kmerstore.h 11.9 KB
Newer Older
Mikaël Salson's avatar
Mikaël Salson committed
1 2 3 4 5 6 7 8
#ifndef KMERSTORE_H
#define KMERSTORE_H

#include <map>
#include <string>
#include <list>
#include <stdexcept>
#include <stdint.h>
9
#include <math.h>
Mikaël Salson's avatar
Mikaël Salson committed
10 11 12 13 14 15 16 17 18 19
#include "fasta.h"
#include "tools.h"

using namespace std;

class Kmer {
public:
  unsigned int count;

  Kmer();
20 21 22 23

  /**
   * This constructor is used via a IKmerStore<Kmer> index (hence the argument list)       
   */
24
  Kmer(const string &label, int strand=1);
25

Mikaël Salson's avatar
Mikaël Salson committed
26 27
  Kmer &operator+=(const Kmer &);
  static bool hasRevcompSymetry();
28 29 30 31 32

  /**
   * @return true if the element is the same as when initialised with default constructor.
   */
  bool isNull();
Mikaël Salson's avatar
Mikaël Salson committed
33 34 35 36 37 38 39 40 41 42 43 44 45
} ;
ostream &operator<<(ostream &os, const Kmer &kmer);


/* K-mer indexing */

template <class T> class IKmerStore
{
protected:
  bool revcomp_indexed;
  int k; // weight of the seed
  int s; // span of the seed (s >= k)
  string seed ;
46
  size_t nb_kmers_inserted;
Mikaël Salson's avatar
Mikaël Salson committed
47 48 49 50 51

public:

  virtual ~IKmerStore();

52 53
  list< pair <T, string> > labels;

Mikaël Salson's avatar
Mikaël Salson committed
54 55 56
  /**
   * @param input: A single FASTA file
   * @param label: label that must be associated to the given files
57
   * @post All the sequences in the FASTA files have been indexed, and the label is stored in the list of labels
Mikaël Salson's avatar
Mikaël Salson committed
58 59 60 61 62 63
   */
  void insert(Fasta& input, const string& label="");

  /**
   * @param input: A list of FASTA files
   * @param label: label that must be associated to the given files
64
   * @post All the sequences in the FASTA files have been indexed, and the label is stored in the list of labels
Mikaël Salson's avatar
Mikaël Salson committed
65 66 67 68 69 70
   */
  void insert(list<Fasta>& input, const string& label="");
  
  /**
   * @param input: A sequence to be cut in k-mers
   * @param label: label that must be associated to the given files
71 72 73 74
   * @param keep_only: if > 0 will keep at most the last keep_only nucleotides
   *                   of the sequence. if < 0 will keep at most the first
   *                   keep_only nucleotides of the sequence. if == 0,
   *                   will keep all the sequence.
Mikaël Salson's avatar
Mikaël Salson committed
75 76
   * @post All the k-mers in the sequence have been indexed.
   */
77
  void insert(const seqtype &sequence,
78
              const string &label,
79
              bool ignore_extended_nucleotides=true, int keep_only = 0);
Mikaël Salson's avatar
Mikaël Salson committed
80 81 82 83 84

  /**
   * @param word: a k-mer
   * @return the value for the kmer as stored in the structure (don't do revcomp)
   */ 
85
  virtual T& get(seqtype &word) = 0;
Mikaël Salson's avatar
Mikaël Salson committed
86

87 88 89 90 91
  /**
   * @return the percentage of kmers that are set in the index
   */
  float getIndexLoad() const;

92 93 94 95 96
  /**
   * @return probability that the number of kmers is 'at_least' or more in a sequence of length 'length'
   */
  double getProbabilityAtLeastOrAbove(int at_least, int length) const;

Mikaël Salson's avatar
Mikaël Salson committed
97 98 99 100 101 102 103 104 105 106
  /**
   * @return the value of k
   */
  int getK() const;

  /**
   * @return the value of s
   */
  int getS() const;

107 108 109 110 111
  /**
   * @return the seed used
   */
  string getSeed() const;

Mikaël Salson's avatar
Mikaël Salson committed
112 113 114 115 116 117 118
  /**
   * @param seq: a sequence
   * @param no_revcomp: force not to revcomp the sequence, even if
   *                    the index was built with revcomp.
   * @return a vector of length seq.length() - getK() + 1 containing
   * for each k-mer the corresponding value in the index.
   */
119
  virtual vector<T> getResults(const seqtype &seq, bool no_revcomp=false) = 0;
Mikaël Salson's avatar
Mikaël Salson committed
120 121 122 123 124 125 126 127 128 129 130 131

  /**
   * @return true iff the revcomp is indexed
   */
  bool isRevcomp() const;

