예제 #1
1
int splicing_dgesdd(const splicing_matrix_t *matrix, 
		    splicing_vector_t *values) {

  splicing_matrix_t tmp;
  int m=splicing_matrix_nrow(matrix);
  int n=splicing_matrix_ncol(matrix);
  int lda=m, minmn= m < n ? m : n, maxmn = m < n ? n : m;
  int lwork=-1;
  int info=0;
  splicing_vector_t work;
  splicing_vector_int_t iwork;
  char jobz='N';
  int dummy=1;
  double dummy2;
  
  SPLICING_CHECK(splicing_matrix_copy(&tmp, matrix));
  SPLICING_FINALLY(splicing_matrix_destroy, &tmp);
  SPLICING_CHECK(splicing_vector_init(&work, 1));
  SPLICING_FINALLY(splicing_vector_destroy, &work);
  SPLICING_CHECK(splicing_vector_int_init(&iwork, 8*minmn));
  SPLICING_FINALLY(splicing_vector_int_destroy, &iwork);

  SPLICING_CHECK(splicing_vector_resize(values, minmn));

  /* Get the optiomal lwork first*/
  splicingdgesdd_(&jobz, &m, &n, &MATRIX(tmp,0,0), &lda, VECTOR(*values),
		  /*U=*/ &dummy2, /*LDU=*/ &dummy, 
		  /*VT=*/ &dummy2, /*LDVT=*/ &dummy, 
		  VECTOR(work), &lwork, VECTOR(iwork), &info);

  lwork = VECTOR(work)[0];
  SPLICING_CHECK(splicing_vector_resize(&work, lwork));

  /* Now do the SVD */
  splicingdgesdd_(&jobz, &m, &n, &MATRIX(tmp,0,0), &lda, VECTOR(*values),
		  /*U=*/ &dummy2, /*LDU=*/ &dummy, 
		  /*VT=*/ &dummy2, /*LDVT=*/ &dummy, 
		  VECTOR(work), &lwork, VECTOR(iwork), &info);

  if (info != 0) { 
    SPLICING_ERROR("Cannot calculate SVD", SPLICING_ELAPACK);
  }

  splicing_vector_destroy(&work);
  splicing_vector_int_destroy(&iwork);
  splicing_matrix_destroy(&tmp);
  SPLICING_FINALLY_CLEAN(3);
  
  return 0;
}
예제 #2
0
파일: miso.c 프로젝트: mlovci/MISO
int splicing_reassign_samples(const splicing_matrix_t *matches, 
			      const splicing_vector_int_t *match_order,
			      const splicing_vector_t *psi, 
			      int noiso, splicing_vector_int_t *result) {

  int noreads = splicing_matrix_ncol(matches);
  int i, w;
  double *prev, *curr;
  double rand, sumpsi;
  int noValid;
  int *order=VECTOR(*match_order);
  splicing_vector_t cumsum;
  splicing_vector_int_t validIso;  

  SPLICING_CHECK(splicing_vector_init(&cumsum, noiso));
  SPLICING_FINALLY(splicing_vector_destroy, &cumsum);
  SPLICING_CHECK(splicing_vector_int_init(&validIso, noiso));
  SPLICING_FINALLY(splicing_vector_int_destroy, &validIso);

  SPLICING_CHECK(splicing_vector_int_resize(result, noreads));

  if (noreads == 0) { return 0; }  

  prev = curr = &MATRIX(*matches, 0, order[0]);
  CUMSUM();

  for (i=0; i<noreads; i++) {
    curr = &MATRIX(*matches, 0, order[i]);

    /* Maybe we need to update the cumulative sum */
    if (memcmp(prev, curr, sizeof(double)*noiso) != 0) { CUMSUM(); }

    if (noValid == 0) {
      VECTOR(*result)[order[i]] = -1;
    } else if (noValid == 1) {
      VECTOR(*result)[order[i]] = VECTOR(validIso)[0];
    } else if (noValid == 2) { 
      rand = RNG_UNIF01() * sumpsi;
      w = (rand < VECTOR(cumsum)[0]) ? VECTOR(validIso)[0] : 
	VECTOR(validIso)[1];
      VECTOR(*result)[order[i]] = w;
    } else {
      /* Draw */
      rand = RNG_UNIF01() * sumpsi;
      /* TODO: Binary search for interval, if many classes */
      for (w=0; rand > VECTOR(cumsum)[w]; w++) ;
      VECTOR(*result)[order[i]] = VECTOR(validIso)[w];
    }

    prev=curr;
  }

  splicing_vector_int_destroy(&validIso);
  splicing_vector_destroy(&cumsum);
  SPLICING_FINALLY_CLEAN(2);

  return 0;
}
예제 #3
0
파일: gff.c 프로젝트: mlovci/MISO
/* TODO: do not ignore size */
int splicing_gff_init(splicing_gff_t *gff, size_t size) {

  if (size < 0) { 
    SPLICING_ERROR("Cannot create GFF, `size' must be non-negative", 
		   SPLICING_EINVAL);
  }

