segment.h 9.33 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 78 79 80 81 82
class AlignBox
{
 public:
  int del_left;
  int start;
  int end;
  int del_right;

83 84 85 86
  AlignBox();
  string getSequence(string sequence);

  /* Identifier, label and sequence of the reference sequence (the best one) */
87 88 89 90
  int ref_nb;
  string ref_label;
  string ref;

91 92
  /* Identifiers and scores of other possible reference sequence */
  vector<pair<int, int> > score;
93 94
};

95
ostream &operator<<(ostream &out, const AlignBox &box);
96

97
/**
98
 * Check whether there is an overlap between two boxes,
99 100 101
 * 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
102
 * @param *box_left, *box_right            the two boxes
103 104
 * @param segment_cost:                    the cost used by the dynamic programing
 *
105 106
 * @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
107 108 109 110 111
 *
 * @return                                 the N segment
 */

string check_and_resolve_overlap(string seq, int seq_begin, int seq_end,
112 113
                                 AlignBox *box_left, AlignBox *box_right,
                                 Cost segment_cost);
114

Mikaël Salson's avatar
Mikaël Salson committed
115 116 117
class Segmenter {
protected:
  string sequence;
Mathieu Giraud's avatar
Mathieu Giraud committed
118 119
  int Vend, Jstart;
  int Dstart, Dend;
Marc Duez's avatar
Marc Duez committed
120
  int CDR3start, CDR3end;
Mikaël Salson's avatar
Mikaël Salson committed
121
  bool reversed, segmented, dSegmented;
122
  int because;
Mikaël Salson's avatar
Mikaël Salson committed
123

124 125 126 127 128 129
  /**
   * 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
130 131 132 133 134
  string removeChevauchement();
  bool finishSegmentation();
  bool finishSegmentationD();

 public:
135
  Germline *segmented_germline;
136
  string label;
Mikaël Salson's avatar
Mikaël Salson committed
137
  string code;
138 139
  string info;        // .vdj.fa header, fixed fields
  string info_extra;  // .vdj.fa header, other information, at the end of the header
140
  string seg_V, seg_N, seg_J, system;
141

142 143
  AlignBox *box_V, *box_D, *box_J;

144
  double evalue;
145 146
  double evalue_left;
  double evalue_right;
Mikaël Salson's avatar
Mikaël Salson committed
147 148 149
  string seg_N1, seg_D, seg_N2;
  Cost segment_cost;

Mathieu Giraud's avatar
Mathieu Giraud committed
150
  virtual ~Segmenter();
Mikaël Salson's avatar
Mikaël Salson committed
151

Mathieu Giraud's avatar
Mathieu Giraud committed
152
  /* Queries */
Mikaël Salson's avatar
Mikaël Salson committed
153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203



  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;

204 205 206 207 208 209 210 211 212
  /**
   * @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
213

214 215 216 217 218
  /**
   * @post status == SEG_PLUS || status == SEG_MINUS <==> isSegmented()
   */
  void setSegmentationStatus(int status);

Mikaël Salson's avatar
Mikaël Salson committed
219 220 221 222 223 224 225 226 227 228 229
  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
230
 private:
231
  KmerAffectAnalyser *kaa;
Mikaël Salson's avatar
Mikaël Salson committed
232 233 234 235
 protected:
  string affects;

 public:
236
  int score;
237 238
  int pvalue_left;
  int pvalue_right;
239
  KmerAffect before, after;
240 241

  KmerSegmenter();
Mikaël Salson's avatar
Mikaël Salson committed
242 243 244
  /**
   * Build a segmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
245
   * @param germline: the germline
Mikaël Salson's avatar
Mikaël Salson committed
246
   */
247
  KmerSegmenter(Sequence seq, Germline *germline, double threshold = THRESHOLD_NB_EXPECTED, int multiplier=1);
Mathieu Giraud's avatar
Mathieu Giraud committed
248

249 250
  KmerSegmenter(const KmerSegmenter &seg);

Mathieu Giraud's avatar
Mathieu Giraud committed
251 252 253 254 255
  ~KmerSegmenter();

  /**
   * @return the KmerAffectAnalyser of the current sequence.
   */
256
  KmerAffectAnalyser *getKmerAffectAnalyser() const;
Mathieu Giraud's avatar
Mathieu Giraud committed
257

258
  string getInfoLineWithAffects() const;
Marc Duez's avatar
Marc Duez committed
259
  json toJson();
260

Mathieu Giraud's avatar
Mathieu Giraud committed
261
 private:
262 263
  void computeSegmentation(int strand, KmerAffect left, KmerAffect right,
                           double threshold, int multiplier);
Mikaël Salson's avatar
Mikaël Salson committed
264 265
};

266 267 268

class KmerMultiSegmenter
{
269 270
 private:
  double threshold_nb_expected;
271 272 273
 public:
  /**
   * @param seq: An object read from a FASTA/FASTQ file
274 275 276 277
   * @param multigermline: the multigerm
   * @param threshold: threshold of randomly expected segmentation
   */
  KmerMultiSegmenter(Sequence seq, MultiGermline *multigermline, ostream *out_unsegmented,
278
                     double threshold = THRESHOLD_NB_EXPECTED, int nb_reads_for_evalue = 1);
279

280 281
  ~KmerMultiSegmenter();

282
  KmerSegmenter *the_kseg;
283
  MultiGermline *multi_germline;
284 285 286
};


Mikaël Salson's avatar
Mikaël Salson committed
287 288 289
class FineSegmenter : public Segmenter
{
 public:
Mikaël Salson's avatar
Mikaël Salson committed
290 291 292 293
   vector<pair<int, int> > score_V;
   vector<pair<int, int> > score_D;
   vector<pair<int, int> > score_J;
   
Mikaël Salson's avatar
Mikaël Salson committed
294 295 296
   /**
   * Build a fineSegmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
297
   * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
298
   */
299 300
   FineSegmenter(Sequence seq, Germline *germline, Cost segment_cost,
                 double threshold = THRESHOLD_NB_EXPECTED, int multiplier=1);
Mikaël Salson's avatar
Mikaël Salson committed
301 302 303
  
  /**
  * extend segmentation from VJ to VDJ
304
  * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
305
  */
306 307
  void FineSegmentD(Germline *germline,
                    double threshold = THRESHOLD_NB_EXPECTED_D, int multiplier=1);
308 309

  bool FineSegmentD(Germline *germline,
310
                    AlignBox *box_Y, AlignBox *box_DD, AlignBox *box_Z,
311 312
                    double threshold = THRESHOLD_NB_EXPECTED_D, int multiplier=1);

Marc Duez's avatar
Marc Duez committed
313
  void findCDR3();
Mathieu Giraud's avatar
Mathieu Giraud committed
314

Marc Duez's avatar
Marc Duez committed
315
  json toJson();
Mikaël Salson's avatar
Mikaël Salson committed
316
  
Mikaël Salson's avatar
Mikaël Salson committed
317 318 319 320 321
};



#endif