEN */

26 27 28 29 30
#define DETECT_THRESHOLD 5      /* This threshold only affects the classification of unsegmented sequences.
                                   If the numbers of both V and J affectations are above this threshold,
                                   then the unsegmented sequence will be marked as AMBIGUOUS.
                                   Otherwise, if the number of either V or J affectations if above this threshold,
                                   then the unsegmented sequence will be marked as either ONLY V or ONLY J. */
31

Mathieu Giraud's avatar
Mathieu Giraud committed
32 33
#define JSON_REMEMBER_BEST  4   /* The number of V/D/J predictions to keep  */

34
#define BAD_EVALUE  1e10
35

36
#define THRESHOLD_NB_EXPECTED 1.0 /* Threshold of the accepted expected value for number of found k-mers */
37
#define THRESHOLD_NB_EXPECTED_D  .05 /* e-value threshold, D-REGION */
38

39 40 41 42
#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). */

43

44

Mikaël Salson's avatar
Mikaël Salson committed
45
using namespace std;
Marc Duez's avatar
Marc Duez committed
46
using json = nlohmann::json;
Mikaël Salson's avatar
Mikaël Salson committed
47

48 49 50 51
enum SEGMENTED { NOT_PROCESSED,
		 TOTAL_SEG_AND_WINDOW,
                 SEG_PLUS, SEG_MINUS,
                 UNSEG_TOO_SHORT, UNSEG_STRAND_NOT_CONSISTENT,
52
		 UNSEG_TOO_FEW_ZERO,  UNSEG_TOO_FEW_J, UNSEG_TOO_FEW_V,
53
                 UNSEG_BAD_DELTA_MIN, UNSEG_AMBIGUOUS,
54
		 UNSEG_TOO_SHORT_FOR_WINDOW,
55

Mikaël Salson's avatar
Mikaël Salson committed
56
		 STATS_SIZE } ;
57 58 59

#define STATS_FIRST_UNSEG UNSEG_TOO_SHORT

60 61 62 63
const char* const segmented_mesg[] = { "?",
                                       "SEG",
                                       "SEG_+", "SEG_-",
                                       "UNSEG too short", "UNSEG strand",
64
				       "UNSEG too few V/J", "UNSEG only V/5'", "UNSEG only J/3'",
65
				       "UNSEG < delta_min", "UNSEG ambiguous",
66
                                       "UNSEG too short w",
Mikaël Salson's avatar
Mikaël Salson committed
67
                                      } ;
Mikaël Salson's avatar
Mikaël Salson committed
68

69 70


71 72 73 74
/**
 * An alignment box (AlignBox) gather all parameters for a recombined gene segment (V, D, J, other D...)
 **/

75 76 77
class AlignBox
{
 public:
78
  string key;
79
  string color;
80

81 82 83 84 85
  int del_left;
  int start;
  int end;
  int del_right;

86
  AlignBox(string key = "", string color="");
87
  string getSequence(string sequence);
88
  void addToJson(json &seg);
89

90 91 92 93 94 95 96 97 98 99 100 101
  /**
   * 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);

102
  /* Identifier, label and sequence of the reference sequence (the best one) */
103 104 105 106
  int ref_nb;
  string ref_label;
  string ref;

107 108 109
  /* Marked position, for Cys104 and Phe118/Trp118 */
  int marked_pos;

110 111
  /* Identifiers and scores of other possible reference sequence */
  vector<pair<int, int> > score;
112 113
};

114
ostream &operator<<(ostream &out, const AlignBox &box);
115

116
/**
117
 * Check whether there is an overlap between two boxes,
118 119 120
 * 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
121
 * @param *box_left, *box_right            the two boxes
122 123
 * @param segment_cost:                    the cost used by the dynamic programing
 *
124 125
 * @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
126 127 128 129 130
 *
 * @return                                 the N segment
 */

string check_and_resolve_overlap(string seq, int seq_begin, int seq_end,
131 132
                                 AlignBox *box_left, AlignBox *box_right,
                                 Cost segment_cost);
133

Mikaël Salson's avatar
Mikaël Salson committed
134 135 136
class Segmenter {
protected:
  string sequence;
137
  string sequence_or_rc;
138

139
  // JUNCTIONstart/end and CDR3start/end are 1-based
140
  int JUNCTIONstart, JUNCTIONend;
141
  string JUNCTIONaa;
142
  bool JUNCTIONproductive;
143

Marc Duez's avatar
Marc Duez committed
144
  int CDR3start, CDR3end;
145 146
  string CDR3nuc;
  string CDR3aa;
147

Mikaël Salson's avatar
Mikaël Salson committed
148
  bool reversed, segmented, dSegmented;
149
  int because;
Mikaël Salson's avatar
Mikaël Salson committed
150

151 152 153 154 155 156
  /**
   * Compares evalue_left, evalue_right and evalue against the provided threshold
   * @post some evalue is above the threshold ==> because is set to UNSEG_TOO_FEW_ZERO, UNSEG_TOO_FEW_V or UNSEG_TOO_FEW_J
   */
  void checkLeftRightEvaluesThreshold(double threshold, int strand);

Mikaël Salson's avatar
Mikaël Salson committed
157 158 159 160 161
  string removeChevauchement();
  bool finishSegmentation();
  bool finishSegmentationD();

 public:
162
  Germline *segmented_germline;
163
  string label;
Mikaël Salson's avatar
Mikaël Salson committed
164
  string code;
165 166
  string info;        // .vdj.fa header, fixed fields
  string info_extra;  // .vdj.fa header, other information, at the end of the header
167
  string seg_V, seg_N, seg_J, system;
168

