/* Function: hmmpgmd2msa()
* Synopsis: Convert an HMMPGMD-derived data stream to an MSA, based
* on the corresponding hmm
*
* Purpose: Given a data stream from HMMPGMD of the form shown
* here, produce an MSA:
* HMMD_SEARCH_STATS
* P7_HITS array of size (nhits) from above?
* then repeats of P7_DOMAIN and P7_ALIDISPLAY data
* for the hits, where each hit with d domains
* produces
* d P7_DOMAINs
* then
* d P7_ALIDISPLAYs
* ... optionally adding a sequence with length matching
* that of the hmm, which will be included in the alignment.
*
* A further extension has been the ability to include or exclude
* sequences form the list of hits.
*
* This function's expected use is as a helper function for
* the hmmer website, which gets the above data stream from
* hmmpgmd.
*
* Args : data: a pointer to binary data in the format given above
* hmm: the HMM against which the alidisplay traces and
* additional sequences/traces are threaded to reach
* the returned msa.
* qsq : optional sequence to be included in the output msa;
* must have the same number of residues as the hmm
* has states, as each residue i will be aligned to
* state i.
* incl: optional array of sequence names, in the case of
* hmmpgmd a list of ints, which are are excluded due
* to the sequence threshold, but have been selected
* to be included in the alignment. This ties in
* with the way jackhmmer is implemented on the
* HMMER website.
* incl_size: required size of the incl array. zero if incl is null.
* excl: optional array of sequence names, in the case of
* hmmpgmd a list of ints, which are are included as they
* score above threshold, but have been selected
* to be excluded from the alignment.
* excl_size: required size of the excl array. zero if excl is null.
*
*
* Returns: Pointer to completed MSA object. NULL on error
*
*/
int
hmmpgmd2msa(void *data, P7_HMM *hmm, ESL_SQ *qsq, int *incl, int incl_size, int *excl, int excl_size, ESL_MSA **ret_msa) {
int i, j;
int c;
int status;
int set_included;
/* trace of the query sequence with N residues onto model with N match states */
P7_TRACE *qtr = NULL;
int extra_sqcnt = 0;
/* vars used to read from the binary data */
HMMD_SEARCH_STATS *stats = NULL; /* pointer to a single stats object, at the beginning of data */
P7_HIT *hits = NULL; /* an array of hits, at the appropriate offset in data */
/* vars used in msa construction */
P7_TOPHITS th;
P7_ALIDISPLAY *ad, *ad2;
ESL_MSA *msa = NULL;
P7_DOMAIN *dom = NULL;
char *p = (char*)data; /*pointer used to walk along data, must be char* to allow pointer arithmetic */
th.N = 0;
th.unsrt = NULL;
th.hit = NULL;
/* optionally build a faux trace for the query sequence: relative to core model (B->M_1..M_L->E) */
if (qsq != NULL) {
if (qsq->n != hmm->M) {
status = eslFAIL;
goto ERROR;
}
if ((qtr = p7_trace_Create()) == NULL) {status = eslFAIL; goto ERROR; }
if ((status = p7_trace_Append(qtr, p7T_B, 0, 0)) != eslOK) goto ERROR;
for (i = 1; i <= qsq->n; i++)
if ((status = p7_trace_Append(qtr, p7T_M, i, i)) != eslOK) goto ERROR;
if ((status = p7_trace_Append(qtr, p7T_E, 0, 0)) != eslOK) goto ERROR;
qtr->M = qsq->n;
qtr->L = qsq->n;
extra_sqcnt = 1;
}