  SPLICING_CHECK(splicing_strvector_init(&gff->seqids, 0));
  SPLICING_FINALLY(splicing_strvector_destroy, &gff->seqids);
  SPLICING_CHECK(splicing_strvector_init(&gff->sources, 0));
  SPLICING_FINALLY(splicing_strvector_destroy, &gff->sources);
  SPLICING_CHECK(splicing_strvector_init(&gff->ID, 0));
  SPLICING_FINALLY(splicing_strvector_destroy, &gff->ID);

  SPLICING_CHECK(splicing_vector_int_init(&gff->genes, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &gff->genes);
  SPLICING_CHECK(splicing_vector_int_init(&gff->transcripts, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &gff->transcripts);
  SPLICING_CHECK(splicing_vector_int_init(&gff->seqid, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &gff->seqid);
  SPLICING_CHECK(splicing_vector_int_init(&gff->source, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &gff->source);
  SPLICING_CHECK(splicing_vector_int_init(&gff->strand, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &gff->strand);
  SPLICING_CHECK(splicing_vector_int_init(&gff->type, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &gff->type);
  SPLICING_CHECK(splicing_vector_int_init(&gff->start, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &gff->start);
  SPLICING_CHECK(splicing_vector_int_init(&gff->end, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &gff->end);
  SPLICING_CHECK(splicing_vector_init(&gff->score, 0));
  SPLICING_FINALLY(splicing_vector_destroy, &gff->score);
  SPLICING_CHECK(splicing_vector_int_init(&gff->phase, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &gff->phase);
  SPLICING_CHECK(splicing_vector_int_init(&gff->parent, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &gff->parent);

  gff->n=0;
  gff->nogenes=0;
  gff->notranscripts=0;

  gff->last_gene_id = gff->last_mrna_id = SPLICING_STRVECTOR_ZERO;
  gff->last_gene_no = gff->last_mrna_no = -1;
  gff->last_seqid = gff->last_source = SPLICING_STRVECTOR_ZERO;

  SPLICING_FINALLY_CLEAN(14);

  return 0;
}
예제 #4
0
파일: pyconvert.c 프로젝트: mlovci/MISO
int pysplicing_to_vector(PyObject *pv, splicing_vector_t *v) {
  int i, n;
  
  if (!PyTuple_Check(pv)) {
    PyErr_SetString(PyExc_TypeError, "Need a tuple");
    return 1;
  }
  
  n=PyTuple_Size(pv);
  splicing_vector_init(v, n);
  for (i=0; i<n; i++) {
    PyObject *it=PyTuple_GetItem(pv, i);
    VECTOR(*v)[i]=PyFloat_AsDouble(it);
  }
  
  return 0;
}
예제 #5
0
파일: pysplicing.c 프로젝트: mlovci/MISO
static PyObject* pysplicing_solve_gene(PyObject *self, PyObject *args) {
  
  PyObject *gff, *readcigar, *position;
  int gene, readLength;
  splicing_gff_t *mygff;
  splicing_vector_int_t myposition;
  splicing_strvector_t myreadcigar;
  splicing_matrix_t match_matrix, assignment_matrix;
  splicing_vector_t expression;
  PyObject *r1, *r2, *r3;
  
  if (!PyArg_ParseTuple(args, "OiiOO", &gff, &gene, &readLength, &position,
			&readcigar)) { return NULL; }
  
  mygff=PyCObject_AsVoidPtr(gff);
  if (pysplicing_to_vector_int(position, &myposition)) { return NULL; }
  SPLICING_FINALLY(splicing_vector_int_destroy, &myposition);
  if (pysplicing_to_strvector(readcigar, &myreadcigar)) { return NULL; }
  SPLICING_FINALLY(splicing_strvector_destroy, &myreadcigar);

  SPLICING_PYCHECK(splicing_matrix_init(&match_matrix, 0, 0));
  SPLICING_FINALLY(splicing_matrix_destroy, &match_matrix);
  SPLICING_PYCHECK(splicing_matrix_init(&assignment_matrix, 0, 0));
  SPLICING_FINALLY(splicing_matrix_destroy, &assignment_matrix);
  SPLICING_PYCHECK(splicing_vector_init(&expression, 0));
  SPLICING_FINALLY(splicing_vector_destroy, &expression);

  SPLICING_PYCHECK(splicing_solve_gene(mygff, gene, readLength, &myposition, 
				       (const char **) myreadcigar.table,
				       &match_matrix, &assignment_matrix,
				       &expression));
  
  r1=pysplicing_from_matrix(&match_matrix);
  r2=pysplicing_from_matrix(&assignment_matrix);
  r3=pysplicing_from_vector(&expression);

  splicing_vector_destroy(&expression);
  splicing_matrix_destroy(&assignment_matrix);
  splicing_matrix_destroy(&match_matrix);
  splicing_strvector_destroy(&myreadcigar);
  splicing_vector_int_destroy(&myposition);
  SPLICING_FINALLY_CLEAN(5);

  return Py_BuildValue("OOO", r1, r2, r3);
}
예제 #6
0
파일: miso.c 프로젝트: mlovci/MISO
int splicing_score_iso(const splicing_vector_t *psi, int noiso, 
		       const splicing_vector_int_t *assignment, int noreads,
		       const splicing_vector_int_t *peffisolen, double *res) {
  int *effisolen = VECTOR(*peffisolen);
  double sum, maxpsieff, score;
  splicing_vector_t logpsi;
  int i;

