segment.h 8.67 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 60
const char* const segmented_mesg[] = { "?",
                                       "SEG",
                                       "SEG_+", "SEG_-",
                                       "UNSEG too short", "UNSEG strand",
61
				       "UNSEG too few V/J", "UNSEG only V", "UNSEG only J",
62
				       "UNSEG < delta_min", "UNSEG ambiguous",
63
                                       "UNSEG too short w",
Mikaël Salson's avatar
Mikaël Salson committed
64
                                      } ;
Mikaël Salson's avatar
Mikaël Salson committed
65

66 67 68 69 70 71 72 73



 /**
  * 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
74
  * @param pos_seq_left, pos_seq_right: the initial end/start position of two overlaping sequences
75
  * @param segment_cost: dp cost
76 77 78 79
  *
  * @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})
80 81 82 83
  */

void best_overlap_split(int overlap, string seq_left, string seq_right,
                        string ref_left, string ref_right,
84 85
                        int *pos_seq_left, int *pos_seq_right,
                        int *trim_ref_left, int *trim_ref_right, Cost segment_cost);
86

Mikaël Salson's avatar
Mikaël Salson committed
87 88 89
class Segmenter {
protected:
  string sequence;
Mathieu Giraud's avatar
Mathieu Giraud committed
90 91
  int Vend, Jstart;
  int Dstart, Dend;
Marc Duez's avatar
Marc Duez committed
92
  int CDR3start, CDR3end;
Mikaël Salson's avatar
Mikaël Salson committed
93
  bool reversed, segmented, dSegmented;
94
  int because;
Mikaël Salson's avatar
Mikaël Salson committed
95

96 97 98 99 100 101
  /**
   * 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
102 103 104 105 106
  string removeChevauchement();
  bool finishSegmentation();
  bool finishSegmentationD();

 public:
107
  Germline *segmented_germline;
108
  string label;
Mikaël Salson's avatar
Mikaël Salson committed
109
  string code;
Mikaël Salson's avatar
Mikaël Salson committed
110
  string code_short;
Mikaël Salson's avatar
Mikaël Salson committed
111
  string code_light;
112 113
  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
114 115
  int best_V, best_J ;
  int del_V, del_D_left, del_D_right, del_J ;
116
  string seg_V, seg_N, seg_J, system;
117

Mikaël Salson's avatar
Mikaël Salson committed
118
  int best_D;
119
  double evalue;
120 121
  double evalue_left;
  double evalue_right;
Mikaël Salson's avatar
Mikaël Salson committed
122 123 124
  string seg_N1, seg_D, seg_N2;
  Cost segment_cost;

Mathieu Giraud's avatar
Mathieu Giraud committed
125
  virtual ~Segmenter();
Mikaël Salson's avatar
Mikaël Salson committed
126

Mathieu Giraud's avatar
Mathieu Giraud committed
127
  /* Queries */
Mikaël Salson's avatar
Mikaël Salson committed
128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178



  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;

179 180 181 182 183 184 185 186 187
  /**
   * @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
188

189 190 191 192 193
  /**
   * @post status == SEG_PLUS || status == SEG_MINUS <==> isSegmented()
   */
  void setSegmentationStatus(int status);

Mikaël Salson's avatar
Mikaël Salson committed
194 195 196 197 198 199 200 201 202 203 204
  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
205
 private:
206
  KmerAffectAnalyser *kaa;
Mikaël Salson's avatar
Mikaël Salson committed
207 208 209 210
 protected:
  string affects;

 public:
211
  int score;
212 213
  int pvalue_left;
  int pvalue_right;
214
  KmerAffect before, after;
215 216

  KmerSegmenter();
Mikaël Salson's avatar
Mikaël Salson committed
217 218 219
  /**
   * Build a segmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
220
   * @param germline: the germline
Mikaël Salson's avatar
Mikaël Salson committed
221
   */
222
  KmerSegmenter(Sequence seq, Germline *germline, double threshold = THRESHOLD_NB_EXPECTED, int multiplier=1);
Mathieu Giraud's avatar
Mathieu Giraud committed
223

224 225
  KmerSegmenter(const KmerSegmenter &seg);

Mathieu Giraud's avatar
Mathieu Giraud committed
226 227 228 229 230
  ~KmerSegmenter();

  /**
   * @return the KmerAffectAnalyser of the current sequence.
   */
231
  KmerAffectAnalyser *getKmerAffectAnalyser() const;
Mathieu Giraud's avatar
Mathieu Giraud committed
232

233
  string getInfoLineWithAffects() const;
Marc Duez's avatar
Marc Duez committed
234
  json toJson();
235

Mathieu Giraud's avatar
Mathieu Giraud committed
236
 private:
237 238
  void computeSegmentation(int strand, KmerAffect left, KmerAffect right,
                           double threshold, int multiplier);
Mikaël Salson's avatar
Mikaël Salson committed
239 240
};

241 242 243

class KmerMultiSegmenter
{
244 245
 private:
  double threshold_nb_expected;
246 247 248
 public:
  /**
   * @param seq: An object read from a FASTA/FASTQ file
249 250 251 252
   * @param multigermline: the multigerm
   * @param threshold: threshold of randomly expected segmentation
   */
  KmerMultiSegmenter(Sequence seq, MultiGermline *multigermline, ostream *out_unsegmented,
253
                     double threshold = THRESHOLD_NB_EXPECTED, int nb_reads_for_evalue = 1);
254

255 256
  ~KmerMultiSegmenter();

257
  KmerSegmenter *the_kseg;
258
  MultiGermline *multi_germline;
259 260 261
};


Mikaël Salson's avatar
Mikaël Salson committed
262 263 264
class FineSegmenter : public Segmenter
{
 public:
Mikaël Salson's avatar
Mikaël Salson committed
265 266 267 268
   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
269 270 271
   /**
   * Build a fineSegmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
272
   * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
273
   */
274 275
   FineSegmenter(Sequence seq, Germline *germline, Cost segment_cost,
                 double threshold = THRESHOLD_NB_EXPECTED, int multiplier=1);
Mikaël Salson's avatar
Mikaël Salson committed
276 277 278
  
  /**
  * extend segmentation from VJ to VDJ
279
  * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
280
  */
281 282
  void FineSegmentD(Germline *germline,
                    double threshold = THRESHOLD_NB_EXPECTED_D, int multiplier=1);
Marc Duez's avatar
Marc Duez committed
283
  void findCDR3();
Mathieu Giraud's avatar
Mathieu Giraud committed
284

Marc Duez's avatar
Marc Duez committed
285
  json toJson();
Mikaël Salson's avatar
Mikaël Salson committed
286
  
Mikaël Salson's avatar
Mikaël Salson committed
287 288 289 290 291
};



#endif