void
MSAFileWrite(FILE *fp, MSA *msa, int outfmt, int do_oneline)
{
switch (outfmt) {
case MSAFILE_A2M: WriteA2M(fp, msa); break;
case MSAFILE_CLUSTAL: WriteClustal(fp, msa); break;
case MSAFILE_MSF: WriteMSF(fp, msa); break;
case MSAFILE_PHYLIP: WritePhylip(fp, msa); break;
case MSAFILE_SELEX: WriteSELEX(fp, msa); break;
case MSAFILE_STOCKHOLM:
if (do_oneline) WriteStockholmOneBlock(fp, msa);
else WriteStockholm(fp, msa);
break;
default:
Die("can't write. no such alignment format %d\n", outfmt);
}
}
/**
* @brief Write alignment to file.
*
* @param[in] mseq
* The mseq_t struct containing the aligned sequences
* @param[in] pcAlnOutfile
* The name of the output file
* @param[in] outfmt
* The alignment output format (defined in squid.h)
* @param[in] iWrap
* length of line for Clustal/Fasta format
*
* @return Non-zero on error
*
* @note We create a temporary squid MSA struct in here because we never
* use it within clustal. We might be better of using the old clustal
* output routines instead.
*
*/
int
WriteAlignment(mseq_t *mseq, const char *pcAlnOutfile, int outfmt, int iWrap, bool bResno)
{
int i; /* aux */
MSA *msa; /* squid's alignment structure */
FILE *pfOut = NULL;
int key; /* MSA struct internal index for sequence */
int alen; /* alignment length */
bool use_stdout;
assert(mseq!=NULL);
if (MSAFILE_UNKNOWN == outfmt) {
Log(&rLog, LOG_ERROR, "Unknown output format chosen");
return -1;
}
if (NULL == pcAlnOutfile) {
pfOut = stdout;
use_stdout = TRUE;
} else {
use_stdout = FALSE;
if (NULL == (pfOut = fopen(pcAlnOutfile, "w"))) {
Log(&rLog, LOG_ERROR, "Could not open file %s for writing", pcAlnOutfile);
return -1;
}
}
/* derive alignment length from first seq */
alen = strlen(mseq->seq[0]);
msa = MSAAlloc(mseq->nseqs, alen);
/* basic structure borrowed code from squid-1.9g/a2m.c:ReadA2M()
* we actually create a copy of mseq. keeping the pointers becomes
* messy when calling MSAFree()
*/
for (i=0; i<mseq->nseqs; i++) {
char *this_name = NULL; /* mseq sequence name */
char *this_seq = NULL; /* mseq sequence */
SQINFO *this_sqinfo = NULL; /* mseq sequence name */
int iI;
/* mseq->tree_order encodes to order in which sequences are listed in the guide-tree,
if the user wants the sequence output in the input-order then mseq->tree_order==NULL,
otherwise mseq->tree_order!=NULL, containing the indices of the sequences, FS, r274 -> */
iI = (NULL == mseq->tree_order) ? i : mseq->tree_order[i];
this_name = mseq->sqinfo[iI].name; /* mseq sequence name */
this_seq = mseq->seq[iI]; /* mseq sequence */
this_sqinfo = &mseq->sqinfo[iI]; /* mseq sequence name */
key = GKIStoreKey(msa->index, this_name);
msa->sqname[key] = sre_strdup(this_name, strlen(this_name));
/* setting msa->sqlen[idx] and msa->aseq[idx] */
msa->sqlen[key] = sre_strcat(&(msa->aseq[key]), msa->sqlen[key],
this_seq, strlen(this_seq));
if (this_sqinfo->flags & SQINFO_DESC) {
/* FIXME never get here ... */
MSASetSeqDescription(msa, key, this_sqinfo->desc);
}
/* FIXME extend this by copying more stuff according to flags.
