segment.h 10.9 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
#include <iostream>
7
#include "bioreader.hpp"
Mikaël Salson's avatar
Mikaël Salson committed
8 9
#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
#define DETECT_THRESHOLD_STRAND 5   /* If the number of total affectations
                                       is above this threshold, then a sequence with no clearly attributed
26
                                       strand will be marked as STRAND_NOT_CONSISTENT */
27

Mathieu Giraud's avatar
Mathieu Giraud committed
28 29
#define JSON_REMEMBER_BEST  4   /* The number of V/D/J predictions to keep  */

30
#define BAD_EVALUE  1e10
31

32
#define THRESHOLD_NB_EXPECTED 1.0 /* Threshold of the accepted expected value for number of found k-mers */
33
#define THRESHOLD_NB_EXPECTED_D  .05 /* e-value threshold, D-REGION */
34

35 36 37 38
#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). */

39 40 41
#define DEFAULT_WINDOW_SHIFT 5  /* Number of nucleotide to try shifting or
                                   reducing the window when it doesn't fit at
                                   its place */
42
#define MINIMAL_WINDOW_LENGTH 30 /* As we now dynamically adapt the window
43 44
                                    length we need to specify a minimum
                                    otherwise we could go as low as
45 46 47
                                    2*DEFAULT_WINDOW_SHIFT. Of course this can
                                    be overriden by the command line by
                                    providing a shorter -w*/
48

49

Mikaël Salson's avatar
Mikaël Salson committed
50
using namespace std;
Marc Duez's avatar
Marc Duez committed
51
using json = nlohmann::json;
Mikaël Salson's avatar
Mikaël Salson committed
52

53 54 55 56
enum SEGMENTED { NOT_PROCESSED,
		 TOTAL_SEG_AND_WINDOW,
                 SEG_PLUS, SEG_MINUS,
                 UNSEG_TOO_SHORT, UNSEG_STRAND_NOT_CONSISTENT,
57
		 UNSEG_TOO_FEW_ZERO,  UNSEG_ONLY_V, UNSEG_ONLY_J,
58
                 UNSEG_BAD_DELTA_MIN, UNSEG_AMBIGUOUS,
59
		 UNSEG_TOO_SHORT_FOR_WINDOW,
60

Mikaël Salson's avatar
Mikaël Salson committed
61
		 STATS_SIZE } ;
62 63 64

#define STATS_FIRST_UNSEG UNSEG_TOO_SHORT

65 66 67 68
const char* const segmented_mesg[] = { "?",
                                       "SEG",
                                       "SEG_+", "SEG_-",
                                       "UNSEG too short", "UNSEG strand",
69
				       "UNSEG too few V/J", "UNSEG only V/5'", "UNSEG only J/3'",
70
				       "UNSEG < delta_min", "UNSEG ambiguous",
71
                                       "UNSEG too short w",
Mikaël Salson's avatar
Mikaël Salson committed
72
                                      } ;
Mikaël Salson's avatar
Mikaël Salson committed
73

74 75


76 77 78 79
/**
 * An alignment box (AlignBox) gather all parameters for a recombined gene segment (V, D, J, other D...)
 **/

80 81 82
class AlignBox
{
 public:
83
  string key;
84
  string color;
85

86 87 88 89 90
  int del_left;
  int start;
  int end;
  int del_right;

91
  AlignBox(string key = "", string color="");
92
  string getSequence(string sequence);
93
  void addToJson(json &seg);
94

95 96 97 98 99
  /**
   * Returns 'V', 'D', 'J', or possibly '5', '4', '3', '?', depending on the ref_label and on the key
   */
  char getInitial();

100 101 102 103 104
  /**
   * Returns the length
   */
  int getLength();

105 106 107 108 109 110 111 112 113 114 115 116
  /**
   * 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);

117
  /* Identifier, label and sequence of the reference sequence (the best one) */
118 119 120 121
  int ref_nb;
  string ref_label;
  string ref;

122 123 124
  /* Marked position, for Cys104 and Phe118/Trp118 */
  int marked_pos;