  /**
   * @param string: a k-mer
   * @return the value in the index associated to the given k-mer.
   *         isRevcomp() ==> the revcomp of the k-mer may be considered
   * @pre word.length() == this->k
   */
132
  virtual T& operator[](seqtype& word) = 0;
Mikaël Salson's avatar
Mikaël Salson committed
133 134 135 136 137 138 139
};


template <class T> 
class MapKmerStore : public IKmerStore<T>
{
public:	
140
  map<seqtype, T> store;
Mikaël Salson's avatar
Mikaël Salson committed
141 142 143 144 145 146 147

  /**
   * @param bool: if true, will also index revcomp
   * @param seed: the seed
   */
  MapKmerStore(string seed, bool=false);
  MapKmerStore(int k, bool=false);
148 149

  vector<T> getResults(const seqtype &seq, bool no_revcomp=false);
150 151
  T& get(seqtype &word);
  T& operator[](seqtype & word);
152 153 154

 private:
  void init();
Mikaël Salson's avatar
Mikaël Salson committed
155 156 157 158 159 160 161
};

template <class T> 
class ArrayKmerStore : public IKmerStore<T>
{
  T* store;
	
162
  int index(const seqtype& word) const;
163 164 165 166 167 168 169 170

  /**
   * Allocates memory for store
   * @throws bad_alloc: when there is not enough memory or when the
   *                    the total number of kmers cannot be stored in an 
   *                    unsigned int
   */
  void init();
Mikaël Salson's avatar
Mikaël Salson committed
171 172 173 174 175 176 177 178
public:
  /**
   * @param bool: if true, will also index revcomp
   * @param seed: the seed
   */
  ArrayKmerStore(string seed, bool=false);
  ArrayKmerStore(int k, bool=false);
  ~ArrayKmerStore();
179 180

  vector<T> getResults(const seqtype &seq, bool no_revcomp=false);	
181 182
  T& get(seqtype &word);
  T& operator[](seqtype & word);
183
  T& operator[](int word);
Mikaël Salson's avatar
Mikaël Salson committed
184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
};


// IKmerStore

template<class T>
IKmerStore<T>::~IKmerStore(){}

template<class T> 
void IKmerStore<T>::insert(list<Fasta>& input,
                           const string &label){
  for(list<Fasta>::iterator it = input.begin() ; it != input.end() ; it++){
    insert(*it, label);
  }
}

template<class T> 
void IKmerStore<T>::insert(Fasta& input,
                           const string &label){
  for (int r = 0; r < input.size(); r++) {
    insert(input.sequence(r), label);
  }
206

207
  labels.push_back(make_pair(T(label, 1), label)) ;
Mikaël Salson's avatar
Mikaël Salson committed
208 209 210
}

template<class T> 
211
void IKmerStore<T>::insert(const seqtype &sequence,
212
                           const string &label,
213 214 215 216 217 218 219 220 221 222
                           bool ignore_extended_nucleotides,
                           int keep_only){
  size_t start_indexing = 0;
  size_t end_indexing = sequence.length();
  if (keep_only > 0 && sequence.length() > (size_t)keep_only) {
    start_indexing = sequence.length() - keep_only;
  } else if (keep_only < 0 && sequence.length() > (size_t) -keep_only) {
    end_indexing = -keep_only;
  }
  for(size_t i = start_indexing ; i + s < end_indexing + 1 ; i++) {
223 224
    seqtype substr = sequence.substr(i, s);
    seqtype kmer = spaced(substr, seed);
225

226
    if (ignore_extended_nucleotides && has_extended_nucleotides(kmer))
227 228
      continue;

Mikaël Salson's avatar
Mikaël Salson committed
229 230
    int strand = 1;
    if (revcomp_indexed && T::hasRevcompSymetry()) {
231
      seqtype rc_kmer = revcomp(kmer);
Mikaël Salson's avatar
Mikaël Salson committed
232 233 234 235 236
      if (rc_kmer.compare(kmer) < 0) {
        strand = -1;
        kmer = rc_kmer;
      }
    }
237 238 239 240 241
    T &this_kmer = this->get(kmer);
    if (this_kmer.isNull()) {
      nb_kmers_inserted++;
    }
    this_kmer += T(label, strand);
Mikaël Salson's avatar
Mikaël Salson committed
242
    if (revcomp_indexed && ! T::hasRevcompSymetry()) {
243
      seqtype rc_kmer = spaced(revcomp(substr), seed);
244 245 246 247
      T &this_rc_kmer = this->get(rc_kmer);
      if (this_rc_kmer.isNull())
        nb_kmers_inserted++;
      this_rc_kmer += T(label, -1);
Mikaël Salson's avatar
Mikaël Salson committed
248 249 250
    }
  }
}
251 252 253 254 255

template<class T>
float IKmerStore<T>::getIndexLoad() const {
  return nb_kmers_inserted*1. / (1 << (2 * k));
}
Mikaël Salson's avatar
Mikaël Salson committed
256

