segment.h 7.94 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 19

#define MIN_D_LENGTH 5          /* If a D-REGION is smaller than this threshold, it is not output */

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 26 27
#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 */

28
#define DETECT_THRESHOLD 5      /* If the number of both V and J affectations
29 30
                                   is above this threshold, then the sequence,
                                   if it not segmented, will be marked as AMBIGUOUS */
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 NO_LIMIT_VALUE  -1
35
#define BAD_EVALUE  1e10
36

37
#define THRESHOLD_NB_EXPECTED 1.0 /* Threshold of the accepted expected value for number of found k-mers */
38 39


40

Mikaël Salson's avatar
Mikaël Salson committed
41
using namespace std;
Marc Duez's avatar
Marc Duez committed
42
using json = nlohmann::json;
Mikaël Salson's avatar
Mikaël Salson committed
43

44 45 46 47
enum SEGMENTED { NOT_PROCESSED,
		 TOTAL_SEG_AND_WINDOW,
                 SEG_PLUS, SEG_MINUS,
                 UNSEG_TOO_SHORT, UNSEG_STRAND_NOT_CONSISTENT,
48
		 UNSEG_TOO_FEW_ZERO,  UNSEG_TOO_FEW_J, UNSEG_TOO_FEW_V,
49
                 UNSEG_BAD_DELTA_MIN, UNSEG_AMBIGUOUS,
50
		 UNSEG_TOO_SHORT_FOR_WINDOW,
51

Mikaël Salson's avatar
Mikaël Salson committed
52
		 STATS_SIZE } ;
53 54 55 56
const char* const segmented_mesg[] = { "?",
                                       "SEG",
                                       "SEG_+", "SEG_-",
                                       "UNSEG too short", "UNSEG strand",
57
				       "UNSEG too few V/J", "UNSEG only V", "UNSEG only J",
58
				       "UNSEG < delta_min", "UNSEG ambiguous",
59
                                       "UNSEG too short w",
Mikaël Salson's avatar
Mikaël Salson committed
60
                                      } ;
Mikaël Salson's avatar
Mikaël Salson committed
61

62 63 64 65 66 67 68 69



 /**
  * Find the best split point when there may be a overlap between a left and a right region
  * @param overlap: the length of the overlap
  * @param seq_left, seq_right: the two sequences that overlap
  * @param ref_left, ref_right: the two reference sequences
70
  * @param pos_seq_left, pos_seq_right: the initial end/start position of two overlaping sequences
71
  * @param segment_cost: dp cost
72 73 74 75
  *
  * @post  trim_ref_left (at the end of ref_left) and trim_ref_right (at the beginning of ref_right)
  *        are set to the best number of nucleotides to trim in order to remove the overlap
  *        pos_seq_left and pos_seq_right are shifted by the good number of nucleotides (relatively to seq_{left,right})
76 77 78 79
  */

void best_overlap_split(int overlap, string seq_left, string seq_right,
                        string ref_left, string ref_right,
80 81
                        int *pos_seq_left, int *pos_seq_right,
                        int *trim_ref_left, int *trim_ref_right, Cost segment_cost);
82

Mikaël Salson's avatar
Mikaël Salson committed
83 84 85
class Segmenter {
protected:
  string sequence;
Mathieu Giraud's avatar
Mathieu Giraud committed
86 87
  int Vend, Jstart;
  int Dstart, Dend;
Marc Duez's avatar
Marc Duez committed
88
  int CDR3start, CDR3end;
Mikaël Salson's avatar
Mikaël Salson committed
89
  bool reversed, segmented, dSegmented;
90
  int because;
Mikaël Salson's avatar
Mikaël Salson committed
91

92 93 94 95 96 97
  /**
   * 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
98 99 100 101 102
  string removeChevauchement();
  bool finishSegmentation();
  bool finishSegmentationD();

 public:
103
  Germline *segmented_germline;
104
  string label;
Mikaël Salson's avatar
Mikaël Salson committed
105
  string code;
Mikaël Salson's avatar
Mikaël Salson committed
106
  string code_short;
Mikaël Salson's avatar
Mikaël Salson committed
107
  string code_light;
108 109
  string info;        // .vdj.fa header, fixed fields
  string info_extra;  // .vdj.fa header, other information, at the end of the header
Mikaël Salson's avatar
Mikaël Salson committed
110 111
  int best_V, best_J ;
  int del_V, del_D_left, del_D_right, del_J ;
112
  string seg_V, seg_N, seg_J, system;
113