* See MSAFileRead() in msa.c and used functions there
*
* Problem is that we never parse MSA information as we use squid'sSeqFile
*/
msa->nseq++;
} /* 0 <= i < mseq->nseqs */
/* FIXME Would like to, but can't use MSAVerifyParse(msa) here, as it
* will die on error. Need to implement our own version
*/
#if 0
MSAVerifyParse(msa);
#endif
/* The below is copy of MSAFileWrite() which originally only writes to stdout.
*/
/* Be sloppy and make a2m and fasta the same. same for vienna (which is
the same). same same. can can. boleh boleh */
if (outfmt==SQFILE_FASTA)
outfmt = MSAFILE_A2M;
if (outfmt==SQFILE_VIENNA)
outfmt = MSAFILE_VIENNA;
switch (outfmt) {
case MSAFILE_A2M:
/*WriteA2M(pfOut, msa, 0);*/
WriteA2M(pfOut, msa, iWrap);
break;
case MSAFILE_VIENNA:
/*WriteA2M(pfOut, msa, 1);*/
WriteA2M(pfOut, msa, INT_MAX);
break;
case MSAFILE_CLUSTAL:
WriteClustal(pfOut, msa, iWrap, TRUE==bResno ? 1 : 0, mseq->seqtype);
break;
case MSAFILE_MSF:
WriteMSF(pfOut, msa);
break;
case MSAFILE_PHYLIP:
WritePhylip(pfOut, msa);
break;
case MSAFILE_SELEX:
WriteSELEX(pfOut, msa);
break;
case MSAFILE_STOCKHOLM:
WriteStockholm(pfOut, msa);
break;
default:
Log(&rLog, LOG_FATAL, "internal error: %s",
"invalid output format should have been detected before");
}
if (use_stdout == FALSE) {
(void) fclose(pfOut);
Log(&rLog, LOG_INFO,
"Alignment written to %s", pcAlnOutfile);
}
MSAFree(msa);
return 0;
}
int main(int argc, char **argv)
{
const char *hmmfile; /* file to read HMMs from */
FILE *fp; /* output file handle */
HMMFILE *hmmfp; /* opened hmmfile for reading */
struct plan7_s *hmm; /* HMM to generate from */
int L; /* length of a sequence */
int i; /* counter over sequences */
char *ofile; /* output sequence file */
int nseq; /* number of seqs to sample */
int seed; /* random number generator seed */
int be_quiet; /* TRUE to silence header/footer */
int do_alignment; /* TRUE to output in aligned format */
int do_consensus; /* TRUE to do a single consensus seq */
AjBool ajselex;
AjBool ajcons;
AjPFile inf=NULL;
AjPFile outf=NULL;
AjPStr instr=NULL;
AjPStr outstr=NULL;
#ifdef MEMDEBUG
unsigned long histid1, histid2, orig_size, current_size;
orig_size = malloc_inuse(&histid1);
fprintf(stderr, "[... memory debugging is ON ...]\n");
#endif
/***********************************************
* Parse command line
***********************************************/
nseq = 10;
be_quiet = FALSE;
do_alignment = FALSE;
do_consensus = FALSE;
ofile = NULL;
embInitPV("ohmmemit",argc,argv,"HMMER",VERSION);
ajselex = ajAcdGetBoolean("selex");
ajcons = ajAcdGetBoolean("consensus");
nseq = ajAcdGetInt("number");
seed = ajAcdGetInt("seed");
inf = ajAcdGetInfile("infile");
outf = ajAcdGetOutfile("outfile");
if(!seed)
seed = time ((time_t *) NULL);
if(ajselex)
do_alignment=TRUE;
else
do_alignment=FALSE;
if(ajcons)
do_consensus=TRUE;
else
do_consensus=FALSE;
instr = ajStrNewC((char *)ajFileGetNameC(inf));
outstr = ajStrNewC((char *)ajFileGetNameC(outf));
hmmfile = ajStrGetPtr(instr);
sre_srandom(seed);
if (do_alignment && do_consensus)
ajFatal("Sorry, -selex and -consensus are incompatible.\n");
if (nseq != 10 && do_consensus)
ajWarn("-consensus overrides -number (# of sampled seqs)");
/***********************************************
* Open HMM file (might be in HMMERDB or current directory).