257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278

template<class T>
double IKmerStore<T>::getProbabilityAtLeastOrAbove(int at_least, int length) const {

  // n: number of kmers in the sequence
  int n = length - getS() + 1;
  float index_load = getIndexLoad() ;

  double proba = 0;
  double probability_having_system = pow(index_load, at_least);
  double probability_not_having_system = pow(1 - index_load, n - at_least);
  for (int i=at_least; i<=n; i++) {
    proba += nChoosek(n, i) * probability_having_system * probability_not_having_system;
    probability_having_system *= index_load;
    probability_not_having_system /= (1 - index_load);
  }

  return proba;
}



Mikaël Salson's avatar
Mikaël Salson committed
279 280 281 282 283 284 285 286 287 288
template<class T>
int IKmerStore<T>::getK() const {
  return k;
}

template<class T>
int IKmerStore<T>::getS() const {
  return s;
}

289 290 291 292 293
template<class T>
string IKmerStore<T>::getSeed() const {
  return seed;
}

294 295

// .getResults()
Mikaël Salson's avatar
Mikaël Salson committed
296
template<class T>
297 298 299 300
vector<T> MapKmerStore<T>::getResults(const seqtype &seq, bool no_revcomp) {

  int s = IKmerStore<T>::getS();

Mikaël Salson's avatar
Mikaël Salson committed
301 302 303 304 305
  if ((int)seq.length() < s - 1) {
    return vector<T>(0);
  }
  vector<T> result(seq.length() - s + 1);
  for (size_t i=0; i + s < seq.length() + 1; i++) {
306
    seqtype kmer = spaced(seq.substr(i, s), IKmerStore<T>::seed);
307
    //    seqtype kmer = seq.substr(i, s);
Mikaël Salson's avatar
Mikaël Salson committed
308
    // cout << kmer << endl << kmer0 << endl << endl ;
309
    if (IKmerStore<T>::revcomp_indexed && no_revcomp) {
Mikaël Salson's avatar
Mikaël Salson committed
310 311 312 313 314 315 316 317
      result[i] = get(kmer);
    } else {
      result[i] = (*this)[kmer];
    }
  }
  return result;
}

318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
template<class T>
vector<T> ArrayKmerStore<T>::getResults(const seqtype &seq, bool no_revcomp) {
  
  int s = IKmerStore<T>::getS();

  int N = (int)seq.length();

  if (N < s - 1) {
    return vector<T>(0);
  }
  vector<T> result(N - s + 1);

  /* Read once the sequence, convert it to int* */
  int* intseq = new int[N];
  for (int i=0; i<N; i++)
    {
334
      intseq[i] = nuc_to_int(seq[i]);
335 336 337 338 339 340 341 342 343
    }

  /* Compute results */
  for (size_t i=0; (int) i+s < N+1; i++) {
    int kmer = spaced_int(intseq + i, IKmerStore<T>::seed);
    if (IKmerStore<T>::revcomp_indexed && no_revcomp) {
      result[i] = store[kmer]; // getfromint(kmer); // store[kmer];
      // cout << i << "/" << N << "  " << kmer << result[i] << endl ;
    } else {
344
      result[i] = (*this)[kmer]; // Deals with revcomp
345 346 347 348 349 350 351
    }
  }

  delete[] intseq ;
  return result;
}

Mikaël Salson's avatar
Mikaël Salson committed
352 353 354 355 356 357 358 359 360 361 362 363 364 365 366

template<class T>
bool IKmerStore<T>::isRevcomp() const {
  return revcomp_indexed;
}

// MapKmerStore

template <class T>
MapKmerStore<T>::MapKmerStore(string seed, bool revcomp){
  this->seed = seed;   
  int k = seed_weight(seed);
  this->k = k;
  this->s = seed.size();
  this->revcomp_indexed = revcomp;
367
  init();
Mikaël Salson's avatar
Mikaël Salson committed
368 369 370 371 372 373 374 375
}

template <class T>
MapKmerStore<T>::MapKmerStore(int k, bool revcomp){
  this->seed = seed_contiguous(k);
  this->k = k;
  this->s = k;
  this->revcomp_indexed = revcomp;
376 377 378 379 380 381
  init();
}