Mikaël Salson's avatar
Mikaël Salson committed
114
  int best_D;
115
  double evalue;
116 117
  double evalue_left;
  double evalue_right;
Mikaël Salson's avatar
Mikaël Salson committed
118 119 120
  string seg_N1, seg_D, seg_N2;
  Cost segment_cost;

Mathieu Giraud's avatar
Mathieu Giraud committed
121
  virtual ~Segmenter();
Mikaël Salson's avatar
Mikaël Salson committed
122

Mathieu Giraud's avatar
Mathieu Giraud committed
123
  /* Queries */
Mikaël Salson's avatar
Mikaël Salson committed
124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174



  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;

175 176 177 178 179 180 181 182 183
  /**
   * @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
184

185 186 187 188 189
  /**
   * @post status == SEG_PLUS || status == SEG_MINUS <==> isSegmented()
   */
  void setSegmentationStatus(int status);

Mikaël Salson's avatar
Mikaël Salson committed
190 191 192 193 194 195 196 197 198 199 200
  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
201
 private:
202
  KmerAffectAnalyser *kaa;
Mikaël Salson's avatar
Mikaël Salson committed
203 204 205 206
 protected:
  string affects;

 public:
207
  int score;
208 209
  int pvalue_left;
  int pvalue_right;
210
  KmerAffect before, after;
211 212

  KmerSegmenter();
Mikaël Salson's avatar
Mikaël Salson committed
213 214 215
  /**
   * Build a segmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
216
   * @param germline: the germline
Mikaël Salson's avatar
Mikaël Salson committed
217
   */
218
  KmerSegmenter(Sequence seq, Germline *germline, double threshold = THRESHOLD_NB_EXPECTED, int multiplier=1);
Mathieu Giraud's avatar
Mathieu Giraud committed
219

220 221
  KmerSegmenter(const KmerSegmenter &seg);

Mathieu Giraud's avatar
Mathieu Giraud committed
222 223 224 225 226
  ~KmerSegmenter();

  /**
   * @return the KmerAffectAnalyser of the current sequence.
   */
227
  KmerAffectAnalyser *getKmerAffectAnalyser() const;
Mathieu Giraud's avatar
Mathieu Giraud committed
228

229
  string getInfoLineWithAffects() const;
Marc Duez's avatar
Marc Duez committed
230
  json toJson();
231

Mathieu Giraud's avatar
Mathieu Giraud committed
232
 private:
233 234
  void computeSegmentation(int strand, KmerAffect left, KmerAffect right,
                           double threshold, int multiplier);
Mikaël Salson's avatar
Mikaël Salson committed
235 236
};

237 238 239

class KmerMultiSegmenter
{
240 241
 private:
  double threshold_nb_expected;
242 243 244
 public:
  /**
   * @param seq: An object read from a FASTA/FASTQ file
245 246 247 248
   * @param multigermline: the multigerm
   * @param threshold: threshold of randomly expected segmentation
   */
  KmerMultiSegmenter(Sequence seq, MultiGermline *multigermline, ostream *out_unsegmented,
249
                     double threshold = THRESHOLD_NB_EXPECTED, int nb_reads_for_evalue = 1);
250

251 252
  ~KmerMultiSegmenter();

253
  KmerSegmenter *the_kseg;
254
  MultiGermline *multi_germline;
255 256 257
};


Mikaël Salson's avatar
Mikaël Salson committed
258 259 260
class FineSegmenter : public Segmenter
{
 public:
Mikaël Salson's avatar
Mikaël Salson committed
261 262 263 264
   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
265 266 267
   /**
   * Build a fineSegmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
268
   * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
269
   */
270 271
   FineSegmenter(Sequence seq, Germline *germline, Cost segment_cost,
                 double threshold = THRESHOLD_NB_EXPECTED, int multiplier=1);
Mikaël Salson's avatar
Mikaël Salson committed
272 273 274
  
  /**
  * extend segmentation from VJ to VDJ
275
  * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
276
  */
277
  void FineSegmentD(Germline *germline);
Marc Duez's avatar
Marc Duez committed
278
  void findCDR3();
Mathieu Giraud's avatar
Mathieu Giraud committed
279

Marc Duez's avatar
Marc Duez committed
280
  json toJson();
Mikaël Salson's avatar
Mikaël Salson committed
281
  
Mikaël Salson's avatar
Mikaël Salson committed
282 283 284 285 286
};



#endif