/* get search stats + hit info */
stats = (HMMD_SEARCH_STATS*)p;
/* sanity check */
if ( ( stats->Z_setby != p7_ZSETBY_NTARGETS && stats->Z_setby != p7_ZSETBY_OPTION && stats->Z_setby != p7_ZSETBY_FILEINFO )
|| ( stats->domZ_setby != p7_ZSETBY_NTARGETS && stats->domZ_setby != p7_ZSETBY_OPTION && stats->domZ_setby != p7_ZSETBY_FILEINFO )
|| stats->nhits > 10000000
|| stats->elapsed > 1000000
) {
status = eslFAIL;
goto ERROR;
}
/* ok, it looks legitimate */
p += sizeof(HMMD_SEARCH_STATS);
hits = (P7_HIT*)p;
p += sizeof(P7_HIT) * stats->nhits;
/* create a tophits object, to be passed to p7_tophits_Alignment() */
ESL_ALLOC( th.unsrt, sizeof(P7_HIT) * stats->nhits);
memcpy( th.unsrt, hits, sizeof(P7_HIT) * stats->nhits);
ESL_ALLOC( th.hit, sizeof(P7_HIT*) * stats->nhits);
for (i=0; i<stats->nhits; i++) {
th.hit[i] = &(th.unsrt[i]);
if ( th.hit[i]->ndom > 10000
|| th.hit[i]->flags > p7_IS_INCLUDED + p7_IS_REPORTED + p7_IS_NEW + p7_IS_DROPPED + p7_IS_DUPLICATE
) {
status = eslFAIL;
goto ERROR;
}
}
// th.unsrt = NULL;
th.N = stats->nhits;
th.nreported = 0;
th.nincluded = 0;
th.is_sorted_by_sortkey = 0;
th.is_sorted_by_seqidx = 0;
for (i = 0; i < th.N; i++) {
ESL_ALLOC( th.hit[i]->dcl, sizeof(P7_DOMAIN) * th.hit[i]->ndom);
/* Go through the hits and set to be excluded or included as necessary */
set_included = 0;
if(th.hit[i]->flags & p7_IS_INCLUDED){
if(excl_size > 0){
for( c = 0; c < excl_size; c++){
if(excl[c] == (long)(th.hit[i]->name) ){
th.hit[i]->flags = p7_IS_DROPPED;
th.hit[i]->nincluded = 0;
break;
}
}
}
}else{
if(incl_size > 0){
for( c = 0; c < incl_size; c++){
if(incl[c] == (long)th.hit[i]->name ){
th.hit[i]->flags = p7_IS_INCLUDED;
set_included = 1;
}
}
}
}
/* first grab all the P7_DOMAINs for the hit */
for (j=0; j < th.hit[i]->ndom; j++) {
dom = th.hit[i]->dcl + j;
memcpy(dom , (P7_DOMAIN*)p, sizeof(P7_DOMAIN));
/* Possibly set domains to be include if being
* externally set via incl list*/
if(set_included) th.hit[i]->dcl[j].is_included = 1;
p += sizeof(P7_DOMAIN);
}
/* then grab the P7_ALIDISPLAYs for the hit */
for (j=0; j < th.hit[i]->ndom; j++) {
ad = (P7_ALIDISPLAY*)p;
ESL_ALLOC(th.hit[i]->dcl[j].ad, sizeof(P7_ALIDISPLAY));
ad2 = th.hit[i]->dcl[j].ad;
ad2->memsize = ad->memsize;
ad2->rfline = ad->rfline;
ad2->mmline = ad->mmline;
ad2->csline = ad->csline ;
ad2->model = ad->model ;
ad2->mline = ad->mline ;
ad2->aseq = ad->aseq ;
ad2->ppline = ad->ppline;
ad2->N = ad->N;
ad2->hmmname = ad->hmmname;
ad2->hmmacc = ad->hmmacc ;
ad2->hmmdesc = ad->hmmdesc;
ad2->hmmfrom = ad->hmmfrom;
ad2->hmmto = ad->hmmto;
ad2->M = ad->M;
ad2->sqname = ad->sqname;
ad2->sqacc = ad->sqacc ;
ad2->sqdesc = ad->sqdesc;
ad2->sqfrom = ad->sqfrom;
ad2->sqto = ad->sqto;
ad2->L = ad->L;
p += sizeof(P7_ALIDISPLAY);
ESL_ALLOC(ad2->mem, ad2->memsize);
memcpy(ad2->mem, p, ad->memsize);
p += ad2->memsize;
p7_alidisplay_Deserialize(ad2);
}
}
/* use the tophits and trace info above to produce an alignment */
if ( (status = p7_tophits_Alignment(&th, hmm->abc, &qsq, &qtr, extra_sqcnt, p7_ALL_CONSENSUS_COLS, &msa)) != eslOK) goto ERROR;
/* free memory */
if (qtr != NULL) free(qtr);
for (i = 0; i < th.N; i++) {
