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
#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 42 43 44 45
#define DEFAULT_WINDOW_SHIFT 5  /* Number of nucleotide to try shifting or
                                   reducing the window when it doesn't fit at
                                   its place */
#define MINIMAL_WINDOW_LENGTH 20 /* As we now dynamically adapt the window
                                    length we need to specify a minimum
                                    otherwise we could go as low as
                                    2*DEFAULT_WINDOW_SHIFT */
46

47

Mikaël Salson's avatar
Mikaël Salson committed
48
using namespace std;
Marc Duez's avatar
Marc Duez committed
49
using json = nlohmann::json;
Mikaël Salson's avatar
Mikaël Salson committed
50

51 52 53 54
enum SEGMENTED { NOT_PROCESSED,
		 TOTAL_SEG_AND_WINDOW,
                 SEG_PLUS, SEG_MINUS,
                 UNSEG_TOO_SHORT, UNSEG_STRAND_NOT_CONSISTENT,
55
		 UNSEG_TOO_FEW_ZERO,  UNSEG_ONLY_V, UNSEG_ONLY_J,
56
                 UNSEG_BAD_DELTA_MIN, UNSEG_AMBIGUOUS,
57
		 UNSEG_TOO_SHORT_FOR_WINDOW,
58

Mikaël Salson's avatar
Mikaël Salson committed
59
		 STATS_SIZE } ;
60 61 62

#define STATS_FIRST_UNSEG UNSEG_TOO_SHORT

63 64 65 66
const char* const segmented_mesg[] = { "?",
                                       "SEG",
                                       "SEG_+", "SEG_-",
                                       "UNSEG too short", "UNSEG strand",
67
				       "UNSEG too few V/J", "UNSEG only V/5'", "UNSEG only J/3'",
68
				       "UNSEG < delta_min", "UNSEG ambiguous",
69
                                       "UNSEG too short w",
Mikaël Salson's avatar
Mikaël Salson committed
70
                                      } ;
Mikaël Salson's avatar
Mikaël Salson committed
71

72 73


74 75 76 77
/**
 * An alignment box (AlignBox) gather all parameters for a recombined gene segment (V, D, J, other D...)
 **/

78 79 80
class AlignBox
{
 public:
81
  string key;
82
  string color;
83

84 85 86 87 88
  int del_left;
  int start;
  int end;
  int del_right;

89
  AlignBox(string key = "", string color="");
90
  string getSequence(string sequence);
91
  void addToJson(json &seg);
92

93 94 95 96 97
  /**
   * Returns 'V', 'D', 'J', or possibly '5', '4', '3', '?', depending on the ref_label and on the key
   */
  char getInitial();

98 99 100 101 102
  /**
   * Returns the length
   */
  int getLength();

103 104 105 106 107 108 109 110 111 112 113 114
  /**
   * 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);

115
  /* Identifier, label and sequence of the reference sequence (the best one) */
116 117 118 119
  int ref_nb;
  string ref_label;
  string ref;

120 121 122
  /* Marked position, for Cys104 and Phe118/Trp118 */
  int marked_pos;

123 124
  /* Identifiers and scores of other possible reference sequence */
  vector<pair<int, int> > score;
125 126
};

127
ostream &operator<<(ostream &out, const AlignBox &box);
128

129
/**
130
 * Check whether there is an overlap between two boxes,
131 132 133
 * 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
134
 * @param *box_left, *box_right            the two boxes
135 136
 * @param segment_cost:                    the cost used by the dynamic programing
 *
137 138
 * @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
139 140 141 142 143
 *
 * @return                                 the N segment
 */

string check_and_resolve_overlap(string seq, int seq_begin, int seq_end,
144 145
                                 AlignBox *box_left, AlignBox *box_right,
                                 Cost segment_cost);
146

Mikaël Salson's avatar
Mikaël Salson committed
147 148 149
class Segmenter {
protected:
  string sequence;
150
  string sequence_or_rc;
151

152
  // JUNCTIONstart/end and CDR3start/end are 1-based
153
  int JUNCTIONstart, JUNCTIONend;
154
  string JUNCTIONaa;
155
  bool JUNCTIONproductive;
156

Marc Duez's avatar
Marc Duez committed
157
  int CDR3start, CDR3end;
158 159
  string CDR3nuc;
  string CDR3aa;
160

Mikaël Salson's avatar
Mikaël Salson committed
161
  bool reversed, segmented, dSegmented;
162
  int because;
Mikaël Salson's avatar
Mikaël Salson committed
163

164 165
  /**
   * Compares evalue_left, evalue_right and evalue against the provided threshold
166
   * @post some evalue is above the threshold ==> because is set to UNSEG_TOO_FEW_ZERO, UNSEG_ONLY_J or UNSEG_ONLY_V
167 168 169
   */
  void checkLeftRightEvaluesThreshold(double threshold, int strand);

Mikaël Salson's avatar
Mikaël Salson committed
170 171 172 173 174
  string removeChevauchement();
  bool finishSegmentation();
  bool finishSegmentationD();

 public:
175
  Germline *segmented_germline;
176
  string label;
Mikaël Salson's avatar
Mikaël Salson committed
177
  string code;
178 179
  string info;        // .vdj.fa header, fixed fields
  string info_extra;  // .vdj.fa header, other information, at the end of the header
180
  string seg_V, seg_N, seg_J, system;
181

182 183
  AlignBox *box_V, *box_D, *box_J;

184
  double evalue;
185 186
  double evalue_left;
  double evalue_right;
Mikaël Salson's avatar
Mikaël Salson committed
187 188 189
  string seg_N1, seg_D, seg_N2;
  Cost segment_cost;