  SPLICING_CHECK(splicing_vector_init(&logpsi, noiso));
  SPLICING_FINALLY(splicing_vector_destroy, &logpsi);

  /* Calculate the normalization factor */
  VECTOR(logpsi)[0] = log(VECTOR(*psi)[0]) + log(effisolen[0]);
  for (maxpsieff=VECTOR(logpsi)[0], i=1; i<noiso; i++) {
    VECTOR(logpsi)[i] = log(VECTOR(*psi)[i]) + log(effisolen[i]);
    if (VECTOR(logpsi)[i] > maxpsieff) { maxpsieff = VECTOR(logpsi)[i]; }
  }
  for (sum=0.0, i=0; i<noiso; i++) {
    sum += exp(VECTOR(logpsi)[i]-maxpsieff);
  }
  sum = log(sum) + maxpsieff;
  
  /* Normalize */
  for (i=0; i<noiso; i++) {
    VECTOR(logpsi)[i] -= sum;
  }
  
  /* Calculate score, based on assignments */
  for (score=0.0, i=0; i<noreads; i++) {
    if (VECTOR(*assignment)[i] != -1) {
      score += VECTOR(logpsi)[ VECTOR(*assignment)[i] ];
    }
  }

  splicing_vector_destroy(&logpsi);
  SPLICING_FINALLY_CLEAN(1);

  *res = score;
  return 0;
}
예제 #7
0
파일: miso.c 프로젝트: mlovci/MISO
int splicing_miso_trinity(const splicing_matrix_t *match_matrix,
			  const splicing_vector_int_t *isolen,
			  int readLength, int noIterations, int noBurnIn,
			  int noLag, const splicing_vector_t *hyperp,
			  splicing_matrix_t *samples, 
			  splicing_vector_t *logLik,
			  splicing_matrix_t *class_templates,
			  splicing_vector_t *class_counts,
			  splicing_vector_int_t *assignment,
			  splicing_miso_rundata_t *rundata) {

  double acceptP, cJS, pJS, sigma;
  int noiso = splicing_matrix_nrow(match_matrix);
  int noReads = splicing_matrix_ncol(match_matrix);
  splicing_vector_int_t *myass=assignment, vass;
  splicing_vector_t vpsi, vpsiNew, valpha, valphaNew, 
    *psi=&vpsi, *psiNew=&vpsiNew, *alpha=&valpha, *alphaNew=&valphaNew;
  int noSamples = (noIterations - noBurnIn + 1) / noLag;
  int i, m, lagCounter=0, noS=0;
  splicing_vector_int_t match_order;
  splicing_vector_int_t effisolen;
  splicing_vector_t isoscores;

  if ( (class_templates ? 1 : 0) + (class_counts ? 1 : 0) == 1) {
    SPLICING_ERROR("Only one of `class_templates' and `class_counts' is "
		   "given", SPLICING_EINVAL);
  }
  
  rundata->noIso=noiso;
  rundata->noIters=noIterations;
  rundata->noBurnIn=noBurnIn;
  rundata->noLag=noLag;
  rundata->noAccepted = rundata->noRejected = 0;

  if (assignment) { 
    SPLICING_CHECK(splicing_vector_int_resize(myass, noReads));
    splicing_vector_int_null(myass);
  } else {
    myass=&vass;
    SPLICING_CHECK(splicing_vector_int_init(myass, noReads));
    SPLICING_FINALLY(splicing_vector_int_destroy, myass);
  }
  SPLICING_CHECK(splicing_vector_init(&vpsi, noiso));
  SPLICING_FINALLY(splicing_vector_destroy, &vpsi);
  SPLICING_CHECK(splicing_vector_init(&vpsiNew, noiso));
  SPLICING_FINALLY(splicing_vector_destroy, &vpsiNew);
  SPLICING_CHECK(splicing_vector_init(&valpha, noiso-1));
  SPLICING_FINALLY(splicing_vector_destroy, &valpha);
  SPLICING_CHECK(splicing_vector_init(&valphaNew, noiso-1));
  SPLICING_FINALLY(splicing_vector_destroy, &valphaNew);

  SPLICING_CHECK(splicing_vector_int_init(&match_order, noReads));
  SPLICING_FINALLY(splicing_vector_int_destroy, &match_order);
  SPLICING_CHECK(splicing_order_matches(match_matrix, &match_order));

  if (class_templates && class_counts) { 
    SPLICING_CHECK(splicing_i_miso_classes(match_matrix, &match_order, 
					   class_templates, class_counts, 
					   /*bin_class_templates=*/ 0, 
					   /*bin_class_counts=*/ 0));
  }