125 126
  /* Identifiers and scores of other possible reference sequence */
  vector<pair<int, int> > score;
127 128
};

129
ostream &operator<<(ostream &out, const AlignBox &box);
130

131
/**
132
 * Check whether there is an overlap between two boxes,
133 134 135
 * 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
136
 * @param *box_left, *box_right            the two boxes
137 138
 * @param segment_cost:                    the cost used by the dynamic programing
 *
139 140
 * @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
141 142 143 144 145
 *
 * @return                                 the N segment
 */

string check_and_resolve_overlap(string seq, int seq_begin, int seq_end,
146 147
                                 AlignBox *box_left, AlignBox *box_right,
                                 Cost segment_cost);
148

Mikaël Salson's avatar
Mikaël Salson committed
149 150 151
class Segmenter {
protected:
  string sequence;
152
  string sequence_or_rc;
153

154
  // JUNCTIONstart/end and CDR3start/end are 1-based
155
  int JUNCTIONstart, JUNCTIONend;
156
  string JUNCTIONaa;
157
  bool JUNCTIONproductive;
158

Marc Duez's avatar
Marc Duez committed
159
  int CDR3start, CDR3end;
160 161
  string CDR3nuc;
  string CDR3aa;
162

Mikaël Salson's avatar
Mikaël Salson committed
163
  bool reversed, segmented, dSegmented;
164
  int because;
Mikaël Salson's avatar
Mikaël Salson committed
165

166 167
  /**
   * Compares evalue_left, evalue_right and evalue against the provided threshold
168
   * @post some evalue is above the threshold ==> because is set to UNSEG_TOO_FEW_ZERO, UNSEG_ONLY_J or UNSEG_ONLY_V
169 170 171
   */
  void checkLeftRightEvaluesThreshold(double threshold, int strand);

Mikaël Salson's avatar
Mikaël Salson committed
172 173 174 175 176
  string removeChevauchement();
  bool finishSegmentation();
  bool finishSegmentationD();

 public:
177
  Germline *segmented_germline;
178
  string label;
Mikaël Salson's avatar
Mikaël Salson committed
179
  string code;
180 181
  string info;        // .vdj.fa header, fixed fields
  string info_extra;  // .vdj.fa header, other information, at the end of the header
182
  string seg_V, seg_N, seg_J, system;
183

184 185
  AlignBox *box_V, *box_D, *box_J;

186
  double evalue;
187 188
  double evalue_left;
  double evalue_right;
Mikaël Salson's avatar
Mikaël Salson committed
189 190 191
  string seg_N1, seg_D, seg_N2;
  Cost segment_cost;

Mathieu Giraud's avatar
Mathieu Giraud committed
192
  virtual ~Segmenter();
Mikaël Salson's avatar
Mikaël Salson committed
193

Mathieu Giraud's avatar
Mathieu Giraud committed
194
  /* Queries */
Mikaël Salson's avatar
Mikaël Salson committed
195 196 197 198 199 200 201 202 203 204 205 206 207 208



  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.
   */
209
  string getJunction(int l, int shift=0) const;
Mikaël Salson's avatar
Mikaël Salson committed
210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245

  /**
   * @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;

246 247 248 249 250 251 252 253 254
  /**
   * @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
255

256 257 258 259 260
  /**
   * @post status == SEG_PLUS || status == SEG_MINUS <==> isSegmented()
   */
  void setSegmentationStatus(int status);

Mikaël Salson's avatar
Mikaël Salson committed
261 262 263 264 265 266 267 268 269 270 271
  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
272
 private:
273
  KmerAffectAnalyser *kaa;
Mikaël Salson's avatar
Mikaël Salson committed
274 275 276 277
 protected:
  string affects;

 public:
278
  int score;
279 280
  int pvalue_left;
  int pvalue_right;
281
  KmerAffect before, after;
282 283