169 170
  AlignBox *box_V, *box_D, *box_J;

171
  double evalue;
172 173
  double evalue_left;
  double evalue_right;
Mikaël Salson's avatar
Mikaël Salson committed
174 175 176
  string seg_N1, seg_D, seg_N2;
  Cost segment_cost;

Mathieu Giraud's avatar
Mathieu Giraud committed
177
  virtual ~Segmenter();
Mikaël Salson's avatar
Mikaël Salson committed
178

Mathieu Giraud's avatar
Mathieu Giraud committed
179
  /* Queries */
Mikaël Salson's avatar
Mikaël Salson committed
180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230



  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.
   */
  string getJunction(int l) const;

  /**
   * @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;

231 232 233 234 235 236 237 238 239
  /**
   * @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
240

241 242 243 244 245
  /**
   * @post status == SEG_PLUS || status == SEG_MINUS <==> isSegmented()
   */
  void setSegmentationStatus(int status);

Mikaël Salson's avatar
Mikaël Salson committed
246 247 248 249 250 251 252 253 254 255 256
  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
257
 private:
258
  KmerAffectAnalyser *kaa;
Mikaël Salson's avatar
Mikaël Salson committed
259 260 261 262
 protected:
  string affects;

 public:
263
  int score;
264 265
  int pvalue_left;
  int pvalue_right;
266
  KmerAffect before, after;
267 268

  KmerSegmenter();
Mikaël Salson's avatar
Mikaël Salson committed
269 270 271
  /**
   * Build a segmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
272
   * @param germline: the germline
Mikaël Salson's avatar
Mikaël Salson committed
273
   */
274
  KmerSegmenter(Sequence seq, Germline *germline, double threshold = THRESHOLD_NB_EXPECTED, int multiplier=1);
Mathieu Giraud's avatar
Mathieu Giraud committed
275

276 277
  KmerSegmenter(const KmerSegmenter &seg);

Mathieu Giraud's avatar
Mathieu Giraud committed
278 279 280 281 282
  ~KmerSegmenter();

  /**
   * @return the KmerAffectAnalyser of the current sequence.
   */
283
  KmerAffectAnalyser *getKmerAffectAnalyser() const;
Mathieu Giraud's avatar
Mathieu Giraud committed
284

285
  string getInfoLineWithAffects() const;
Marc Duez's avatar
Marc Duez committed
286
  json toJson();
287

Mathieu Giraud's avatar
Mathieu Giraud committed
288
 private:
289 290
  void computeSegmentation(int strand, KmerAffect left, KmerAffect right,
                           double threshold, int multiplier);
Mikaël Salson's avatar
Mikaël Salson committed
291 292
};

293 294 295

class KmerMultiSegmenter
{
296 297
 private:
  double threshold_nb_expected;
298 299 300
 public:
  /**
   * @param seq: An object read from a FASTA/FASTQ file
301 302 303 304
   * @param multigermline: the multigerm
   * @param threshold: threshold of randomly expected segmentation
   */
  KmerMultiSegmenter(Sequence seq, MultiGermline *multigermline, ostream *out_unsegmented,
305
                     double threshold = THRESHOLD_NB_EXPECTED, int nb_reads_for_evalue = 1);
306

307 308
  ~KmerMultiSegmenter();

309
  KmerSegmenter *the_kseg;
310
  MultiGermline *multi_germline;
311 312 313
};


Mikaël Salson's avatar
Mikaël Salson committed
314 315 316
class FineSegmenter : public Segmenter
{
 public:
Mikaël Salson's avatar
Mikaël Salson committed
317 318 319
   vector<pair<int, int> > score_V;
   vector<pair<int, int> > score_D;
   vector<pair<int, int> > score_J;
320 321

   vector <AlignBox*> boxes ;
Mikaël Salson's avatar
Mikaël Salson committed
322
   
Mikaël Salson's avatar
Mikaël Salson committed
323 324 325
   /**
   * Build a fineSegmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
326
   * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
327
   */
328 329
   FineSegmenter(Sequence seq, Germline *germline, Cost segment_cost,
                 double threshold = THRESHOLD_NB_EXPECTED, int multiplier=1);
330 331 332

   ~FineSegmenter();

Mikaël Salson's avatar
Mikaël Salson committed
333 334
  /**
  * extend segmentation from VJ to VDJ
335
  * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
336
  */
337
  void FineSegmentD(Germline *germline, bool several_D,
338
                    double threshold = THRESHOLD_NB_EXPECTED_D, int multiplier=1);
339 340

  bool FineSegmentD(Germline *germline,
341
                    AlignBox *box_Y, AlignBox *box_DD, AlignBox *box_Z,
342
                    int forbidden_id,
343
                    int extend_DD_on_Y, int extend_DD_on_Z,
344 345
                    double threshold = THRESHOLD_NB_EXPECTED_D, int multiplier=1);

Mathieu Giraud's avatar
Mathieu Giraud committed
346 347 348 349
  /**
   * 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
350
  void findCDR3();
Mathieu Giraud's avatar
Mathieu Giraud committed
351

Marc Duez's avatar
Marc Duez committed
352
  json toJson();
Mikaël Salson's avatar
Mikaël Salson committed
353
  
Mikaël Salson's avatar
Mikaël Salson committed
354 355 356
};


357 358 359
void align_against_collection(string &read, Fasta &rep, int forbidden_rep_id,
                              bool reverse_ref, bool reverse_both, bool local,
                              AlignBox *box, Cost segment_cost);
Mikaël Salson's avatar
Mikaël Salson committed
360 361

#endif