* Read a single HMM from it.
***********************************************/
if ((hmmfp = HMMFileOpen(hmmfile, "HMMERDB")) == NULL)
ajFatal("Failed to open HMM file %s\n", hmmfile);
if (!HMMFileRead(hmmfp, &hmm))
ajFatal("Failed to read any HMMs from %s\n", hmmfile);
HMMFileClose(hmmfp);
if (hmm == NULL)
ajFatal("HMM file %s corrupt or in incorrect format? Parse failed",
hmmfile);
/* Configure the HMM to shut off N,J,C emission: so we
* do a simple single pass through the model.
*/
Plan7NakedConfig(hmm);
Plan7Renormalize(hmm);
/***********************************************
* Open the output file, or stdout
***********************************************/
fp = ajFileGetFileptr(outf);
/***********************************************
* Show the options banner
***********************************************/
be_quiet=TRUE;
if (! be_quiet)
{
printf("HMM file: %s\n", hmmfile);
if (! do_consensus)
{
printf("Number of seqs: %d\n", nseq);
printf("Random seed: %d\n", seed);
}
printf("- - - - - - - - - - - - - - - - - - - - - - - - - "
"- - - - - - -\n\n");
}
/***********************************************
* Do the work.
* If we're generating an alignment, we have to collect
* all our traces, then output. If we're generating unaligned
* sequences, we can emit one at a time.
***********************************************/
if (do_consensus)
{
char *seq;
SQINFO sqinfo; /* info about sequence (name/desc) */
EmitConsensusSequence(hmm, &seq, NULL, &L, NULL);
strcpy(sqinfo.name, "consensus");
sqinfo.len = L;
sqinfo.flags = SQINFO_NAME | SQINFO_LEN;
WriteSeq(fp, kPearson, seq, &sqinfo);
free(seq);
}
else if (do_alignment)
{
struct p7trace_s **tr;
char **dsq;
SQINFO *sqinfo;
char **aseq;
AINFO ainfo;
float *wgt;
dsq = MallocOrDie(sizeof(char *) * nseq);
tr = MallocOrDie(sizeof(struct p7trace_s *) * nseq);
sqinfo = MallocOrDie(sizeof(SQINFO) * nseq);
wgt = MallocOrDie(sizeof(float) * nseq);
FSet(wgt, nseq, 1.0);
for (i = 0; i < nseq; i++)
{
EmitSequence(hmm, &(dsq[i]), &L, &(tr[i]));
sprintf(sqinfo[i].name, "seq%d", i+1);
sqinfo[i].len = L;
sqinfo[i].flags = SQINFO_NAME | SQINFO_LEN;
}
P7Traces2Alignment(dsq, sqinfo, wgt, nseq, hmm->M, tr, FALSE,
&aseq, &ainfo);
/* Output the alignment */
WriteSELEX(fp, aseq, &ainfo, 50);
if (ofile != NULL && !be_quiet)
printf("Alignment saved in file %s\n", ofile);
/* Free memory
*/
for (i = 0; i < nseq; i++)
{
P7FreeTrace(tr[i]);
free(dsq[i]);
}
FreeAlignment(aseq, &ainfo);
free(sqinfo);
free(dsq);
free(wgt);
free(tr);
}
else /* unaligned sequence output */
{
struct p7trace_s *tr;
char *dsq;
char *seq;
SQINFO sqinfo;
for (i = 0; i < nseq; i++)
{
EmitSequence(hmm, &dsq, &L, &tr);
sprintf(sqinfo.name, "seq%d", i+1);
sqinfo.len = L;
sqinfo.flags = SQINFO_NAME | SQINFO_LEN;
seq = DedigitizeSequence(dsq, L);
WriteSeq(fp, kPearson, seq, &sqinfo);
P7FreeTrace(tr);
free(dsq);
free(seq);
}
}
ajFileClose(&outf);
FreePlan7(hmm);
SqdClean();
#ifdef MEMDEBUG
current_size = malloc_inuse(&histid2);