  SPLICING_CHECK(splicing_vector_int_init(&effisolen, noiso));
  SPLICING_FINALLY(splicing_vector_int_destroy, &effisolen);
  SPLICING_CHECK(splicing_vector_init(&isoscores, noiso));
  SPLICING_FINALLY(splicing_vector_destroy, &isoscores);
  for (i=0; i<noiso; i++) { 
    int l=VECTOR(*isolen)[i]-readLength+1;
    VECTOR(effisolen)[i] = l > 0 ? l : 0;
    VECTOR(isoscores)[i] = -log((double) l);
  }

  SPLICING_CHECK(splicing_matrix_resize(samples, noiso, noSamples));
  SPLICING_CHECK(splicing_vector_resize(logLik, noSamples));

  /* Initialize Psi(0) randomly */

  SPLICING_CHECK(splicing_drift_proposal(/* mode= */ 0, 0, 0, 0, 0, 0, 
					 noiso, psi, alpha, &sigma, 0));
  SPLICING_CHECK(splicing_drift_proposal(/* mode= */ 1, psi, alpha, sigma,
					 0, 0, noiso, psi, alpha, 0, 0));

  /* Initialize assignments of reads */  
  
  SPLICING_CHECK(splicing_reassign_samples(match_matrix, &match_order, psi, 
					   noiso, myass));

  /* foreach Iteration m=1, ..., M do */

  for (m=0; m < noIterations; m++) {

    SPLICING_CHECK(splicing_drift_proposal(/* mode= */ 1, psi, alpha, sigma,
					   0, 0, noiso, psiNew, alphaNew, 0,
					   0));
    
    SPLICING_CHECK(splicing_metropolis_hastings_ratio(myass, noReads, psiNew,
						      alphaNew, psi, alpha,
						      sigma, noiso, 
						      &effisolen, hyperp,
						      &isoscores, 
						      m > 0 ? 1 : 0, 
						      &acceptP, &cJS, &pJS));
    
    if (acceptP >= 1 || RNG_UNIF01() < acceptP) {
      splicing_vector_t *tmp;
      tmp=psi; psi=psiNew; psiNew=tmp;
      tmp=alpha; alpha=alphaNew; alphaNew=tmp;
      cJS = pJS;
      rundata->noAccepted ++;
    } else {
      rundata->noRejected ++;
    }
    
    if (m >= noBurnIn) {
      if (lagCounter == noLag - 1) {
	memcpy(&MATRIX(*samples, 0, noS), VECTOR(*psi), 
	       noiso * sizeof(double));
	VECTOR(*logLik)[noS] = cJS;
	noS++;
	lagCounter = 0;
      } else {
	lagCounter ++;
      }
    }
    
    SPLICING_CHECK(splicing_reassign_samples(match_matrix, &match_order, 
					     psi, noiso, myass));

  } /* for m < noIterations */

  splicing_vector_destroy(&isoscores);
  splicing_vector_int_destroy(&effisolen);
  splicing_vector_int_destroy(&match_order);
  splicing_vector_destroy(&valphaNew);
  splicing_vector_destroy(&valpha);
  splicing_vector_destroy(&vpsiNew);
  splicing_vector_destroy(&vpsi);
  SPLICING_FINALLY_CLEAN(7);

  if (!assignment) { 
    splicing_vector_int_destroy(myass);
    SPLICING_FINALLY_CLEAN(1);
  }  
  
  return 0;
}
예제 #8
0
파일: pysplicing.c 프로젝트: mlovci/MISO
static PyObject* pysplicing_miso(PyObject *self, PyObject *args) {
  PyObject *gff, *readpos, *readcigar, *hyperp=0;
  int gene, readLength, noIterations=5000, noBurnIn=500, noLag=10;
  splicing_gff_t *mygff;
  splicing_strvector_t myreadcigar;
  splicing_vector_int_t myreadpos;
  splicing_vector_t myhyperp;
  splicing_matrix_t samples;
  splicing_vector_t logLik;
  splicing_matrix_t class_templates;
  splicing_vector_t class_counts;
  splicing_vector_int_t assignment;
  splicing_miso_rundata_t rundata;
  PyObject *r1, *r2, *r3, *r4, *r5, *r6;
  
  if (!PyArg_ParseTuple(args, "OiOOi|iiiO", &gff, &gene, &readpos, &readcigar,
			&readLength, &noIterations, &noBurnIn, &noLag, 
			&hyperp)) { return NULL; }
  
  mygff=PyCObject_AsVoidPtr(gff);

  SPLICING_PYCHECK(splicing_matrix_init(&samples, 0, 0));
  SPLICING_FINALLY(splicing_matrix_destroy, &samples);
  SPLICING_PYCHECK(splicing_vector_init(&logLik, 0));
  SPLICING_FINALLY(splicing_vector_destroy, &logLik);
  SPLICING_PYCHECK(splicing_matrix_init(&class_templates, 0, 0));
  SPLICING_FINALLY(splicing_matrix_destroy, &class_templates);
  SPLICING_PYCHECK(splicing_vector_init(&class_counts, 0));
  SPLICING_FINALLY(splicing_vector_destroy, &class_counts);
  SPLICING_PYCHECK(splicing_vector_int_init(&assignment, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &assignment);
  if (pysplicing_to_vector_int(readpos, &myreadpos)) { return NULL; }
  SPLICING_FINALLY(splicing_vector_int_destroy, &myreadpos);
  if (hyperp) { 
    if (pysplicing_to_vector(hyperp, &myhyperp)) { return NULL; }
    SPLICING_FINALLY(splicing_vector_destroy, &myhyperp);
  } else {
    size_t i, noiso;
    SPLICING_PYCHECK(splicing_gff_noiso_one(mygff, gene, &noiso));
    SPLICING_PYCHECK(splicing_vector_init(&myhyperp, noiso));
    SPLICING_FINALLY(splicing_vector_destroy, &myhyperp);
    for (i=0; i<noiso; i++) { VECTOR(myhyperp)[i] = 1.0; }
  }
  if (pysplicing_to_strvector(readcigar, &myreadcigar)) { return NULL; };
  SPLICING_FINALLY(splicing_strvector_destroy, &myreadcigar);
  
