int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_CreateDefaultApp(options, 1, argc, argv, banner, usage);
char *hmmfile = esl_opt_GetArg(go, 1);
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_GMX *gx1 = NULL;
P7_GMX *gx2 = NULL;
int L = esl_opt_GetInteger(go, "-L");
int N = esl_opt_GetInteger(go, "-N");
ESL_DSQ *dsq = malloc(sizeof(ESL_DSQ) * (L+2));
float null2[p7_MAXCODE];
int i;
float fsc, bsc;
double Mcs;
if (p7_hmmfile_OpenE(hmmfile, NULL, &hfp, NULL) != eslOK) p7_Fail("Failed to open HMM file %s", hmmfile);
if (p7_hmmfile_Read(hfp, &abc, &hmm) != eslOK) p7_Fail("Failed to read HMM");
bg = p7_bg_Create(abc);
p7_bg_SetLength(bg, L);
gm = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, L, p7_LOCAL);
gx1 = p7_gmx_Create(gm->M, L);
gx2 = p7_gmx_Create(gm->M, L);
esl_rsq_xfIID(r, bg->f, abc->K, L, dsq);
p7_GForward (dsq, L, gm, gx1, &fsc);
p7_GBackward(dsq, L, gm, gx2, &bsc);
p7_GDecoding(gm, gx1, gx2, gx2);
esl_stopwatch_Start(w);
for (i = 0; i < N; i++)
p7_GNull2_ByExpectation(gm, gx2, null2);
esl_stopwatch_Stop(w);
Mcs = (double) N * (double) L * (double) gm->M * 1e-6 / w->user;
esl_stopwatch_Display(stdout, w, "# CPU time: ");
printf("# M = %d\n", gm->M);
printf("# %.1f Mc/s\n", Mcs);
free(dsq);
p7_gmx_Destroy(gx1);
p7_gmx_Destroy(gx2);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
p7_hmmfile_Close(hfp);
esl_alphabet_Destroy(abc);
esl_stopwatch_Destroy(w);
esl_randomness_Destroy(r);
esl_getopts_Destroy(go);
return 0;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = esl_getopts_CreateDefaultApp(options, 1, argc, argv, banner, usage);
ESL_STOPWATCH *w = esl_stopwatch_Create();
char *hmmfile = esl_opt_GetArg(go, 1);
ESL_ALPHABET *abc = NULL;
P7_HMMFILE *hfp = NULL;
P7_HMM *hmm = NULL;
P7_BG *bg = NULL;
int L = esl_opt_GetInteger(go, "-L");
int N = esl_opt_GetInteger(go, "-N");
int i;
/* Read one HMM from <hmmfile> */
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);
esl_stopwatch_Start(w);
for (i = 0; i < N; i++)
p7_bg_SetFilterByHMM(bg, hmm);
esl_stopwatch_Stop(w);
esl_stopwatch_Display(stdout, w, "# CPU time: ");
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
esl_alphabet_Destroy(abc);
esl_stopwatch_Destroy(w);
esl_getopts_Destroy(go);
return 0;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_CreateDefaultApp(options, 1, argc, argv, banner, usage);
ESL_STOPWATCH *w = esl_stopwatch_Create();
char *hmmfile = esl_opt_GetArg(go, 1);
P7_HMMCACHE *hcache = NULL;
char errbuf[eslERRBUFSIZE];
size_t tot_mem;
int status;
esl_stopwatch_Start(w);
status = p7_hmmcache_Open(hmmfile, &hcache, errbuf);
if (status == eslENOTFOUND) p7_Fail("Failed to read %s\n %s\n", hmmfile, errbuf);
else if (status == eslEFORMAT) p7_Fail("Failed to parse %s\n %s\n", hmmfile, errbuf);