for (j=0; j < th.hit[i]->ndom; j++)
p7_alidisplay_Destroy(th.hit[i]->dcl[j].ad);
if (th.hit[i]->dcl != NULL) free (th.hit[i]->dcl);
}
if (th.unsrt != NULL) free (th.unsrt);
if (th.hit != NULL) free (th.hit);
*ret_msa = msa;
return eslOK;
ERROR:
/* free memory */
if (qtr != NULL) free(qtr);
for (i = 0; i < th.N; i++) {
for (j=0; j < th.hit[i]->ndom; j++)
p7_alidisplay_Destroy(th.hit[i]->dcl[j].ad);
if (th.hit[i]->dcl != NULL) free (th.hit[i]->dcl);
}
if (th.unsrt != NULL) free (th.unsrt);
if (th.hit != NULL) free (th.hit);
return status;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = esl_getopts_CreateDefaultApp(options, 1, argc, argv, banner, usage);
char *hmmfile = esl_opt_GetArg(go, 1);
int N = esl_opt_GetInteger(go, "-N");
ESL_STOPWATCH *w = esl_stopwatch_Create();
ESL_RANDOMNESS *r = esl_randomness_CreateFast(esl_opt_GetInteger(go, "-s"));
ESL_ALPHABET *abc = NULL;
P7_HMMFILE *hfp = NULL;
P7_HMM *hmm = NULL;
P7_BG *bg = NULL;
P7_PROFILE *gm = NULL;
P7_OPROFILE *om = NULL;
P7_TRACE *tr = NULL;
ESL_SQ *sq = NULL;
P7_ALIDISPLAY *ad = NULL;
int i,z;
if (p7_hmmfile_Open(hmmfile, NULL, &hfp) != eslOK) p7_Fail("Failed to open HMM file %s", hmmfile);
if (p7_hmmfile_Read(hfp, &abc, &hmm) != eslOK) p7_Fail("Failed to read HMM");
p7_hmmfile_Close(hfp);
bg = p7_bg_Create(abc);
p7_bg_SetLength(bg, 0);
gm = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, 0, p7_UNIGLOCAL); /* that sets N,C,J to generate nothing */
om = p7_oprofile_Create(gm->M, abc);
p7_oprofile_Convert(gm, om);
if (esl_opt_GetBoolean(go, "-p")) tr = p7_trace_CreateWithPP();
else tr = p7_trace_Create();
sq = esl_sq_CreateDigital(abc);
esl_stopwatch_Start(w);
for (i = 0; i < N; i++)
{
p7_ProfileEmit(r, hmm, gm, bg, sq, tr);
esl_sq_SetName(sq, "random");
if (! esl_opt_GetBoolean(go, "-b"))
{
if (esl_opt_GetBoolean(go, "-p"))
for (z = 0; z < tr->N; z++)
if (tr->i[z] > 0) tr->pp[z] = esl_random(r);
ad = p7_alidisplay_Create(tr, 0, om, sq);
p7_alidisplay_Print(stdout, ad, 40, 80, FALSE);
p7_alidisplay_Destroy(ad);
}
p7_trace_Reuse(tr);
esl_sq_Reuse(sq);
}
esl_stopwatch_Stop(w);
esl_stopwatch_Display(stdout, w, "# CPU time: ");
esl_sq_Destroy(sq);
p7_trace_Destroy(tr);
p7_oprofile_Destroy(om);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
esl_alphabet_Destroy(abc);
esl_randomness_Destroy(r);
esl_stopwatch_Destroy(w);
esl_getopts_Destroy(go);
return 0;
}
/* Function: hmmpgmd2stats()
* Synopsis: Use a HMMPGMD-derived data stream to extract some simple
* statistics regarding its alignment.
* Purpose: Given a data stream from HMMPGMD of the form shown
* here, produce a vector of floats:
* positions 0 to hmm->M-1 are:
* Fraction of alignments which cover model at that position
* positions hmm->M to hmm->M*2-1
* Fraction of alignments which cover model at that position (mod hmm->M)
* with a similar residue
* positions hmm->M*2 to hmm->M*3-1
* Fraction of alignments which cover model at that position (mod hmm->M)
* with the consensus residue
*
* Args : data: a pointer to binary data in the format given above
* hmm: the HMM against which the alidisplay traces and
* additional sequences/traces are threaded to reach
* the returned msa.