  SPLICING_PYCHECK(splicing_miso(mygff, gene, &myreadpos, 
				 (const char**) myreadcigar.table, 
				 readLength, noIterations, noBurnIn, noLag,
				 &myhyperp, &samples, &logLik, 
				 /*match_matrix=*/ 0, &class_templates,
				 &class_counts, &assignment, &rundata));

  splicing_vector_destroy(&myhyperp);
  splicing_vector_int_destroy(&myreadpos);
  splicing_strvector_destroy(&myreadcigar);
  SPLICING_FINALLY_CLEAN(3);
  
  r6=pysplicing_from_miso_rundata(&rundata);

  r5=pysplicing_from_vector_int(&assignment);
  splicing_vector_int_destroy(&assignment); SPLICING_FINALLY_CLEAN(1);

  r4=pysplicing_from_vector(&class_counts);
  splicing_vector_destroy(&class_counts); SPLICING_FINALLY_CLEAN(1);

  splicing_matrix_transpose(&class_templates);
  r3=pysplicing_from_matrix(&class_templates);
  splicing_matrix_destroy(&class_templates); SPLICING_FINALLY_CLEAN(1);

  r2=pysplicing_from_vector(&logLik);
  splicing_vector_destroy(&logLik); SPLICING_FINALLY_CLEAN(1);

  r1=pysplicing_from_matrix(&samples);
  splicing_matrix_destroy(&samples); SPLICING_FINALLY_CLEAN(1);
  
  return Py_BuildValue("OOOOOO", r1, r2, r3, r4, r5, r6);
}
예제 #9
0
파일: complexity.c 프로젝트: mlovci/MISO
int splicing_gene_complexity(const splicing_gff_t *gff, size_t gene,
			     int readLength, splicing_complexity_t type,
			     splicing_norm_t norm, int paired,
			     const splicing_vector_t *fragmentProb,
			     int fragmentStart, double normalMean, 
			     double normalVar, double numDevs,
			     double *complexity) {
  
  splicing_matrix_t assignment_matrix;

  SPLICING_CHECK(splicing_matrix_init(&assignment_matrix, 0, 0));
  SPLICING_FINALLY(splicing_matrix_destroy, &assignment_matrix);

  if (!paired) {
    SPLICING_CHECK(splicing_assignment_matrix(gff, gene, readLength, 
					      &assignment_matrix));
  } else {
    SPLICING_CHECK(splicing_paired_assignment_matrix(gff, gene, readLength, 
						     fragmentProb, 
						     fragmentStart,
						     normalMean, normalVar,
						     numDevs, 
						     &assignment_matrix));
  }

  switch (type) {
  case SPLICING_COMPLEXITY_RELATIVE:
    switch (norm) {
      splicing_vector_t values;
      int i, n;

    case SPLICING_NORM_2:

      SPLICING_CHECK(splicing_vector_init(&values, 0));
      SPLICING_FINALLY(splicing_vector_destroy, &values);
      SPLICING_CHECK(splicing_dgesdd(&assignment_matrix, &values));
      n=splicing_vector_size(&values);
      for (i=n-1; i>=0 && VECTOR(values)[i] < 1e-14; i--) ;
      *complexity = VECTOR(values)[0] / VECTOR(values)[i];
      splicing_vector_destroy(&values);
      SPLICING_FINALLY_CLEAN(1);
      break;

    case SPLICING_NORM_1:

      SPLICING_ERROR("One norm not implemented", SPLICING_UNIMPLEMENTED);
      break;

    case SPLICING_NORM_INFINITY:

      SPLICING_ERROR("Infinity norm not implemented", SPLICING_UNIMPLEMENTED);
      break;

    }
    break;
  case SPLICING_COMPLEXITY_ABSOLUTE:
    SPLICING_ERROR("Absolute complexity not implemented", 
		   SPLICING_UNIMPLEMENTED);
    break;
  }

  splicing_matrix_destroy(&assignment_matrix);
  SPLICING_FINALLY_CLEAN(1);

  return 0;
}
예제 #10
0
int splicing_simulate_reads(const splicing_gff_t *gff, int gene,
			    const splicing_vector_t *expression,
			    int noreads, int readLength,
			    splicing_vector_int_t *isoform, 
			    splicing_vector_int_t *position, 
			    splicing_strvector_t *cigar, 
			    splicing_vector_t *sample_prob) {
  
  size_t i, p, noiso, goodiso=0, nogenes;
  splicing_vector_int_t effisolen;
  splicing_vector_t sampleprob;
  double rand, sumpsi=0.0;
  splicing_vector_int_t exstart, exend, exidx;