else if (status == eslEINCOMPAT) p7_Fail("Mixed profile types in %s\n %s\n", hmmfile, errbuf);
else if (status != eslOK) p7_Fail("Failed to cache %s: error code %d\n", hmmfile, status);
p7_hmmcache_SetNumericNames(hcache);
tot_mem = p7_hmmcache_Sizeof(hcache);
esl_stopwatch_Stop(w);
esl_stopwatch_Display(stdout, w, "# CPU time: ");
printf("models = %d\n", hcache->n);
printf("tot memory = %" PRIu64 "\n", (uint64_t) tot_mem);
p7_hmmcache_Close(hcache);
esl_getopts_Destroy(go);
esl_stopwatch_Destroy(w);
return 0;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_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_ALPHABET *abc = NULL;
P7_HMMFILE *hfp = NULL;
P7_HMM *hmm = NULL;
if (p7_hmmfile_OpenE(hmmfile, NULL, &hfp, NULL) != 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);
esl_stopwatch_Start(w);
while (N--)
{ /* cfg rng bg gm om */
p7_Calibrate(hmm, NULL, NULL, NULL, NULL, NULL);
}
esl_stopwatch_Stop(w);
esl_stopwatch_Display(stdout, w, "# CPU time: ");
p7_hmm_Destroy(hmm);
esl_alphabet_Destroy(abc);
esl_stopwatch_Destroy(w);
esl_getopts_Destroy(go);
return 0;
}
int
main(int argc, char **argv)
{
char *hmmfile = argv[1]; /* name of HMM file to read one HMM from */
ESL_ALPHABET *abc = NULL; /* sequence alphabet */
ESL_RANDOMNESS *r = NULL; /* source of randomness */
P7_HMMFILE *hfp = NULL; /* open hmmfile */
P7_HMM *hmm = NULL; /* HMM to emit from */
P7_PROFILE *gm = NULL; /* profile HMM (scores) */
P7_BG *bg = NULL; /* null model */
P7_TRACE *tr = NULL; /* sampled trace */
ESL_SQ *sq = NULL; /* sampled digital sequence */
int n = 1000;
int counts[p7T_NSTATETYPES];
int i;
float sc;
float nullsc;
double bitscore;
r = esl_randomness_CreateFast(0);
tr = p7_trace_Create();
if (p7_hmmfile_OpenE(hmmfile, NULL, &hfp, NULL) != eslOK) p7_Fail("failed to open %s", hmmfile);
if (p7_hmmfile_Read(hfp, &abc, &hmm) != eslOK) p7_Fail("failed to read HMM");
sq = esl_sq_CreateDigital(abc);
bg = p7_bg_Create(abc);
gm = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, sq->n, p7_LOCAL);
for (i = 0; i < n; i++)
{
p7_ProfileEmit(r, hmm, gm, bg, sq, tr);
p7_trace_GetStateUseCounts(tr, counts);
p7_ReconfigLength(gm, sq->n);
p7_bg_SetLength(bg, sq->n);
p7_trace_Score(tr, sq->dsq, gm, &sc);
p7_bg_NullOne (bg, sq->dsq, sq->n, &nullsc);
bitscore = (sc - nullsc)/ eslCONST_LOG2;
printf("%d %8.4f\n",
counts[p7T_M] + (counts[p7T_I] + counts[p7T_D])/2,
bitscore);
}
p7_profile_Destroy(gm);
esl_sq_Destroy(sq);
p7_trace_Destroy(tr);
esl_randomness_Destroy(r);
esl_alphabet_Destroy(abc);
p7_hmmfile_Close(hfp);
p7_hmm_Destroy(hmm);
return eslOK;
}
/* Create an SSI index file for open HMM file <hfp>.
* Both name and accession of HMMs are stored as keys.
