int splicing_drift_proposal(int mode,
const splicing_vector_t *psi,
const splicing_vector_t *alpha,
double sigma,
const splicing_vector_t *otherpsi,
const splicing_vector_t *otheralpha, int noiso,
splicing_vector_t *respsi,
splicing_vector_t *resalpha,
double *ressigma, double *resscore) {
switch (mode) {
case 0: /* init */
{
SPLICING_CHECK(splicing_vector_resize(respsi, noiso));
SPLICING_CHECK(splicing_vector_resize(resalpha, noiso-1));
if (noiso != 2) {
int i;
for (i=0; i<noiso; i++) {
VECTOR(*respsi)[i] = 1.0/noiso;
}
for (i=0; i<noiso-1; i++) {
VECTOR(*resalpha)[i] = 1.0/(noiso-1);
}
*ressigma = 0.05;
} else {
VECTOR(*respsi)[0] = RNG_UNIF01();
VECTOR(*respsi)[1] = 1 - VECTOR(*respsi)[0];
VECTOR(*resalpha)[0] = 0.0;
VECTOR(*resalpha)[1] = 0.0;
*ressigma = 0.05;
}
}
break;
case 1: /* propose */
{
int len=noiso-1;
double sumpsi=0.0;
SPLICING_CHECK(splicing_vector_reserve(respsi, len+1));
SPLICING_CHECK(splicing_mvrnorm(alpha, sigma, resalpha, len));
SPLICING_CHECK(splicing_logit_inv(resalpha, respsi, len));
sumpsi = splicing_vector_sum(respsi);
SPLICING_CHECK(splicing_vector_resize(respsi, len+1));
VECTOR(*respsi)[len] = 1-sumpsi;
}
break;
case 2: /* score */
SPLICING_CHECK(splicing_mvplogisnorm(psi, otheralpha, sigma, noiso-1,
resscore));
break;
}
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;
}