if (current_size != orig_size)
malloc_list(2, histid1, histid2);
else
fprintf(stderr, "[No memory leaks.]\n");
#endif
ajStrDel(&instr);
ajStrDel(&outstr);
ajFileClose(&inf);
ajFileClose(&outf);
embExit();
return 0;
}
int main(int argc, char **argv)
{
const char *hmmfile; /* file to read HMMs from */
HMMFILE *hmmfp; /* opened hmmfile for reading */
const char *seqfile; /* file to read target sequence from */
char **rseq; /* raw, unaligned sequences */
SQINFO *sqinfo; /* info associated with sequences */
char **dsq; /* digitized raw sequences */
int nseq; /* number of sequences */
char **aseq; /* aligned sequences */
AINFO ainfo; /* alignment information */
float *wgt; /* per-sequence weights */
int i;
struct plan7_s *hmm; /* HMM to align to */
struct p7trace_s **tr; /* traces for aligned sequences */
int be_quiet; /* TRUE to suppress verbose banner */
int matchonly; /* TRUE to show only match state syms */
const char *outfile; /* optional alignment output file */
FILE *ofp; /* handle on alignment output file */
AjPFile ajwithali; /* name of additional alignment file to align */
AjPFile ajmapali; /* name of additional alignment file to map */
AjBool ajmatch=ajFalse;
AjPFile outf=NULL;
AjPStr outfname=NULL;
AjPFile inf=NULL;
AjPStr infname=NULL;
AjPSeqset seqset=NULL;
AjPStr ajseqfile=NULL;
char* mapali=NULL;
char* withali=NULL;
#ifdef MEMDEBUG
unsigned long histid1, histid2, orig_size, current_size;
orig_size = malloc_inuse(&histid1);
fprintf(stderr, "[... memory debugging is ON ...]\n");
#endif
/***********************************************
* Parse command line
***********************************************/
matchonly = FALSE;
outfile = NULL;
be_quiet = FALSE;
withali = NULL;
mapali = NULL;
embInitPV("ohmmalign",argc,argv,"HMMER",VERSION);
ajmatch = ajAcdGetBoolean("matchonly");
if(ajmatch)
matchonly=TRUE;
else
matchonly=FALSE;
ajmapali = ajAcdGetInfile("mapalifile");
if (ajmapali)
mapali = ajCharNewS(ajFileGetNameS(ajmapali));
ajFileClose(&ajmapali);
ajwithali = ajAcdGetInfile("withalifile");
if (ajwithali)
withali = ajCharNewS(ajFileGetNameS(ajwithali));
ajFileClose(&ajwithali);
be_quiet=TRUE;
outf = ajAcdGetOutfile("outfile");
outfname = ajStrNewC((char *)ajFileGetNameC(outf));
if(*ajStrGetPtr(outfname)>31)
ajFileClose(&outf);
outfile = ajStrGetPtr(outfname);
inf = ajAcdGetInfile("hmmfile");
infname = ajStrNewC((char *)ajFileGetNameC(inf));
ajFileClose(&inf);
hmmfile = ajStrGetPtr(infname);
seqset = ajAcdGetSeqset("sequences");
ajseqfile = ajStrNewC(ajStrGetPtr(seqset->Filename));
seqfile = ajStrGetPtr(ajseqfile);
/***********************************************
* Open HMM file (might be in HMMERDB or current directory).
* Read a single HMM from it.
*
* Currently hmmalign disallows the J state and
* only allows one domain per sequence. To preserve
* the S/W entry information, the J state is explicitly
* disallowed, rather than calling a Plan7*Config() function.
* this is a workaround in 2.1 for the 2.0.x "yo!" bug.