*/
static void
create_ssi_index(ESL_GETOPTS *go, P7_HMMFILE *hfp)
{
ESL_NEWSSI *ns = NULL;
ESL_ALPHABET *abc = NULL;
P7_HMM *hmm = NULL;
int nhmm = 0;
char *ssifile = NULL;
uint16_t fh;
int status;
if (esl_sprintf(&ssifile, "%s.ssi", hfp->fname) != eslOK) p7_Die("esl_sprintf() failed");
status = esl_newssi_Open(ssifile, FALSE, &ns);
if (status == eslENOTFOUND) esl_fatal("failed to open SSI index %s", ssifile);
else if (status == eslEOVERWRITE) esl_fatal("SSI index %s already exists; delete or rename it", ssifile);
else if (status != eslOK) esl_fatal("failed to create a new SSI index");
if (esl_newssi_AddFile(ns, hfp->fname, 0, &fh) != eslOK) /* 0 = format code (HMMs don't have any yet) */
esl_fatal("Failed to add HMM file %s to new SSI index\n", hfp->fname);
printf("Working... ");
fflush(stdout);
while ((status = p7_hmmfile_Read(hfp, &abc, &hmm)) != eslEOF)
{
if (status == eslEOD) p7_Fail("read failed, HMM file %s may be truncated?", hfp->fname);
else if (status == eslEFORMAT) p7_Fail("bad file format in HMM file %s", hfp->fname);
else if (status == eslEINCOMPAT) p7_Fail("HMM file %s contains different alphabets", hfp->fname);
else if (status != eslOK) p7_Fail("Unexpected error in reading HMMs from %s", hfp->fname);
nhmm++;
if (hmm->name == NULL) p7_Fail("Every HMM must have a name to be indexed. Failed to find name of HMM #%d\n", nhmm);
if (esl_newssi_AddKey(ns, hmm->name, fh, hmm->offset, 0, 0) != eslOK)
p7_Fail("Failed to add key %s to SSI index", hmm->name);
if (hmm->acc) {
if (esl_newssi_AddAlias(ns, hmm->acc, hmm->name) != eslOK)
p7_Fail("Failed to add secondary key %s to SSI index", hmm->acc);
}
p7_hmm_Destroy(hmm);
}
if (esl_newssi_Write(ns) != eslOK)
p7_Fail("Failed to write keys to ssi file %s\n", ssifile);
printf("done.\n");
if (ns->nsecondary > 0)
printf("Indexed %d HMMs (%ld names and %ld accessions).\n", nhmm, (long) ns->nprimary, (long) ns->nsecondary);
else
printf("Indexed %d HMMs (%ld names).\n", nhmm, (long) ns->nprimary);
printf("SSI index written to file %s\n", ssifile);
free(ssifile);
esl_alphabet_Destroy(abc);
esl_newssi_Close(ns);
return;
}
static int
serial_loop(WORKER_INFO *info, struct cfg_s *cfg)
{
P7_BUILDER *bld = NULL;
ESL_MSA *msa = NULL;
ESL_MSA *postmsa = NULL;
ESL_MSA **postmsa_ptr = (cfg->postmsafile != NULL) ? &postmsa : NULL;
P7_HMM *hmm = NULL;
char errmsg[eslERRBUFSIZE];
int status;
double entropy;
cfg->nali = 0;
while ((status = esl_msa_Read(cfg->afp, &msa)) == eslOK)
{
cfg->nali++;
if ((status = set_msa_name(cfg, errmsg, msa)) != eslOK) p7_Fail("%s\n", errmsg); /* cfg->nnamed gets incremented in this call */
/* bg new-HMM trarr gm om */
if ((status = p7_Builder(info->bld, msa, info->bg, &hmm, NULL, NULL, NULL, postmsa_ptr)) != eslOK) p7_Fail("build failed: %s", bld->errbuf);
entropy = p7_MeanMatchRelativeEntropy(hmm, info->bg);
if ((status = output_result(cfg, errmsg, cfg->nali, msa, hmm, postmsa, entropy)) != eslOK) p7_Fail(errmsg);
p7_hmm_Destroy(hmm);
esl_msa_Destroy(msa);
esl_msa_Destroy(postmsa);
}
return status;
}
/* mpi_worker()
* The main control for an MPI worker process.