  SPLICING_CHECK(splicing_gff_nogenes(gff, &nogenes));
  if (gene < 0 || gene >= nogenes) {
    SPLICING_ERROR("Invalid gene id", SPLICING_EINVAL);
  }

  /* TODO: more error checks */

  SPLICING_CHECK(splicing_gff_noiso_one(gff, gene, &noiso));
    
  SPLICING_CHECK(splicing_vector_int_init(&effisolen, noiso));
  SPLICING_FINALLY(splicing_vector_int_destroy, &effisolen);
  SPLICING_CHECK(splicing_vector_init(&sampleprob, noiso));
  SPLICING_FINALLY(splicing_vector_destroy, &sampleprob);
  SPLICING_CHECK(splicing_vector_int_resize(isoform, noreads));
  SPLICING_CHECK(splicing_gff_isolength_one(gff, gene, &effisolen));
  for (i=0; i<noiso; i++) {
    int l=VECTOR(effisolen)[i]-readLength+1;
    VECTOR(effisolen)[i] = l > 0 ? l : 0;
    VECTOR(sampleprob)[i] = VECTOR(*expression)[i] * VECTOR(effisolen)[i];
    if (VECTOR(sampleprob)[i] != 0) { goodiso++; }
    sumpsi += VECTOR(sampleprob)[i];
  }

  if (goodiso==0) {
    SPLICING_ERROR("No isoform is possible", SPLICING_FAILURE);
  }

  if (sample_prob) {
    SPLICING_CHECK(splicing_vector_update(sample_prob, &sampleprob));
  }

  for (i=1; i<noiso; i++) {
    VECTOR(sampleprob)[i] += VECTOR(sampleprob)[i-1];
  }

  for (i=0; i<noreads; i++) {
    int w;
    if (noiso==1) {
      w=0;
    } else if (noiso==2) {
      rand = RNG_UNIF01() * sumpsi;
      w = (rand < VECTOR(sampleprob)[0]) ? 0 : 1;
    } else {
      rand = RNG_UNIF01() * sumpsi;
      for (w=0; rand > VECTOR(sampleprob)[w]; w++) ;
    }
    VECTOR(*isoform)[i]=w;
  }
  
  splicing_vector_destroy(&sampleprob);
  SPLICING_FINALLY_CLEAN(1);

  /* OK, we have the isoforms, now we need the read positions, 
     these are uniformly sampled from the individual isoforms. */

  SPLICING_CHECK(splicing_vector_int_resize(position, noreads));
  SPLICING_CHECK(splicing_vector_int_init(&exstart, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &exstart);
  SPLICING_CHECK(splicing_vector_int_init(&exend, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &exend);
  SPLICING_CHECK(splicing_vector_int_init(&exidx, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &exidx);
  SPLICING_CHECK(splicing_gff_exon_start_end(gff, &exstart, &exend, &exidx,
					     gene));

  /* Positions in isoform coordinates first */

  for (i=0; i<noreads; i++) { 
    int iso=VECTOR(*isoform)[i];
    int len=VECTOR(effisolen)[iso];
    VECTOR(*position)[i]=RNG_INTEGER(1, len);
  }

  /* Translate isoform coordinates to genomic coordintes */

  /* TODO: some of this is already calculated */
  SPLICING_CHECK(splicing_iso_to_genomic(gff, gene, isoform, /*converter=*/ 0,
					 position));

  /* CIGAR strings */

  splicing_strvector_clear(cigar);
  SPLICING_CHECK(splicing_strvector_reserve(cigar, noreads));
  for (i=0; i<noreads; i++) {
    char tmp[1000], *tmp2=tmp;
    int iso=VECTOR(*isoform)[i];
    size_t rs=VECTOR(*position)[i];
    int ex=0;
    int rl=readLength;
    for (ex=VECTOR(exidx)[iso]; VECTOR(exend)[ex] < rs; ex++) ;
    while (VECTOR(exend)[ex] < rs+rl-1) {
      tmp2 += snprintf(tmp2, sizeof(tmp)/sizeof(char)-(tmp2-tmp)-1, "%iM%iN",
		       (int) (VECTOR(exend)[ex]-rs+1), 
		       (int) (VECTOR(exstart)[ex+1]-VECTOR(exend)[ex]-1));
      if (tmp2 >= tmp + sizeof(tmp)/sizeof(char)) {
	SPLICING_ERROR("CIGAR string too long", SPLICING_EINVAL);
      }
      rl -= (VECTOR(exend)[ex] - rs + 1);
      rs = VECTOR(exstart)[ex+1];
      ex++;
    }
    tmp2 += snprintf(tmp2, sizeof(tmp)/sizeof(char)-(tmp2-tmp)-1, "%iM", rl);
    if (tmp2 >= tmp + sizeof(tmp)/sizeof(char)) {
      SPLICING_ERROR("CIGAR string too long", SPLICING_EINVAL); }
    SPLICING_CHECK(splicing_strvector_append(cigar, tmp));
  }

