int splicing_gff_isolength(const splicing_gff_t *gff, splicing_vector_int_t *isolength, splicing_vector_int_t *isolength_idx) { size_t idx1; size_t nogenes=splicing_vector_int_size(&gff->genes); size_t notrans=splicing_vector_int_size(&gff->transcripts); int pos=-1, ipos=0; SPLICING_CHECK(splicing_vector_int_resize(isolength, notrans)); SPLICING_CHECK(splicing_vector_int_resize(isolength_idx, nogenes)); for (idx1=VECTOR(gff->genes)[0]; idx1 < gff->n; idx1++) { if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_GENE) { VECTOR(*isolength_idx)[ipos++]=pos+1; } else if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_MRNA) { VECTOR(*isolength)[++pos] = 0; } else if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_EXON) { VECTOR(*isolength)[pos] += (VECTOR(gff->end)[idx1] - VECTOR(gff->start)[idx1] + 1); } } return 0; }
int splicing_i_gff_exon_start_end_sort(const splicing_vector_int_t *start, const splicing_vector_int_t *end, const splicing_vector_int_t *idx, int iso, splicing_vector_int_t *tmp, splicing_vector_int_t *tmp2) { int i, j, from=VECTOR(*idx)[iso], to=VECTOR(*idx)[iso+1], len=to-from; SPLICING_CHECK(splicing_vector_int_resize(tmp, len)); SPLICING_CHECK(splicing_vector_int_resize(tmp2, len)); for (i=0; i<len; i++) { VECTOR(*tmp)[i]=i; } splicing_qsort_r(VECTOR(*tmp), len, sizeof(int), (void*) (VECTOR(*start)+from), splicing_i_gff_exon_start_end_sort_cmp); /* Store the order */ for (i=0, j=from; i<len; i++, j++) { VECTOR(*tmp2)[i]=VECTOR(*start)[j]; } for (i=0, j=from; i<len; i++, j++) { VECTOR(*start)[j] = VECTOR(*tmp2)[ VECTOR(*tmp)[i] ]; } for (i=0, j=from; i<len; i++, j++) { VECTOR(*tmp2)[i]=VECTOR(*end)[j]; } for (i=0, j=from; i<len; i++, j++) { VECTOR(*end)[j] = VECTOR(*tmp2)[ VECTOR(*tmp)[i] ]; } return 0; }
int splicing_gff_noexons_one(const splicing_gff_t *gff, size_t gene, splicing_vector_int_t *noexons) { size_t nogenes, idx1, idx2, noiso, pos, il; SPLICING_CHECK(splicing_gff_nogenes(gff, &nogenes)); if (gene >= nogenes) { SPLICING_ERROR("Invalid gene id", SPLICING_EINVAL); } idx1=VECTOR(gff->genes)[gene]; idx2= gene+1 == nogenes ? gff->n : VECTOR(gff->genes)[gene+1]; for (noiso=0; idx1 < idx2; idx1++) { if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_MRNA) { noiso += 1; } } SPLICING_CHECK(splicing_vector_int_resize(noexons, noiso)); idx1=VECTOR(gff->genes)[gene]; idx2= gene+1 == nogenes ? gff->n : VECTOR(gff->genes)[gene+1]; for (; idx1 < idx2 && VECTOR(gff->type)[idx1] != SPLICING_TYPE_MRNA; idx1++) ; idx1++; for (pos=0, il=0; idx1 < idx2; idx1++) { if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_MRNA) { VECTOR(*noexons)[pos++]=il; il=0; } else if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_EXON) { il++; } } VECTOR(*noexons)[pos++]=il; return 0; }
int splicing_gff_gene_start_end(const splicing_gff_t *gff, splicing_vector_int_t *start, splicing_vector_int_t *end) { size_t i, nogenes=splicing_vector_int_size(&gff->genes); SPLICING_CHECK(splicing_vector_int_resize(start, nogenes)); SPLICING_CHECK(splicing_vector_int_resize(end, nogenes)); for (i=0; i<nogenes; i++) { size_t idx=VECTOR(gff->genes)[i]; VECTOR(*start)[i] = VECTOR(gff->start)[idx]; VECTOR(*end)[i] = VECTOR(gff->end)[idx]; } return 0; }
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; }
int splicing_gff_exon_start_end(const splicing_gff_t *gff, splicing_vector_int_t *start, splicing_vector_int_t *end, splicing_vector_int_t *idx, int gene) { size_t noiso; int i=0, p=0, n=splicing_gff_size(gff); int pos; size_t nogenes; splicing_vector_int_t tmp, tmp2; SPLICING_CHECK(splicing_vector_int_init(&tmp, 10)); SPLICING_FINALLY(splicing_vector_int_destroy, &tmp); SPLICING_CHECK(splicing_vector_int_init(&tmp2, 10)); SPLICING_FINALLY(splicing_vector_int_destroy, &tmp2); SPLICING_CHECK(splicing_gff_nogenes(gff, &nogenes)); if (gene < 0 || gene >= nogenes) { SPLICING_ERROR("Invalid gene id", SPLICING_EINVAL); } pos=VECTOR(gff->genes)[gene]+1; SPLICING_CHECK(splicing_gff_noiso_one(gff, gene, &noiso)); splicing_vector_int_clear(start); splicing_vector_int_clear(end); SPLICING_CHECK(splicing_vector_int_resize(idx, noiso+1)); while (pos < n) { if (VECTOR(gff->type)[pos] == SPLICING_TYPE_EXON) { int s=VECTOR(gff->start)[pos]; int e=VECTOR(gff->end)[pos]; SPLICING_CHECK(splicing_vector_int_push_back(start, s)); p++; SPLICING_CHECK(splicing_vector_int_push_back(end, e)); } else if (VECTOR(gff->type)[pos] == SPLICING_TYPE_MRNA) { VECTOR(*idx)[i] = p; if (i!=0) { SPLICING_CHECK(splicing_i_gff_exon_start_end_sort(start, end, idx, i-1, &tmp, &tmp2)); } i++; } else if (VECTOR(gff->type)[pos] == SPLICING_TYPE_GENE) { break; } pos++; } VECTOR(*idx)[i] = p; SPLICING_CHECK(splicing_i_gff_exon_start_end_sort(start, end, idx, i-1, &tmp, &tmp2)); splicing_vector_int_destroy(&tmp2); splicing_vector_int_destroy(&tmp); SPLICING_FINALLY_CLEAN(1); return 0; }