*
* Returns: the location where the output vector will be placed.
* caller is responsible for freeing it later
*
*/
int hmmpgmd2stats(void *data, P7_HMM *hmm, float** statsOut)
{
int i, j, k;
int status;
/* trace of the query sequence with N residues onto model with N match states */
P7_TRACE *qtr = NULL;
/* vars used to read from the binary data */
HMMD_SEARCH_STATS *stats = NULL; /* pointer to a single stats object, at the beginning of data */
P7_HIT *hits = NULL; /* an array of hits, at the appropriate offset in data */
P7_TOPHITS th;
P7_DOMAIN *dom;
P7_ALIDISPLAY *ad, *ad2;
int *cover, *id, *similar; //store statistics result per hit
int readPos, writePos; //for converting alignment contents into model indexing
char *p = (char*)data; /*pointer used to walk along data, must be char* to allow pointer arithmetic */
th.N = 0;
th.unsrt = NULL;
th.hit = NULL;
//storage for output
ESL_ALLOC( *statsOut, sizeof(float) * hmm->M * 3);
//storage for accumulation per hit
ESL_ALLOC( cover, sizeof(int) * hmm->M);
ESL_ALLOC( id, sizeof(int) * hmm->M);
ESL_ALLOC( similar, sizeof(int) * hmm->M);
for(k = 0; k < hmm->M; k++)
{
cover[k] = 0;
id[k] = 0;
similar[k] = 0;
(*statsOut)[k ] = 0;
(*statsOut)[k+hmm->M ] = 0;
(*statsOut)[k+hmm->M*2] = 0;
}
/* get search stats + hit info */
stats = (HMMD_SEARCH_STATS*)p;
/* sanity check */
if ( ( stats->Z_setby != p7_ZSETBY_NTARGETS && stats->Z_setby != p7_ZSETBY_OPTION && stats->Z_setby != p7_ZSETBY_FILEINFO )
|| ( stats->domZ_setby != p7_ZSETBY_NTARGETS && stats->domZ_setby != p7_ZSETBY_OPTION && stats->domZ_setby != p7_ZSETBY_FILEINFO )
|| stats->nhits > 10000000
|| stats->elapsed > 1000000
) {
status = eslFAIL;
goto ERROR;
}
/* ok, it looks legitimate */
p += sizeof(HMMD_SEARCH_STATS);
hits = (P7_HIT*)p;
p += sizeof(P7_HIT) * stats->nhits;
/* create a tophits object, use it to step through the alignments */
ESL_ALLOC( th.unsrt, sizeof(P7_HIT) * stats->nhits);
memcpy( th.unsrt, hits, sizeof(P7_HIT) * stats->nhits);
ESL_ALLOC( th.hit, sizeof(P7_HIT*) * stats->nhits);
for (i=0; i<stats->nhits; i++) {
th.hit[i] = &(th.unsrt[i]);
if ( th.hit[i]->ndom > 10000
|| th.hit[i]->flags > p7_IS_INCLUDED + p7_IS_REPORTED + p7_IS_NEW + p7_IS_DROPPED + p7_IS_DUPLICATE
) {
status = eslFAIL;
goto ERROR;
}
}
th.N = stats->nhits;
th.nreported = 0;
th.nincluded = 0;
th.is_sorted_by_sortkey = 0;
th.is_sorted_by_seqidx = 0;
for (i = 0; i < th.N; i++)
{
ESL_ALLOC( th.hit[i]->dcl, sizeof(P7_DOMAIN) * th.hit[i]->ndom);
if(th.hit[i]->flags & p7_IS_INCLUDED) th.nincluded++;
/* first grab all the P7_DOMAINs for the hit */
for (j=0; j < th.hit[i]->ndom; j++)
{
dom = (P7_DOMAIN*)p;
th.hit[i]->dcl[j].is_included = dom->is_included;
p += sizeof(P7_DOMAIN);
}
/* then grab the P7_ALIDISPLAYs for the hit */