  splicing_vector_int_destroy(&exidx);
  splicing_vector_int_destroy(&exend);
  splicing_vector_int_destroy(&exstart);
  splicing_vector_int_destroy(&effisolen);
  SPLICING_FINALLY_CLEAN(4);
  
  return 0;
}
예제 #11
0
int splicing_simulate_paired_reads(const splicing_gff_t *gff, int gene,
				   const splicing_vector_t *expression,
				   int noreads, int readLength,
				   const splicing_vector_t *fragmentProb,
				   int fragmentStart, double normalMean,
				   double normalVar, double numDevs,
				   splicing_vector_int_t *isoform,
				   splicing_vector_int_t *position,
				   splicing_strvector_t *cigar, 
				   splicing_vector_t *sampleprob) {
  
  size_t i, j, noiso, il, nogenes;
  splicing_vector_t *mysampleprob=sampleprob, vsampleprob;
  splicing_vector_t px, cpx;
  double sumpx, sumpsi=0.0;
  splicing_vector_int_t isolen;
  int goodiso=0;
  splicing_vector_int_t exstart, exend, exidx;
  splicing_vector_t *myfragmentProb=(splicing_vector_t*) fragmentProb,
    vfragmentProb;
  int fs, fl;

  SPLICING_CHECK(splicing_gff_nogenes(gff, &nogenes));
  if (gene < 0 || gene >= nogenes) {
    SPLICING_ERROR("Invalid gene id", SPLICING_EINVAL);
  }

  /* TODO: more error checks */

  if (!fragmentProb) { 
    myfragmentProb=&vfragmentProb;
    SPLICING_CHECK(splicing_vector_init(&vfragmentProb, 0));
    SPLICING_FINALLY(splicing_vector_destroy, &vfragmentProb);
    SPLICING_CHECK(splicing_normal_fragment(normalMean, normalVar, numDevs, 
					    readLength, myfragmentProb,
					    &fragmentStart));
    splicing_vector_scale(myfragmentProb, 
			  1.0/splicing_vector_sum(myfragmentProb));
  }

  il=splicing_vector_size(myfragmentProb);
  fs=fragmentStart;
  fl=fragmentStart+il-1;

  SPLICING_CHECK(splicing_gff_noiso_one(gff, gene, &noiso));
    
  if ( fabs(splicing_vector_sum(myfragmentProb) - 1.0) > 1e-10 ) {
    SPLICING_ERROR("Fragment length distribution does not sum up to 1", 
		   SPLICING_EINVAL);
  }

  SPLICING_CHECK(splicing_vector_int_init(&isolen, noiso));
  SPLICING_FINALLY(splicing_vector_int_destroy, &isolen);
  SPLICING_CHECK(splicing_gff_isolength_one(gff, gene, &isolen));
  
  SPLICING_CHECK(splicing_vector_copy(&px, myfragmentProb));
  SPLICING_FINALLY(splicing_vector_destroy, &px);
  SPLICING_CHECK(splicing_vector_init(&cpx, il));
  SPLICING_FINALLY(splicing_vector_destroy, &cpx);

  if (!sampleprob) {
    mysampleprob=&vsampleprob;
    SPLICING_CHECK(splicing_vector_init(mysampleprob, noiso));
    SPLICING_FINALLY(splicing_vector_destroy, mysampleprob);
  } else {
    SPLICING_CHECK(splicing_vector_resize(mysampleprob, noiso));
  }

  for (sumpx=VECTOR(px)[0], i=1; i<il; i++) {
    VECTOR(px)[i] += VECTOR(px)[i-1];
    sumpx += VECTOR(px)[i];
  }
  VECTOR(cpx)[0] = VECTOR(px)[0];
  for (i=1; i<il; i++) {
    VECTOR(cpx)[i] = VECTOR(cpx)[i-1] + VECTOR(px)[i];
  }

  for (i=0; i<noiso; i++) {
    int ilen=VECTOR(isolen)[i];
    int r1= ilen >= fl ? ilen - fl + 1 : 0;
    int r2= ilen >= fs ? (ilen >= fl ? fl - fs : ilen - fs + 1) : 0;
    /* int r3= fs - 1; */
    double sp=0.0;
    if (r1 > 0) { sp += r1; } 
    if (r2 > 0) { sp += VECTOR(cpx)[r2-1]; }
    VECTOR(*mysampleprob)[i] = sp * VECTOR(*expression)[i];
    if (VECTOR(*mysampleprob)[i] != 0) { goodiso += 1; }
    sumpsi += VECTOR(*mysampleprob)[i];
  }

  if (goodiso == 0) {
    SPLICING_ERROR("No isoform is possible", SPLICING_FAILURE);
  }

  for (i=1; i<noiso; i++) {
    VECTOR(*mysampleprob)[i] += VECTOR(*mysampleprob)[i-1];
  }