int splicing_iso_to_genomic_all(const splicing_gff_t *gff, size_t gene, int position, const splicing_gff_converter_t *converter, splicing_vector_int_t *result) { size_t i; splicing_gff_converter_t vconverter, *myconverter = (splicing_gff_converter_t*) converter; if (position < 1) { SPLICING_ERROR("Invalid isoform coordinate, must the larger than zero", SPLICING_EINVAL); } if (!converter) { myconverter=&vconverter; SPLICING_CHECK(splicing_gff_converter_init(gff, gene, myconverter)); SPLICING_FINALLY(splicing_gff_converter_destroy, myconverter); } SPLICING_CHECK(splicing_vector_int_resize(result, myconverter->noiso)); /* TODO: find impossible positions */ for (i=0; i<myconverter->noiso; i++) { int ex; for (ex=VECTOR(myconverter->exidx)[i]; ex < VECTOR(myconverter->exidx)[i+1] && VECTOR(myconverter->exlim)[ex] <= position; ex++) ; if (ex < VECTOR(myconverter->exidx)[i+1]) { VECTOR(*result)[i] = position + VECTOR(myconverter->shift)[ex]; } else { VECTOR(*result)[i] = -1; } } if (!converter) { splicing_gff_converter_destroy(myconverter); SPLICING_FINALLY_CLEAN(1); } return 0; }
int splicing_i_gff_noiso_one(const splicing_gff_t *gff, size_t gene, size_t *noiso, splicing_vector_int_t *isolen) { size_t nogenes, idx1, idx2; SPLICING_CHECK(splicing_gff_nogenes(gff, &nogenes)); if (gene < 0 || gene >= nogenes) { SPLICING_ERROR("Invalid gene id", SPLICING_EINVAL); } idx1=VECTOR(gff->genes)[gene]; idx2= gene+1 == nogenes ? gff->n : VECTOR(gff->genes)[gene+1]; *noiso = 0; for ( ; idx1 < idx2; idx1++) { if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_MRNA) { *noiso += 1; } } if (isolen) { size_t il=0, pos=0; SPLICING_CHECK(splicing_vector_int_resize(isolen, *noiso)); idx1=VECTOR(gff->genes)[gene]; idx2= gene+1 == nogenes ? gff->n : VECTOR(gff->genes)[gene+1]; for (; idx1 < idx2 && VECTOR(gff->type)[idx1] != SPLICING_TYPE_MRNA; idx1++) ; idx1++; for (; idx1 < idx2; idx1++) { if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_MRNA) { VECTOR(*isolen)[pos++]=il; il = 0; } else if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_EXON) { il += VECTOR(gff->end)[idx1] - VECTOR(gff->start)[idx1] + 1; } } VECTOR(*isolen)[pos++]=il; } return 0; }
int splicing_gff_noiso(const splicing_gff_t *gff, splicing_vector_int_t *noiso) { size_t nogenes, idx1, idx2, pos=0; SPLICING_CHECK(splicing_gff_nogenes(gff, &nogenes)); idx1=VECTOR(gff->genes)[0]; idx2=gff->n; SPLICING_CHECK(splicing_vector_int_resize(noiso, nogenes)); splicing_vector_int_null(noiso); if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_GENE) { idx1++; } for (; idx1 < gff->n; idx1++) { if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_MRNA) { VECTOR(*noiso)[pos] += 1; } else if (VECTOR(gff->type)[idx1] == SPLICING_TYPE_GENE) { pos++; } } return 0; }
int splicing_genomic_to_iso_all(const splicing_gff_t *gff, size_t gene, int position, const splicing_gff_converter_t *converter, splicing_vector_int_t *result) { int i; splicing_gff_converter_t vconverter, *myconverter = (splicing_gff_converter_t*) converter; if (!converter) { myconverter=&vconverter; SPLICING_CHECK(splicing_gff_converter_init(gff, gene, myconverter)); SPLICING_FINALLY(splicing_gff_converter_destroy, myconverter); } SPLICING_CHECK(splicing_vector_int_resize(result, myconverter->noiso)); for (i=0; i<myconverter->noiso; i++) { size_t startpos=VECTOR(myconverter->exidx)[i]; size_t endpos=VECTOR(myconverter->exidx)[i+1]; int ex; for (ex=startpos; ex < endpos && VECTOR(myconverter->exend)[ex] < position; ex++) ; if (ex < endpos && VECTOR(myconverter->exstart)[ex] <= position && position <= VECTOR(myconverter->exend)[ex]) { VECTOR(*result)[i] = position - VECTOR(myconverter->shift)[ex]; } else { VECTOR(*result)[i] = -1; } } if (!converter) { splicing_gff_converter_destroy(myconverter); SPLICING_FINALLY_CLEAN(1); } return 0; }
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; }
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; }
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; }