segment.h 10.7 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 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 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 29 30 31 32
#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. */
33

Mathieu Giraud's avatar
Mathieu Giraud committed
34 35
#define JSON_REMEMBER_BEST  4   /* The number of V/D/J predictions to keep  */

36
#define BAD_EVALUE  1e10
37

38
#define THRESHOLD_NB_EXPECTED 1.0 /* Threshold of the accepted expected value for number of found k-mers */
39
#define THRESHOLD_NB_EXPECTED_D  .05 /* e-value threshold, D-REGION */
40

41 42 43 44
#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). */

45

46

Mikaël Salson's avatar
Mikaël Salson committed
47
using namespace std;
Marc Duez's avatar
Marc Duez committed
48
using json = nlohmann::json;
Mikaël Salson's avatar
Mikaël Salson committed
49

50 51 52 53
enum SEGMENTED { NOT_PROCESSED,
		 TOTAL_SEG_AND_WINDOW,
                 SEG_PLUS, SEG_MINUS,
                 UNSEG_TOO_SHORT, UNSEG_STRAND_NOT_CONSISTENT,
54
		 UNSEG_TOO_FEW_ZERO,  UNSEG_TOO_FEW_J, UNSEG_TOO_FEW_V,
55
                 UNSEG_BAD_DELTA_MIN, UNSEG_AMBIGUOUS,
56
		 UNSEG_TOO_SHORT_FOR_WINDOW,
57

Mikaël Salson's avatar
Mikaël Salson committed
58
		 STATS_SIZE } ;
59 60 61

#define STATS_FIRST_UNSEG UNSEG_TOO_SHORT

62 63 64 65
const char* const segmented_mesg[] = { "?",
                                       "SEG",
                                       "SEG_+", "SEG_-",
                                       "UNSEG too short", "UNSEG strand",
66
				       "UNSEG too few V/J", "UNSEG only V/5'", "UNSEG only J/3'",
67
				       "UNSEG < delta_min", "UNSEG ambiguous",
68
                                       "UNSEG too short w",
Mikaël Salson's avatar
Mikaël Salson committed
69
                                      } ;
Mikaël Salson's avatar
Mikaël Salson committed
70

71 72


73 74 75 76
/**
 * An alignment box (AlignBox) gather all parameters for a recombined gene segment (V, D, J, other D...)
 **/

77 78 79
class AlignBox
{
 public:
80
  string key;
81
  string color;
82

83 84 85 86 87
  int del_left;
  int start;
  int end;
  int del_right;

88
  AlignBox(string key = "", string color="");
89
  string getSequence(string sequence);
90
  void addToJson(json &seg);
91

92 93 94 95 96
  /**
   * Returns 'V', 'D', 'J', or possibly '5', '4', '3', '?', depending on the ref_label and on the key
   */
  char getInitial();

97 98 99 100 101 102 103 104 105 106 107 108
  /**
   * 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);

109
  /* Identifier, label and sequence of the reference sequence (the best one) */
110 111 112 113
  int ref_nb;
  string ref_label;
  string ref;

114 115 116
  /* Marked position, for Cys104 and Phe118/Trp118 */
  int marked_pos;

117 118
  /* Identifiers and scores of other possible reference sequence */
  vector<pair<int, int> > score;
119 120
};

121
ostream &operator<<(ostream &out, const AlignBox &box);
122

123
/**
124
 * Check whether there is an overlap between two boxes,
125 126 127
 * 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
128
 * @param *box_left, *box_right            the two boxes
129 130
 * @param segment_cost:                    the cost used by the dynamic programing
 *
131 132
 * @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
133 134 135 136 137
 *
 * @return                                 the N segment
 */

string check_and_resolve_overlap(string seq, int seq_begin, int seq_end,
138 139
                                 AlignBox *box_left, AlignBox *box_right,
                                 Cost segment_cost);
140

Mikaël Salson's avatar
Mikaël Salson committed
141 142 143
class Segmenter {
protected:
  string sequence;
144
  string sequence_or_rc;
145

146
  // JUNCTIONstart/end and CDR3start/end are 1-based
147
  int JUNCTIONstart, JUNCTIONend;
148
  string JUNCTIONaa;
149
  bool JUNCTIONproductive;
150

Marc Duez's avatar
Marc Duez committed
151
  int CDR3start, CDR3end;
152 153
  string CDR3nuc;
  string CDR3aa;
154

Mikaël Salson's avatar
Mikaël Salson committed
155
  bool reversed, segmented, dSegmented;
156
  int because;
Mikaël Salson's avatar
Mikaël Salson committed
157

158 159 160 161 162 163
  /**
   * 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
164 165 166 167 168
  string removeChevauchement();
  bool finishSegmentation();
  bool finishSegmentationD();

 public:
169
  Germline *segmented_germline;
170
  string label;
Mikaël Salson's avatar
Mikaël Salson committed
171
  string code;
172 173
  string info;        // .vdj.fa header, fixed fields
  string info_extra;  // .vdj.fa header, other information, at the end of the header
174
  string seg_V, seg_N, seg_J, system;
175

176 177
  AlignBox *box_V, *box_D, *box_J;