  SPLICING_CHECK(splicing_vector_int_resize(isoform, noreads*2));

  for (i=0; i<2*noreads; i+=2) {
    int w;
    double rand;
    if (noiso==1) {
      w=0;
    } else if (noiso==2) {
      rand = RNG_UNIF01() * sumpsi;
      w = (rand < VECTOR(*mysampleprob)[0]) ? 0 : 1;
    } else {
      rand = RNG_UNIF01() * sumpsi;
      for (w=0; rand > VECTOR(*mysampleprob)[w]; w++) ;
    }
    VECTOR(*isoform)[i]=VECTOR(*isoform)[i+1]=w;
  }

  if (!sampleprob) { 
    splicing_vector_destroy(mysampleprob);
    SPLICING_FINALLY_CLEAN(1);
  } else {
    for (i=noiso-1; i>0; i--) {
      VECTOR(*mysampleprob)[i] -= VECTOR(*mysampleprob)[i-1];
    }
  }

  /* We have the isoforms, now get the read positions. */
  
  SPLICING_CHECK(splicing_vector_int_resize(position, noreads*2));
  SPLICING_CHECK(splicing_vector_int_init(&exstart, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &exstart);
  SPLICING_CHECK(splicing_vector_int_init(&exend, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &exend);
  SPLICING_CHECK(splicing_vector_int_init(&exidx, 0));
  SPLICING_FINALLY(splicing_vector_int_destroy, &exidx);
  SPLICING_CHECK(splicing_gff_exon_start_end(gff, &exstart, &exend, &exidx,
					     gene));
  
  /* Positions in isoform coordinates first. 
     These are sampled based on the fragment length distribution. */

  for (i=0, j=0; i<noreads; i++) {
    int iso=VECTOR(*isoform)[2*i];
    int ilen=VECTOR(isolen)[iso];
    int r1= ilen >= fl ? ilen - fl + 1 : 0;
    int r2= ilen >= fs ? (ilen >= fl ? fl - fs : ilen - fs + 1) : 0;
    /* int r3= fs - 1; */
    int pos, fragment;
    double sp=0.0;
    if (r1 > 0) { sp += r1; } 
    if (r2 > 0) { sp += VECTOR(cpx)[r2-1]; }
    double rand=RNG_UNIF(0, sp);
    if (rand < r1) { 
      pos = ceil(rand);
    } else {
      int w;
      rand -= r1;
      for (w=0; VECTOR(cpx)[w] < rand; w++) ;
      pos = r1 + r2 - w;
    }

    if (pos <= r1) {
      rand=RNG_UNIF(0, 1.0);
    } else {
      rand=RNG_UNIF(0, VECTOR(px)[r1+r2-pos]);
    }
    for (fragment=0; VECTOR(px)[fragment] < rand; fragment++) ;
    fragment += fragmentStart;

    VECTOR(*position)[j++] = pos;
    VECTOR(*position)[j++] = pos+fragment-readLength;
    
  }

  /* Translate positions to genomic coordinates */

  /* TODO: some of this is already calculated */
  SPLICING_CHECK(splicing_iso_to_genomic(gff, gene, isoform, /*converter=*/ 0,
					 position));

  /* CIGAR strings */

  splicing_strvector_clear(cigar);
  SPLICING_CHECK(splicing_strvector_reserve(cigar, 2*noreads));
  for (j=0; j<2*noreads; j++) {
    char tmp[1000], *tmp2=tmp;
    int iso=VECTOR(*isoform)[j];
    size_t rs=VECTOR(*position)[j];
    int ex=0;
    int rl=readLength;
    for (ex=VECTOR(exidx)[iso]; VECTOR(exend)[ex] < rs; ex++) ;
    while (rs + rl - 1 > VECTOR(exend)[ex]) {
      tmp2 += snprintf(tmp2, sizeof(tmp)/sizeof(char)-(tmp2-tmp)-1, "%iM%iN",
		       (int) (VECTOR(exend)[ex]-rs+1), 
		       (int) (VECTOR(exstart)[ex+1]-VECTOR(exend)[ex]-1));
      if (tmp2 >= tmp + sizeof(tmp)/sizeof(char)) {
	SPLICING_ERROR("CIGAR string too long", SPLICING_EINVAL);
      }
      rl -= (VECTOR(exend)[ex] - rs + 1);
      rs = VECTOR(exstart)[ex+1];
      ex++;
    }
    tmp2 += snprintf(tmp2, sizeof(tmp)/sizeof(char)-(tmp2-tmp)-1, "%iM", rl);
    if (tmp2 >= tmp + sizeof(tmp)/sizeof(char)) {
      SPLICING_ERROR("CIGAR string too long", SPLICING_EINVAL);
    }
    SPLICING_CHECK(splicing_strvector_append(cigar, tmp));
  }

  splicing_vector_int_destroy(&exidx);
  splicing_vector_int_destroy(&exend);
  splicing_vector_int_destroy(&exstart);
  splicing_vector_destroy(&cpx);
  splicing_vector_destroy(&px);
  splicing_vector_int_destroy(&isolen);
  SPLICING_FINALLY_CLEAN(6);

  if (!fragmentProb) { 
    splicing_vector_destroy(myfragmentProb); 
    SPLICING_FINALLY_CLEAN(1);
  }

  return 0;
}