for (j=0; j < th.hit[i]->ndom; j++)
{
ad = (P7_ALIDISPLAY*)p;
ESL_ALLOC(th.hit[i]->dcl[j].ad, sizeof(P7_ALIDISPLAY));
ad2 = th.hit[i]->dcl[j].ad;
ad2->memsize = ad->memsize;
ad2->rfline = ad->rfline;
ad2->mmline = ad->mmline;
ad2->csline = ad->csline ;
ad2->model = ad->model ;
ad2->mline = ad->mline ;
ad2->aseq = ad->aseq ;
ad2->ppline = ad->ppline;
ad2->N = ad->N;
ad2->hmmname = ad->hmmname;
ad2->hmmacc = ad->hmmacc ;
ad2->hmmdesc = ad->hmmdesc;
ad2->hmmfrom = ad->hmmfrom;
ad2->hmmto = ad->hmmto;
ad2->M = ad->M;
ad2->sqname = ad->sqname;
ad2->sqacc = ad->sqacc ;
ad2->sqdesc = ad->sqdesc;
ad2->sqfrom = ad->sqfrom;
ad2->sqto = ad->sqto;
ad2->L = ad->L;
p += sizeof(P7_ALIDISPLAY);
ESL_ALLOC(ad2->mem, ad2->memsize);
memcpy(ad2->mem, p, ad->memsize);
p += ad2->memsize;
p7_alidisplay_Deserialize(ad2);
if(th.hit[i]->flags & p7_IS_INCLUDED && th.hit[i]->dcl[j].is_included)
{
writePos = ad2->hmmfrom-1;
readPos = 0;
while(readPos < ad2->N)
{
//check if model covers residue
if(isupper(ad2->aseq[readPos]) || ad2->aseq[readPos] == '-')
{
cover[writePos]++;
//check mline for id
if(isalpha(ad2->mline[readPos]))
{
id[writePos]++;
similar[writePos]++;
}
//check mline for not-a-space
else if(ad2->mline[readPos] == '+')
{
similar[writePos]++;
}
writePos++;
}
readPos++;
}
}
}
//increment output, adjusting for overlaps
for(k = 0; k < hmm->M; k++)
{
if(cover[k]) (*statsOut)[k]+=1.0;
if(id[k]) (*statsOut)[k+hmm->M]+=(id[k]/cover[k]);
id[k] = 0;
if(similar[k]) (*statsOut)[k+hmm->M*2]+=(similar[k]/cover[k]);
similar[k] = 0;
cover[k] = 0;
}
}
for(i = 0; i < hmm->M*3; i++)
{
(*statsOut)[i] = (*statsOut)[i]/(th.nincluded);
}
for(i = hmm->M; i < hmm->M*3; i++)
{
if((*statsOut)[i%hmm->M])
{
(*statsOut)[i] = (*statsOut)[i]/(*statsOut)[i%hmm->M];
}
else
{
(*statsOut)[i] = 0.0;
}
}
/* free memory */
if (qtr != NULL) free(qtr);
qtr = NULL;
for (i = 0; i < th.N; i++) {
for (j=0; j < th.hit[i]->ndom; j++)
if(th.hit[i]->dcl[j].ad)
{
p7_alidisplay_Destroy(th.hit[i]->dcl[j].ad);
th.hit[i]->dcl[j].ad = NULL;
}
if (th.hit[i]->dcl != NULL) free (th.hit[i]->dcl);
th.hit[i]->dcl = NULL;
}
if (th.unsrt != NULL) free (th.unsrt);
th.unsrt = NULL;
if (th.hit != NULL) free (th.hit);
th.hit = NULL;
return eslOK;
ERROR:
/* free memory */
if (qtr != NULL) free(qtr);
qtr = NULL;
for (i = 0; i < th.N; i++) {
for (j=0; j < th.hit[i]->ndom; j++)
if(th.hit[i]->dcl[j].ad)
{
p7_alidisplay_Destroy(th.hit[i]->dcl[j].ad);
th.hit[i]->dcl[j].ad = NULL;
}
if (th.hit[i]->dcl != NULL) free (th.hit[i]->dcl);
th.hit[i]->dcl = NULL;
}
if (th.unsrt != NULL) free (th.unsrt);
th.unsrt = NULL;
if (th.hit != NULL) free (th.hit);
th.hit = NULL;
return status;
}
/* Function: p7_alidisplay_Create()
* Synopsis: Create an alignment display, from trace and oprofile.