Mathieu Giraud's avatar
Mathieu Giraud committed
190
  virtual ~Segmenter();
Mikaël Salson's avatar
Mikaël Salson committed
191

Mathieu Giraud's avatar
Mathieu Giraud committed
192
  /* Queries */
Mikaël Salson's avatar
Mikaël Salson committed
193 194 195 196 197 198 199 200 201 202 203 204 205 206



  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.
   */
207
  string getJunction(int l, int shift=0) const;
Mikaël Salson's avatar
Mikaël Salson committed
208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243

  /**
   * @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;

244 245 246 247 248 249 250 251 252
  /**
   * @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
253

254 255 256 257 258
  /**
   * @post status == SEG_PLUS || status == SEG_MINUS <==> isSegmented()
   */
  void setSegmentationStatus(int status);

Mikaël Salson's avatar
Mikaël Salson committed
259 260 261 262 263 264 265 266 267 268 269
  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
270
 private:
271
  KmerAffectAnalyser *kaa;
Mikaël Salson's avatar
Mikaël Salson committed
272 273 274 275
 protected:
  string affects;

 public:
276
  int score;
277 278
  int pvalue_left;
  int pvalue_right;
279
  KmerAffect before, after;
280 281

  KmerSegmenter();
Mikaël Salson's avatar
Mikaël Salson committed
282 283 284
  /**
   * Build a segmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
285
   * @param germline: the germline
Mikaël Salson's avatar
Mikaël Salson committed
286
   */
287
  KmerSegmenter(Sequence seq, Germline *germline, double threshold = THRESHOLD_NB_EXPECTED, double multiplier=1.0);
Mathieu Giraud's avatar
Mathieu Giraud committed
288

289 290
  KmerSegmenter(const KmerSegmenter &seg);

Mathieu Giraud's avatar
Mathieu Giraud committed
291 292 293 294 295
  ~KmerSegmenter();

  /**
   * @return the KmerAffectAnalyser of the current sequence.
   */
296
  KmerAffectAnalyser *getKmerAffectAnalyser() const;
Mathieu Giraud's avatar
Mathieu Giraud committed
297

298
  string getInfoLineWithAffects() const;
Marc Duez's avatar
Marc Duez committed
299
  json toJson();
300

Mathieu Giraud's avatar
Mathieu Giraud committed
301
 private:
302
  void computeSegmentation(int strand, KmerAffect left, KmerAffect right,
303
                           double threshold, double multiplier);
Mikaël Salson's avatar
Mikaël Salson committed
304 305
};

306 307 308

class KmerMultiSegmenter
{
309 310
 private:
  double threshold_nb_expected;
311 312 313
 public:
  /**
   * @param seq: An object read from a FASTA/FASTQ file
314 315 316 317
   * @param multigermline: the multigerm
   * @param threshold: threshold of randomly expected segmentation
   */
  KmerMultiSegmenter(Sequence seq, MultiGermline *multigermline, ostream *out_unsegmented,
318
                     double threshold = THRESHOLD_NB_EXPECTED, int nb_reads_for_evalue = 1);
319

320 321
  ~KmerMultiSegmenter();

322
  KmerSegmenter *the_kseg;
323
  MultiGermline *multi_germline;
324 325 326
};


Mikaël Salson's avatar
Mikaël Salson committed
327 328 329
class FineSegmenter : public Segmenter
{
 public:
Mikaël Salson's avatar
Mikaël Salson committed
330 331 332
   vector<pair<int, int> > score_V;
   vector<pair<int, int> > score_D;
   vector<pair<int, int> > score_J;
333 334

   vector <AlignBox*> boxes ;
Mikaël Salson's avatar
Mikaël Salson committed
335
   
Mikaël Salson's avatar
Mikaël Salson committed
336 337 338
   /**
   * Build a fineSegmenter based on KmerSegmentation
   * @param seq: An object read from a FASTA/FASTQ file
339
   * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
340
   */
341
   FineSegmenter(Sequence seq, Germline *germline, Cost segment_cost,
342
                 double threshold = THRESHOLD_NB_EXPECTED, double multiplier=1.0);
343 344 345

   ~FineSegmenter();

Mikaël Salson's avatar
Mikaël Salson committed
346 347
  /**
  * extend segmentation from VJ to VDJ
348
  * @param germline: germline used
Mikaël Salson's avatar
Mikaël Salson committed
349
  */
350
  void FineSegmentD(Germline *germline, bool several_D,
351
                    double threshold = THRESHOLD_NB_EXPECTED_D, double multiplier=1.0);
352 353

  bool FineSegmentD(Germline *germline,
354
                    AlignBox *box_Y, AlignBox *box_DD, AlignBox *box_Z,
355
                    int forbidden_id,
356
                    int extend_DD_on_Y, int extend_DD_on_Z,
357
                    double threshold = THRESHOLD_NB_EXPECTED_D, double multiplier=1.0);
358

Mathieu Giraud's avatar
Mathieu Giraud committed
359 360 361 362
  /**
   * 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
363
  void findCDR3();
Mathieu Giraud's avatar
Mathieu Giraud committed
364

Marc Duez's avatar
Marc Duez committed
365
  json toJson();
Mikaël Salson's avatar
Mikaël Salson committed
366
  
Mikaël Salson's avatar
Mikaël Salson committed
367 368 369
};


370
void align_against_collection(string &read, BioReader &rep, int forbidden_rep_id,
371 372
                              bool reverse_ref, bool reverse_both, bool local,
                              AlignBox *box, Cost segment_cost);
Mikaël Salson's avatar
Mikaël Salson committed
373 374

#endif