template <class T>
void MapKmerStore<T>::init() {
  this->nb_kmers_inserted = 0;
Mikaël Salson's avatar
Mikaël Salson committed
382 383 384
}

template <class T> 
385
T& MapKmerStore<T>::get(seqtype& word){
Mikaël Salson's avatar
Mikaël Salson committed
386 387 388 389
  return store[word];
}

template <class T> 
390
T& MapKmerStore<T>::operator[](seqtype& word){
Mikaël Salson's avatar
Mikaël Salson committed
391
  if (this->isRevcomp() && T::hasRevcompSymetry()) {
392
    seqtype rc_kmer = revcomp(word);
Mikaël Salson's avatar
Mikaël Salson committed
393 394 395 396 397 398 399 400 401
    if (rc_kmer.compare(word) < 0)
      word = rc_kmer;
  }
  return store[word];
}

// ArrayKmerStore

template <class T> 
402
ArrayKmerStore<T>::ArrayKmerStore(int k, bool revcomp) {
Mikaël Salson's avatar
Mikaël Salson committed
403 404 405 406
  this->seed = seed_contiguous(k);
  this->k = k;
  this->s = k;
  this->revcomp_indexed = revcomp;
407
  init();
Mikaël Salson's avatar
Mikaël Salson committed
408 409 410 411 412 413 414 415 416 417
}


template <class T> 
ArrayKmerStore<T>::ArrayKmerStore(string seed, bool revcomp){
  this->seed = seed; 
  int k = seed_weight(seed);
  this->k = k;
  this->s = seed.size();
  this->revcomp_indexed = revcomp;
418 419 420 421 422
  init();
}

template <class T>
void ArrayKmerStore<T>::init() {
423
  this->nb_kmers_inserted = 0;
424
  if ((size_t)(this->k << 1) >= sizeof(int) * 8)
Mikaël Salson's avatar
Mikaël Salson committed
425
    throw std::bad_alloc();
426
  store = new T[(unsigned int)1 << (this->k << 1)];
Mikaël Salson's avatar
Mikaël Salson committed
427 428 429 430 431 432 433 434 435 436 437 438 439 440
  if (! store)
    throw std::bad_alloc();
}

template <class T> 
ArrayKmerStore<T>::~ArrayKmerStore(){
  delete [] store;
}

/**	Returns index of word in an array of size 4^k
	Considering B(A) = 0, B(C) = 1, B(T) = 2, B(G) = 3
	index_word = \sum_{i=0}^{k-1}B(word[i])*4^(k-i-1)
**/
template <class T> 
441
int ArrayKmerStore<T>::index(const seqtype& word) const{
Mikaël Salson's avatar
Mikaël Salson committed
442
  return dna_to_int(word, this->k);
Mikaël Salson's avatar
Mikaël Salson committed
443 444 445
}

template <class T> 
446
T& ArrayKmerStore<T>::get(seqtype& word){
Mikaël Salson's avatar
Mikaël Salson committed
447 448 449 450
  return store[index(word)];
}

template <class T> 
451
T& ArrayKmerStore<T>::operator[](seqtype& word){
452 453 454 455 456
  return (*this)[index(word)];
}

template <class T> 
T& ArrayKmerStore<T>::operator[](int word){
Mikaël Salson's avatar
Mikaël Salson committed
457
  if (this->isRevcomp() && T::hasRevcompSymetry()) {
458 459
    int rc_kmer = revcomp_int(word, IKmerStore<T>::k);
    if (rc_kmer < word)
Mikaël Salson's avatar
Mikaël Salson committed
460 461
      word = rc_kmer;
  }
462
  return store[word];
Mikaël Salson's avatar
Mikaël Salson committed
463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483
}


/**
 * KmerStoreFactory is a factory that allows to create an index that best fits
 * your needs!
 */
class KmerStoreFactory {
 public:
  template<class T>
    static IKmerStore<T> *createIndex(string seed, bool revcomp=false);
  template<class T>
    static IKmerStore<T> *createIndex(int k, bool revcomp=false);
};

template<class T>
IKmerStore<T> *KmerStoreFactory::createIndex(string seed, bool revcomp) {
  IKmerStore<T> *index;
  try{
    index = new ArrayKmerStore<T>(seed, revcomp);
  }catch(exception e){
484
    cout << "  (using a MapKmer to fit into memory)" << endl;
Mikaël Salson's avatar
Mikaël Salson committed
485 486 487 488 489 490 491 492 493 494 495
    index = new MapKmerStore<T>(seed, revcomp);
  }
  return index;
}

template<class T>
IKmerStore<T> *KmerStoreFactory::createIndex(int k, bool revcomp) {
  return createIndex<T>(seed_contiguous(k), revcomp);
}

#endif