*/
static void
mpi_worker(ESL_GETOPTS *go, struct cfg_s *cfg)
{
int xstatus = eslOK;
int status;
P7_HMM *hmm = NULL;
char *wbuf = NULL;
double *xv = NULL; /* result: array of N scores */
int *av = NULL; /* optional result: array of N alignment lengths */
int wn = 0;
char errbuf[eslERRBUFSIZE];
int pos;
/* Worker initializes */
if ((status = minimum_mpi_working_buffer(go, cfg->N, &wn)) != eslOK) xstatus = status;
ESL_ALLOC(wbuf, wn * sizeof(char));
ESL_ALLOC(xv, cfg->N * sizeof(double) + 2);
if (esl_opt_GetBoolean(go, "-a"))
ESL_ALLOC(av, cfg->N * sizeof(int));
/* Main worker loop */
while (p7_hmm_mpi_Recv(0, 0, MPI_COMM_WORLD, &wbuf, &wn, &(cfg->abc), &hmm) == eslOK)
{
if (esl_opt_GetBoolean(go, "--recal")) {
if (( status = recalibrate_model(go, cfg, errbuf, hmm)) != eslOK) goto CLEANERROR;
}
if ((status = process_workunit(go, cfg, errbuf, hmm, xv, av)) != eslOK) goto CLEANERROR;
pos = 0;
MPI_Pack(&status, 1, MPI_INT, wbuf, wn, &pos, MPI_COMM_WORLD);
MPI_Pack(xv, cfg->N, MPI_DOUBLE, wbuf, wn, &pos, MPI_COMM_WORLD);
if (esl_opt_GetBoolean(go, "-a"))
MPI_Pack(av, cfg->N, MPI_INT, wbuf, wn, &pos, MPI_COMM_WORLD);
MPI_Send(wbuf, pos, MPI_PACKED, 0, 0, MPI_COMM_WORLD);
p7_hmm_Destroy(hmm);
}
free(wbuf);
free(xv);
if (av != NULL) free(av);
return;
CLEANERROR:
pos = 0;
MPI_Pack(&status, 1, MPI_INT, wbuf, wn, &pos, MPI_COMM_WORLD);
MPI_Pack(errbuf, eslERRBUFSIZE, MPI_CHAR, wbuf, wn, &pos, MPI_COMM_WORLD);
MPI_Send(wbuf, pos, MPI_PACKED, 0, 0, MPI_COMM_WORLD);
if (wbuf != NULL) free(wbuf);
if (hmm != NULL) p7_hmm_Destroy(hmm);
if (xv != NULL) free(xv);
if (av != NULL) free(av);
return;
ERROR:
p7_Fail("Allocation error in mpi_worker");
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_CreateDefaultApp(options, 1, argc, argv, banner, usage);
ESL_RANDOMNESS *rng = esl_randomness_CreateFast(esl_opt_GetInteger(go, "-s"));
char *hmmfile = esl_opt_GetArg(go, 1);
int L = esl_opt_GetInteger(go, "-L");
int N = esl_opt_GetInteger(go, "-N");
ESL_ALPHABET *abc = NULL;
P7_HMMFILE *hfp = NULL;
P7_HMM *hmm = NULL;
P7_BG *bg = NULL;
P7_PROFILE *gm = NULL;
P7_TRACE *tr = p7_trace_Create();
ESL_SQ *sq = NULL;
char errbuf[eslERRBUFSIZE];
int i;
int status;
status = p7_hmmfile_OpenE(hmmfile, NULL, &hfp, errbuf);
if (status == eslENOTFOUND) p7_Fail("File existence/permissions problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf);
else if (status == eslEFORMAT) p7_Fail("File format problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf);
else if (status != eslOK) p7_Fail("Unexpected error %d in opening HMM file %s.\n%s\n", status, hmmfile, errbuf);
status = p7_hmmfile_Read(hfp, &abc, &hmm);
if (status == eslEFORMAT) p7_Fail("Bad file format in HMM file %s:\n%s\n", hfp->fname, hfp->errbuf);
else if (status == eslEINCOMPAT) p7_Fail("HMM in %s is not in the expected %s alphabet\n", hfp->fname, esl_abc_DecodeType(abc->type));
else if (status == eslEOF) p7_Fail("Empty HMM file %s? No HMM data found.\n", hfp->fname);
else if (status != eslOK) p7_Fail("Unexpected error in reading HMMs from %s\n", hfp->fname);
p7_hmmfile_Close(hfp);
bg = p7_bg_Create(abc); p7_bg_SetLength(bg, L);
gm = p7_profile_Create(hmm->M, abc); p7_ProfileConfig(hmm, bg, gm, L, p7_LOCAL);
sq = esl_sq_CreateDigital(abc);
for (i = 0; i < N; i++)
{
p7_ProfileEmit(rng, hmm, gm, bg, sq, tr);
esl_sq_FormatName(sq, "%s-sample%d", hmm->name, i);
esl_sqio_Write(stdout, sq, eslSQFILE_FASTA, FALSE);