***********************************************/
if ((hmmfp = HMMFileOpen(hmmfile, "HMMERDB")) == NULL)
ajFatal("Failed to open HMM file %s\n", hmmfile);
if (!HMMFileRead(hmmfp, &hmm))
ajFatal("Failed to read any HMMs from %s\n", hmmfile);
HMMFileClose(hmmfp);
if (hmm == NULL)
ajFatal("HMM file %s corrupt or in incorrect format? Parse failed", hmmfile);
hmm->xt[XTE][MOVE] = 1.; /* only 1 domain/sequence ("global" alignment) */
hmm->xt[XTE][LOOP] = 0.;
P7Logoddsify(hmm, TRUE);
/* do we have the map we might need? */
if (mapali != NULL && ! (hmm->flags & PLAN7_MAP))
ajFatal("HMMER: HMM file %s has no map; you can't use --mapali.", hmmfile);
/***********************************************
* Open sequence file in current directory.
* Read all seqs from it.
***********************************************/
/*
if (! SeqfileFormat(seqfile, &format, NULL))
switch (squid_errno) {
case SQERR_NOFILE:
ajFatal("Sequence file %s could not be opened for reading", seqfile);
case SQERR_FORMAT:
default:
ajFatal("Failed to determine format of sequence file %s", seqfile);
}
if (! ReadMultipleRseqs(seqfile, format, &rseq, &sqinfo, &nseq))
ajFatal("Failed to read any sequences from file %s", seqfile);
*/
emboss_rseqs(seqset,&rseq,&sqinfo,&nseq);
/***********************************************
* Show the banner
***********************************************/
be_quiet=TRUE;
if (! be_quiet)
{
/* Banner(stdout, banner); */
printf( "HMM file: %s\n", hmmfile);
printf( "Sequence file: %s\n", seqfile);
printf("- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -\n\n");
}
/***********************************************
* Do the work
***********************************************/
/* Allocations and initializations.
*/
dsq = MallocOrDie(sizeof(char *) * nseq);
tr = MallocOrDie(sizeof(struct p7trace_s *) * nseq);
/* Align each sequence to the model, collect traces
*/
for (i = 0; i < nseq; i++)
{
dsq[i] = DigitizeSequence(rseq[i], sqinfo[i].len);
if (P7ViterbiSize(sqinfo[i].len, hmm->M) <= RAMLIMIT)
(void) P7Viterbi(dsq[i], sqinfo[i].len, hmm, &(tr[i]));
else
(void) P7SmallViterbi(dsq[i], sqinfo[i].len, hmm, &(tr[i]));
}
/* Include an aligned alignment, if desired.
*/
if (mapali != NULL)
include_alignment(mapali, hmm, TRUE, &rseq, &dsq, &sqinfo, &tr, &nseq);
if (withali != NULL)
include_alignment(withali, hmm, FALSE, &rseq, &dsq, &sqinfo, &tr, &nseq);
/* Turn traces into a multiple alignment
*/
wgt = MallocOrDie(sizeof(float) * nseq);
FSet(wgt, nseq, 1.0);
P7Traces2Alignment(dsq, sqinfo, wgt, nseq, hmm->M, tr, matchonly,
&aseq, &ainfo);
/***********************************************
* Output the alignment
***********************************************/
if (outfile != NULL && (ofp = fopen(outfile, "w")) != NULL)
{
WriteSELEX(ofp, aseq, &ainfo, 50);
printf("Alignment saved in file %s\n", outfile);
fclose(ofp);
}
else
WriteSELEX(stdout, aseq, &ainfo, 50);
/***********************************************
* Cleanup and exit
***********************************************/
for (i = 0; i < nseq; i++)
{
P7FreeTrace(tr[i]);
FreeSequence(rseq[i], &(sqinfo[i]));
free(dsq[i]);
}
FreeAlignment(aseq, &ainfo);
FreePlan7(hmm);
free(sqinfo);
free(rseq);
free(dsq);
free(wgt);
free(tr);
SqdClean();
ajStrDel(&outfname);
ajStrDel(&infname);
ajStrDel(&ajseqfile);
#ifdef MEMDEBUG
current_size = malloc_inuse(&histid2);
if (current_size != orig_size) malloc_list(2, histid1, histid2);
else fprintf(stderr, "[No memory leaks.]\n");
#endif
ajSeqsetDel(&seqset);
ajFileClose(&ajwithali);
ajFileClose(&ajmapali);
embExit();
return 0;
}