* Incept: SRE, Sun Dec 30 09:13:31 2007 [Janelia]
*
* Purpose: Creates and returns an alignment display for domain number
* <which> in traceback <tr>, where the traceback
* corresponds to an alignment of optimized profile <om> to digital sequence
* <dsq>, and the unique name of that target
* sequence <dsq> is <sqname>. The <which> index starts at 0.
*
* It will be a little faster if the trace is indexed with
* <p7_trace_Index()> first. The number of domains is then
* in <tr->ndom>. If the caller wants to create alidisplays
* for all of these, it would loop <which> from
* <0..tr->ndom-1>.
*
* However, even without an index, the routine will work fine.
*
* Args: tr - traceback
* which - domain number, 0..tr->ndom-1
* om - optimized profile (query)
* sq - digital sequence (target)
*
* Returns: <eslOK> on success.
*
* Throws: <NULL> on allocation failure, or if something's internally corrupt
* in the data.
*/
P7_ALIDISPLAY *
p7_alidisplay_Create(const P7_TRACE *tr, int which, const P7_OPROFILE *om, const ESL_SQ *sq)
{
P7_ALIDISPLAY *ad = NULL;
char *Alphabet = om->abc->sym;
int n, pos, z;
int z1,z2;
int k,x,i,s;
int hmm_namelen, hmm_acclen, hmm_desclen;
int sq_namelen, sq_acclen, sq_desclen;
int status;
/* First figure out which piece of the trace (from first match to last match)
* we're going to represent, and how big it is.
*/
if (tr->ndom > 0) { /* if we have an index, this is a little faster: */
for (z1 = tr->tfrom[which]; z1 < tr->N; z1++) if (tr->st[z1] == p7T_M) break; /* find next M state */
if (z1 == tr->N) return NULL; /* no M? corrupt trace */
for (z2 = tr->tto[which]; z2 >= 0 ; z2--) if (tr->st[z2] == p7T_M) break; /* find prev M state */
if (z2 == -1) return NULL; /* no M? corrupt trace */
} else { /* without an index, we can still do it fine: */
for (z1 = 0; which >= 0 && z1 < tr->N; z1++) if (tr->st[z1] == p7T_B) which--; /* find the right B state */
if (z1 == tr->N) return NULL; /* no such domain <which> */
for (; z1 < tr->N; z1++) if (tr->st[z1] == p7T_M) break; /* find next M state */
if (z1 == tr->N) return NULL; /* no M? corrupt trace */
for (z2 = z1; z2 < tr->N; z2++) if (tr->st[z2] == p7T_E) break; /* find the next E state */
for (; z2 >= 0; z2--) if (tr->st[z2] == p7T_M) break; /* find prev M state */
if (z2 == -1) return NULL; /* no M? corrupt trace */
}
/* Now we know that z1..z2 in the trace will be represented in the
* alidisplay; that's z2-z1+1 positions. We need a \0 trailer on all
* our display lines, so allocate z2-z1+2. We know each position is
* M, D, or I, so there's a 1:1 correspondence of trace positions
* with alignment display positions. We also know the display
* starts and ends with M states.
*
* So now let's allocate. The alidisplay is packed into a single
* memory space, so this appears to be intricate, but it's just
* bookkeeping.