  KmerSegmenter();
Mikaël Salson's avatar
Mikaël Salson committed
284 285 286
  /**
   * Build a segmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
287
   * @param germline: the germline
Mikaël Salson's avatar
Mikaël Salson committed
288
   */
289
  KmerSegmenter(Sequence seq, Germline *germline, double threshold = THRESHOLD_NB_EXPECTED, double multiplier=1.0);
Mathieu Giraud's avatar
Mathieu Giraud committed
290

291 292
  KmerSegmenter(const KmerSegmenter &seg);

Mathieu Giraud's avatar
Mathieu Giraud committed
293 294 295 296 297
  ~KmerSegmenter();

  /**
   * @return the KmerAffectAnalyser of the current sequence.
   */
298
  KmerAffectAnalyser *getKmerAffectAnalyser() const;
Mathieu Giraud's avatar
Mathieu Giraud committed
299

300
  string getInfoLineWithAffects() const;
Marc Duez's avatar
Marc Duez committed
301
  json toJson();
302

Mathieu Giraud's avatar
Mathieu Giraud committed
303
 private:
304
  void computeSegmentation(int strand, KmerAffect left, KmerAffect right,
305
                           double threshold, double multiplier);
Mikaël Salson's avatar
Mikaël Salson committed
306 307
};

308 309 310

class KmerMultiSegmenter
{
311 312
 private:
  double threshold_nb_expected;
313 314 315
 public:
  /**
   * @param seq: An object read from a FASTA/FASTQ file
316 317 318 319
   * @param multigermline: the multigerm
   * @param threshold: threshold of randomly expected segmentation
   */
  KmerMultiSegmenter(Sequence seq, MultiGermline *multigermline, ostream *out_unsegmented,
320
                     double threshold = THRESHOLD_NB_EXPECTED, int nb_reads_for_evalue = 1);
321

322 323
  ~KmerMultiSegmenter();

324
  KmerSegmenter *the_kseg;
325
  MultiGermline *multi_germline;
326 327 328
};


Mikaël Salson's avatar
Mikaël Salson committed
329 330 331
class FineSegmenter : public Segmenter
{
 public:
Mikaël Salson's avatar
Mikaël Salson committed
332 333 334
   vector<pair<int, int> > score_V;
   vector<pair<int, int> > score_D;
   vector<pair<int, int> > score_J;
335 336

   vector <AlignBox*> boxes ;
Mikaël Salson's avatar
Mikaël Salson committed
337
   
Mikaël Salson's avatar
Mikaël Salson committed
338 339 340
   /**
   * Build a fineSegmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
341
   * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
342
   */
343
   FineSegmenter(Sequence seq, Germline *germline, Cost segment_cost,
344
                 double threshold = THRESHOLD_NB_EXPECTED, double multiplier=1.0);
345 346 347

   ~FineSegmenter();

Mikaël Salson's avatar
Mikaël Salson committed
348 349
  /**
  * extend segmentation from VJ to VDJ
350
  * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
351
  */
352
  void FineSegmentD(Germline *germline, bool several_D,
353
                    double threshold = THRESHOLD_NB_EXPECTED_D, double multiplier=1.0);
354 355

  bool FineSegmentD(Germline *germline,
356
                    AlignBox *box_Y, AlignBox *box_DD, AlignBox *box_Z,
357
                    int forbidden_id,
358
                    int extend_DD_on_Y, int extend_DD_on_Z,
359
                    double threshold = THRESHOLD_NB_EXPECTED_D, double multiplier=1.0);
360

Mathieu Giraud's avatar
Mathieu Giraud committed
361 362 363 364
  /**
   * 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
365
  void findCDR3();
Mathieu Giraud's avatar
Mathieu Giraud committed
366

Marc Duez's avatar
Marc Duez committed
367
  json toJson();
Mikaël Salson's avatar
Mikaël Salson committed
368
  
Mikaël Salson's avatar
Mikaël Salson committed
369 370 371
};


372
void align_against_collection(string &read, BioReader &rep, int forbidden_rep_id,
373 374
                              bool reverse_ref, bool reverse_both, bool local,
                              AlignBox *box, Cost segment_cost);
Mikaël Salson's avatar
Mikaël Salson committed
375 376

#endif