if (p7_trace_Validate(tr, abc, sq->dsq, errbuf) != eslOK) esl_fatal(errbuf);
esl_sq_Reuse(sq);
p7_trace_Reuse(tr);
}
esl_sq_Destroy(sq);
p7_trace_Destroy(tr);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
esl_alphabet_Destroy(abc);
esl_randomness_Destroy(rng);
esl_getopts_Destroy(go);
return 0;
}
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);
ESL_STOPWATCH *w = esl_stopwatch_Create();
ESL_ALPHABET *abc = NULL;
P7_HMMFILE *hfp = NULL;
P7_HMM *hmm = NULL;
P7_BG *bg = NULL;
P7_PROFILE *gm = NULL;
P7_OPROFILE *om = NULL;
int L = esl_opt_GetInteger(go, "-L");
int N = esl_opt_GetInteger(go, "-N");
int i;
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");
bg = p7_bg_Create(abc);
p7_bg_SetLength(bg, L);
gm = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, L, p7_LOCAL);
om = p7_oprofile_Create(gm->M, abc);
esl_stopwatch_Start(w);
for (i = 0; i < N; i++)
p7_oprofile_Convert(gm, om);
esl_stopwatch_Stop(w);
esl_stopwatch_Display(stdout, w, "# CPU time: ");
printf("# M = %d\n", gm->M);
p7_oprofile_Destroy(om);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
p7_hmmfile_Close(hfp);
esl_alphabet_Destroy(abc);
esl_stopwatch_Destroy(w);
esl_getopts_Destroy(go);
return 0;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = NULL;
struct cfg_s cfg;
int status = eslOK;
impl_Init(); /* processor specific initialization */
p7_FLogsumInit(); /* we're going to use table-driven Logsum() approximations at times */
/* Initialize what we can in the config structure (without knowing the alphabet yet)
*/
cfg.hmmfile = NULL;
cfg.dbfile = NULL;
cfg.do_mpi = FALSE; /* this gets reset below, if we init MPI */
cfg.nproc = 0; /* this gets reset below, if we init MPI */
cfg.my_rank = 0; /* this gets reset below, if we init MPI */
cfg.firstseq_key = NULL;
cfg.n_targetseq = -1;
process_commandline(argc, argv, &go, &cfg.hmmfile, &cfg.dbfile);
/* is the range restricted? */
if (esl_opt_IsUsed(go, "--restrictdb_stkey") )
if ((cfg.firstseq_key = esl_opt_GetString(go, "--restrictdb_stkey")) == NULL) p7_Fail("Failure capturing --restrictdb_stkey\n");
if (esl_opt_IsUsed(go, "--restrictdb_n") )
cfg.n_targetseq = esl_opt_GetInteger(go, "--restrictdb_n");
if ( cfg.n_targetseq != -1 && cfg.n_targetseq < 1 )
p7_Fail("--restrictdb_n must be >= 1\n");
{
status = serial_master(go, &cfg);
}
esl_getopts_Destroy(go);
return status;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_CreateDefaultApp(options, 1, argc, argv, banner, usage);
char *hmmfile = esl_opt_GetArg(go, 1);
ESL_ALPHABET *abc = NULL;
P7_HMMFILE *hfp = NULL;
P7_HMM *hmm = NULL;
P7_BG *bg = NULL;
P7_PROFILE *gm = NULL;
float ftol = 1e-4; /* floating-point tolerance for checking parameters against expected probs or summing to 1 */
char errbuf[eslERRBUFSIZE];
/* Read in one HMM; sets alphabet to the HMM's alphabet */
if (p7_hmmfile_OpenE(hmmfile, NULL, &hfp, NULL) != 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);
/* Set up a null model */
bg = p7_bg_Create(abc);
/* Allocate and configure a profile from HMM and null model */
gm = p7_profile_Create(hmm->M, abc);
p7_profile_Config(gm, hmm, bg);
p7_profile_SetLength(gm, 400); /* 400 is arbitrary here; this is whatever your target seq length L is */
printf("profile memory consumed: %" PRId64 " bytes\n", (int64_t) p7_profile_Sizeof(gm));
/* Debugging tools allow dumping, validating the object */
if (p7_profile_Validate(gm, errbuf, ftol) != eslOK) p7_Fail("profile validation failed\n %s\n", errbuf);