*/
n = (z2-z1+2) * 3; /* model, mline, aseq mandatory */
if (om->rf[0] != 0) n += z2-z1+2; /* optional reference line */
if (om->cs[0] != 0) n += z2-z1+2; /* optional structure line */
if (tr->pp != NULL) n += z2-z1+2; /* optional posterior prob line */
hmm_namelen = strlen(om->name); n += hmm_namelen + 1;
hmm_acclen = (om->acc != NULL ? strlen(om->acc) : 0); n += hmm_acclen + 1;
hmm_desclen = (om->desc != NULL ? strlen(om->desc) : 0); n += hmm_desclen + 1;
sq_namelen = strlen(sq->name); n += sq_namelen + 1;
sq_acclen = strlen(sq->acc); n += sq_acclen + 1; /* sq->acc is "\0" when unset */
sq_desclen = strlen(sq->desc); n += sq_desclen + 1; /* same for desc */
ESL_ALLOC(ad, sizeof(P7_ALIDISPLAY));
ad->mem = NULL;
pos = 0;
ad->memsize = sizeof(char) * n;
ESL_ALLOC(ad->mem, ad->memsize);
if (om->rf[0] != 0) { ad->rfline = ad->mem + pos; pos += z2-z1+2; } else { ad->rfline = NULL; }
if (om->cs[0] != 0) { ad->csline = ad->mem + pos; pos += z2-z1+2; } else { ad->csline = NULL; }
ad->model = ad->mem + pos; pos += z2-z1+2;
ad->mline = ad->mem + pos; pos += z2-z1+2;
ad->aseq = ad->mem + pos; pos += z2-z1+2;
if (tr->pp != NULL) { ad->ppline = ad->mem + pos; pos += z2-z1+2;} else { ad->ppline = NULL; }
ad->hmmname = ad->mem + pos; pos += hmm_namelen +1;
ad->hmmacc = ad->mem + pos; pos += hmm_acclen +1;
ad->hmmdesc = ad->mem + pos; pos += hmm_desclen +1;
ad->sqname = ad->mem + pos; pos += sq_namelen +1;
ad->sqacc = ad->mem + pos; pos += sq_acclen +1;
ad->sqdesc = ad->mem + pos; pos += sq_desclen +1;
strcpy(ad->hmmname, om->name);
if (om->acc != NULL) strcpy(ad->hmmacc, om->acc); else ad->hmmacc[0] = 0;
if (om->desc != NULL) strcpy(ad->hmmdesc, om->desc); else ad->hmmdesc[0] = 0;
strcpy(ad->sqname, sq->name);
strcpy(ad->sqacc, sq->acc);
strcpy(ad->sqdesc, sq->desc);
/* Determine hit coords */
ad->hmmfrom = tr->k[z1];
ad->hmmto = tr->k[z2];
ad->M = om->M;
ad->sqfrom = tr->i[z1];
ad->sqto = tr->i[z2];
ad->L = sq->n;
/* optional rf line */
if (ad->rfline != NULL) {
for (z = z1; z <= z2; z++) ad->rfline[z-z1] = ((tr->st[z] == p7T_I) ? '.' : om->rf[tr->k[z]]);
ad->rfline[z-z1] = '\0';
}
/* optional cs line */
if (ad->csline != NULL) {
for (z = z1; z <= z2; z++) ad->csline[z-z1] = ((tr->st[z] == p7T_I) ? '.' : om->cs[tr->k[z]]);
ad->csline[z-z1] = '\0';
}
/* optional pp line */
if (ad->ppline != NULL) {
for (z = z1; z <= z2; z++) ad->ppline[z-z1] = ( (tr->st[z] == p7T_D) ? '.' : p7_alidisplay_EncodePostProb(tr->pp[z]));
ad->ppline[z-z1] = '\0';
}
/* mandatory three alignment display lines: model, mline, aseq */
for (z = z1; z <= z2; z++)
{
k = tr->k[z];
i = tr->i[z];
x = sq->dsq[i];
s = tr->st[z];
switch (s) {
case p7T_M:
ad->model[z-z1] = om->consensus[k];
if (x == esl_abc_DigitizeSymbol(om->abc, om->consensus[k])) ad->mline[z-z1] = ad->model[z-z1];
else if (p7_oprofile_FGetEmission(om, k, x) > 1.0) ad->mline[z-z1] = '+'; /* >1 not >0; om has odds ratios, not scores */
else ad->mline[z-z1] = ' ';
ad->aseq [z-z1] = toupper(Alphabet[x]);
break;
case p7T_I:
ad->model [z-z1] = '.';
ad->mline [z-z1] = ' ';
ad->aseq [z-z1] = tolower(Alphabet[x]);
break;
case p7T_D:
ad->model [z-z1] = om->consensus[k];
ad->mline [z-z1] = ' ';
ad->aseq [z-z1] = '-';
break;
default: ESL_XEXCEPTION(eslEINVAL, "invalid state in trace: not M,D,I");
}
}
ad->model [z2-z1+1] = '\0';
ad->mline [z2-z1+1] = '\0';
ad->aseq [z2-z1+1] = '\0';
ad->N = z2-z1+1;
return ad;
ERROR:
p7_alidisplay_Destroy(ad);
return NULL;
}