178
  double evalue;
179 180
  double evalue_left;
  double evalue_right;
Mikaël Salson's avatar
Mikaël Salson committed
181 182 183
  string seg_N1, seg_D, seg_N2;
  Cost segment_cost;

Mathieu Giraud's avatar
Mathieu Giraud committed
184
  virtual ~Segmenter();
Mikaël Salson's avatar
Mikaël Salson committed
185

Mathieu Giraud's avatar
Mathieu Giraud committed
186
  /* Queries */
Mikaël Salson's avatar
Mikaël Salson committed
187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237



  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;

238 239 240 241 242 243 244 245 246
  /**
   * @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
247

248 249 250 251 252
  /**
   * @post status == SEG_PLUS || status == SEG_MINUS <==> isSegmented()
   */
  void setSegmentationStatus(int status);

Mikaël Salson's avatar
Mikaël Salson committed
253 254 255 256 257 258 259 260 261 262 263
  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
264
 private:
265
  KmerAffectAnalyser *kaa;
Mikaël Salson's avatar
Mikaël Salson committed
266 267 268 269
 protected:
  string affects;

 public:
270
  int score;
271 272
  int pvalue_left;
  int pvalue_right;
273
  KmerAffect before, after;
274 275

  KmerSegmenter();
Mikaël Salson's avatar
Mikaël Salson committed
276 277 278
  /**
   * Build a segmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
279
   * @param germline: the germline
Mikaël Salson's avatar
Mikaël Salson committed
280
   */
281
  KmerSegmenter(Sequence seq, Germline *germline, double threshold = THRESHOLD_NB_EXPECTED, double multiplier=1.0);
Mathieu Giraud's avatar
Mathieu Giraud committed
282

283 284
  KmerSegmenter(const KmerSegmenter &seg);

Mathieu Giraud's avatar
Mathieu Giraud committed
285 286 287 288 289
  ~KmerSegmenter();

  /**
   * @return the KmerAffectAnalyser of the current sequence.
   */
290
  KmerAffectAnalyser *getKmerAffectAnalyser() const;
Mathieu Giraud's avatar
Mathieu Giraud committed
291

292
  string getInfoLineWithAffects() const;
Marc Duez's avatar
Marc Duez committed
293
  json toJson();
294

Mathieu Giraud's avatar
Mathieu Giraud committed
295
 private:
296
  void computeSegmentation(int strand, KmerAffect left, KmerAffect right,
297
                           double threshold, double multiplier);
Mikaël Salson's avatar
Mikaël Salson committed
298 299
};

300 301 302

class KmerMultiSegmenter
{
303 304
 private:
  double threshold_nb_expected;
305 306 307
 public:
  /**
   * @param seq: An object read from a FASTA/FASTQ file
308 309 310 311
   * @param multigermline: the multigerm
   * @param threshold: threshold of randomly expected segmentation
   */
  KmerMultiSegmenter(Sequence seq, MultiGermline *multigermline, ostream *out_unsegmented,
312
                     double threshold = THRESHOLD_NB_EXPECTED, int nb_reads_for_evalue = 1);
313

314 315
  ~KmerMultiSegmenter();

316
  KmerSegmenter *the_kseg;
317
  MultiGermline *multi_germline;
318 319 320
};


Mikaël Salson's avatar
Mikaël Salson committed
321 322 323
class FineSegmenter : public Segmenter
{
 public:
Mikaël Salson's avatar
Mikaël Salson committed
324 325 326
   vector<pair<int, int> > score_V;
   vector<pair<int, int> > score_D;
   vector<pair<int, int> > score_J;
327 328

   vector <AlignBox*> boxes ;
Mikaël Salson's avatar
Mikaël Salson committed
329
   
Mikaël Salson's avatar
Mikaël Salson committed
330 331 332
   /**
   * Build a fineSegmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
333
   * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
334
   */
335
   FineSegmenter(Sequence seq, Germline *germline, Cost segment_cost,
336
                 double threshold = THRESHOLD_NB_EXPECTED, double multiplier=1.0);
337 338 339

   ~FineSegmenter();

Mikaël Salson's avatar
Mikaël Salson committed
340 341
  /**
  * extend segmentation from VJ to VDJ
342
  * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
343
  */
344
  void FineSegmentD(Germline *germline, bool several_D,
345
                    double threshold = THRESHOLD_NB_EXPECTED_D, double multiplier=1.0);
346 347

  bool FineSegmentD(Germline *germline,
348
                    AlignBox *box_Y, AlignBox *box_DD, AlignBox *box_Z,
349
                    int forbidden_id,
350
                    int extend_DD_on_Y, int extend_DD_on_Z,
351
                    double threshold = THRESHOLD_NB_EXPECTED_D, double multiplier=1.0);
352

Mathieu Giraud's avatar
Mathieu Giraud committed
353 354 355 356
  /**
   * 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
357
  void findCDR3();
Mathieu Giraud's avatar
Mathieu Giraud committed
358

Marc Duez's avatar
Marc Duez committed
359
  json toJson();
Mikaël Salson's avatar
Mikaël Salson committed
360
  
Mikaël Salson's avatar
Mikaël Salson committed
361 362 363
};


364 365 366
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
367 368

#endif