if (esl_opt_GetBoolean(go, "--vv"))
p7_profile_Dump(stdout, gm);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
esl_alphabet_Destroy(abc);
esl_getopts_Destroy(go);
return 0;
}
static void
serial_master(ESL_GETOPTS *go, struct cfg_s *cfg)
{
P7_HMM *hmm = NULL;
double *xv = NULL; /* results: array of N scores */
int *av = NULL; /* optional results: array of N alignment lengths */
char errbuf[eslERRBUFSIZE];
int status;
if ((status = init_master_cfg(go, cfg, errbuf)) != eslOK) p7_Fail(errbuf);
if ((xv = malloc(sizeof(double) * cfg->N)) == NULL) p7_Fail("allocation failed");
if (esl_opt_GetBoolean(go, "-a") &&
(av = malloc(sizeof(int) * cfg->N)) == NULL) p7_Fail("allocation failed");
while ((status = p7_hmmfile_Read(cfg->hfp, &(cfg->abc), &hmm)) != eslEOF)
{
if (status == eslEOD) p7_Fail("read failed, HMM file %s may be truncated?", cfg->hmmfile);
else if (status == eslEFORMAT) p7_Fail("bad file format in HMM file %s", cfg->hmmfile);
else if (status == eslEINCOMPAT) p7_Fail("HMM file %s contains different alphabets", cfg->hmmfile);
else if (status != eslOK) p7_Fail("Unexpected error in reading HMMs from %s", cfg->hmmfile);
if (cfg->bg == NULL) {
if (esl_opt_GetBoolean(go, "--bgflat")) cfg->bg = p7_bg_CreateUniform(cfg->abc);
else cfg->bg = p7_bg_Create(cfg->abc);
p7_bg_SetLength(cfg->bg, esl_opt_GetInteger(go, "-L")); /* set the null model background length in both master and workers. */
}
if (esl_opt_GetBoolean(go, "--recal")) {
if (recalibrate_model(go, cfg, errbuf, hmm) != eslOK) p7_Fail(errbuf);
}
if (process_workunit(go, cfg, errbuf, hmm, xv, av) != eslOK) p7_Fail(errbuf);
if (output_result (go, cfg, errbuf, hmm, xv, av) != eslOK) p7_Fail(errbuf);
p7_hmm_Destroy(hmm);
}
free(xv);
if (av != NULL) free(av);
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = esl_getopts_CreateDefaultApp(options, 1, argc, argv, banner, usage);
ESL_STOPWATCH *w = esl_stopwatch_Create();
ESL_ALPHABET *abc = NULL;
char *msvfile = esl_opt_GetArg(go, 1);
FILE *msvfp = NULL;
P7_OPROFILE *om = NULL;
int nmodel = 0;
uint64_t totM = 0;
int status;
esl_stopwatch_Start(w);
if ((msvfp = fopen(msvfile, "r")) == NULL) p7_Fail("Failed to open MSV file %s for reading.\n", msvfile);
while ((status = p7_oprofile_ReadMSV(msvfp, &abc, NULL, &om)) == eslOK)
{
nmodel++;
totM += om->M;
p7_oprofile_Destroy(om);
}
if (status == eslEFORMAT) p7_Fail("bad file format in profile file %s", msvfile);
else if (status == eslEINCOMPAT) p7_Fail("profile file %s contains different alphabets", msvfile);
else if (status != eslEOF) p7_Fail("Unexpected error in reading profiles from %s", msvfile);
esl_stopwatch_Stop(w);
esl_stopwatch_Display(stdout, w, "# CPU time: ");
printf("# number of models: %d\n", nmodel);
printf("# total M: %" PRId64 "\n", totM);
fclose(msvfp);
esl_alphabet_Destroy(abc);
esl_stopwatch_Destroy(w);
esl_getopts_Destroy(go);
return 0;
}
/* multifetch:
* given a file containing lines with one name or key per line;
* parse the file line-by-line;
* if we have an SSI index available, retrieve the HMMs by key
* as we see each line;
* else, without an SSI index, store the keys in a hash, then
* read the entire HMM file in a single pass, outputting HMMs
* that are in our keylist.
*
* Note that with an SSI index, you get the HMMs in the order they
* appear in the <keyfile>, but without an SSI index, you get HMMs in
* the order they occur in the HMM file.
*/
static void
multifetch(ESL_GETOPTS *go, FILE *ofp, char *keyfile, P7_HMMFILE *hfp)
{
ESL_KEYHASH *keys = esl_keyhash_Create();
ESL_FILEPARSER *efp = NULL;
ESL_ALPHABET *abc = NULL;
P7_HMM *hmm = NULL;
int nhmm = 0;
char *key;
int keylen;
int keyidx;
int status;
if (esl_fileparser_Open(keyfile, NULL, &efp) != eslOK) p7_Fail("Failed to open key file %s\n", keyfile);
esl_fileparser_SetCommentChar(efp, '#');
while (esl_fileparser_NextLine(efp) == eslOK)
{
if (esl_fileparser_GetTokenOnLine(efp, &key, &keylen) != eslOK)
p7_Fail("Failed to read HMM name on line %d of file %s\n", efp->linenumber, keyfile);
status = esl_key_Store(keys, key, &keyidx);
if (status == eslEDUP) p7_Fail("HMM key %s occurs more than once in file %s\n", key, keyfile);
if (hfp->ssi != NULL) { onefetch(go, ofp, key, hfp); nhmm++; }
}
if (hfp->ssi == NULL)
{
while ((status = p7_hmmfile_Read(hfp, &abc, &hmm)) != eslEOF)
{
if (status == eslEOD) p7_Fail("read failed, HMM file %s may be truncated?", hfp->fname);
else if (status == eslEFORMAT) p7_Fail("bad file format in HMM file %s", hfp->fname);
else if (status == eslEINCOMPAT) p7_Fail("HMM file %s contains different alphabets", hfp->fname);
else if (status != eslOK) p7_Fail("Unexpected error in reading HMMs from %s", hfp->fname);
if (esl_key_Lookup(keys, hmm->name, &keyidx) == eslOK ||
((hmm->acc) && esl_key_Lookup(keys, hmm->acc, &keyidx) == eslOK))
{
p7_hmmfile_WriteASCII(ofp, -1, hmm);
nhmm++;
}
p7_hmm_Destroy(hmm);
}
}
if (ofp != stdout) printf("\nRetrieved %d HMMs.\n", nhmm);
if (abc != NULL) esl_alphabet_Destroy(abc);
esl_keyhash_Destroy(keys);
esl_fileparser_Close(efp);
return;
}
int
main(int argc, char **argv)
{
char *hmmfile = argv[1];
ESL_ALPHABET *abc = NULL;
P7_HMMFILE *hfp = NULL;
P7_HMM *hmm = NULL;
P7_BG *bg = NULL;
P7_PROFILE *gm = NULL;
P7_OPROFILE *om1 = NULL;
P7_OPROFILE *om2 = NULL;
int status;
char errbuf[eslERRBUFSIZE];
status = p7_hmmfile_OpenE(hmmfile, NULL, &hfp, errbuf);
if (status == eslENOTFOUND) p7_Fail("File existence/permissions problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf);
else if (status == eslEFORMAT) p7_Fail("File format problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf);
else if (status != eslOK) p7_Fail("Unexpected error %d in opening HMM file %s.\n%s\n", status, hmmfile, errbuf);
status = p7_hmmfile_Read(hfp, &abc, &hmm);
if (status == eslEFORMAT) p7_Fail("Bad file format in HMM file %s:\n%s\n", hfp->fname, hfp->errbuf);
else if (status == eslEINCOMPAT) p7_Fail("HMM in %s is not in the expected %s alphabet\n", hfp->fname, esl_abc_DecodeType(abc->type));
else if (status == eslEOF) p7_Fail("Empty HMM file %s? No HMM data found.\n", hfp->fname);
else if (status != eslOK) p7_Fail("Unexpected error in reading HMMs from %s\n", hfp->fname);
bg = p7_bg_Create(abc);
gm = p7_profile_Create(hmm->M, abc);
om1 = p7_oprofile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, 400, p7_LOCAL);
p7_oprofile_Convert(gm, om1);
om2 = p7_oprofile_Copy(om1);
if (p7_oprofile_Compare(om1, om2, 0.001f, errbuf) != eslOK) p7_Fail("Compare failed %s\n", errbuf);
p7_oprofile_Destroy(om1);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
p7_hmmfile_Close(hfp);
esl_alphabet_Destroy(abc);
return eslOK;
}
static void
pipeline_thread(void *arg)
{
int workeridx;
int status;
WORK_ITEM *item;
void *newItem;
WORKER_INFO *info;
ESL_THREADS *obj;
obj = (ESL_THREADS *) arg;
esl_threads_Started(obj, &workeridx);
info = (WORKER_INFO *) esl_threads_GetData(obj, workeridx);
status = esl_workqueue_WorkerUpdate(info->queue, NULL, &newItem);
if (status != eslOK) esl_fatal("Work queue worker failed");
/* loop until all blocks have been processed */
item = (WORK_ITEM *) newItem;
while (item->msa != NULL)
{
status = p7_Builder(info->bld, item->msa, info->bg, &item->hmm, NULL, NULL, NULL, &item->postmsa);
if (status != eslOK) p7_Fail("build failed: %s", info->bld->errbuf);
item->entropy = p7_MeanMatchRelativeEntropy(item->hmm, info->bg);
item->processed = TRUE;
status = esl_workqueue_WorkerUpdate(info->queue, item, &newItem);
if (status != eslOK) esl_fatal("Work queue worker failed");
item = (WORK_ITEM *) newItem;
}
status = esl_workqueue_WorkerUpdate(info->queue, item, NULL);
if (status != eslOK) esl_fatal("Work queue worker failed");
esl_threads_Finished(obj, workeridx);
return;
}
/* Function: p7_Alimask_MakeModel2AliMap()
* Synopsis: Compute map of coordinate in the alignment corresponding to each model position.
*
* Args: msa - The alignment for which the mapped model is to be computed. We assume
* the MSA has already been manipulated to account for model building
* flags (e.g. weighting).
* do_hand - TRUE when the model is to follow a hand-build RF line (which must be
* part of the file.
* symfraq - if weighted occupancy exceeds this value, include the column in the model.
* map - int array into which the map values will be stored. Calling function
* must allocate (msa->alen+1) ints.
*
* Returns: The number of mapped model positions.
*/
int
p7_Alimask_MakeModel2AliMap(ESL_MSA *msa, int do_hand, float symfrac, int *map )
{
int i = 0;
int apos, idx;
float r; /* weighted residue count */
float totwgt; /* weighted residue+gap count */
i = 0;
if ( do_hand ) {
if (msa->rf == NULL) p7_Fail("Model file does not contain an RF line, required for --hand.\n");
/* Watch for off-by-one. rf is [0..alen-1]*/
for (apos = 1; apos <= msa->alen; apos++) {
if (!esl_abc_CIsGap(msa->abc, msa->rf[apos-1]) ) {
map[i] = apos;
i++;
}
}
} else {
for (apos = 1; apos <= msa->alen; apos++)
{
r = totwgt = 0.;
for (idx = 0; idx < msa->nseq; idx++)
{
if (esl_abc_XIsResidue(msa->abc, msa->ax[idx][apos])) { r += msa->wgt[idx]; totwgt += msa->wgt[idx]; }
else if (esl_abc_XIsGap(msa->abc, msa->ax[idx][apos])) { totwgt += msa->wgt[idx]; }
else if (esl_abc_XIsMissing(msa->abc, msa->ax[idx][apos])) continue;
}
if (r > 0. && r / totwgt >= symfrac) {
map[i] = apos;
i++;
}
}
}
return i;
}
/* onefetch():
* Given one <key> (an HMM name or accession), retrieve the corresponding HMM.
* In SSI mode, we can do this quickly by positioning the file, then reading
* and writing the HMM that's at that position.
* Without an SSI index, we have to parse the HMMs sequentially 'til we find
* the one we're after.
*/
static void
onefetch(ESL_GETOPTS *go, FILE *ofp, char *key, P7_HMMFILE *hfp)
{
ESL_ALPHABET *abc = NULL;
P7_HMM *hmm = NULL;
int status;
if (hfp->ssi != NULL)
{
status = p7_hmmfile_PositionByKey(hfp, key);
if (status == eslENOTFOUND) p7_Fail("HMM %s not found in SSI index for file %s\n", key, hfp->fname);
else if (status == eslEFORMAT) p7_Fail("Failed to parse SSI index for %s\n", hfp->fname);
else if (status != eslOK) p7_Fail("Failed to look up location of HMM %s in SSI index of file %s\n", key, hfp->fname);
}
while ((status = p7_hmmfile_Read(hfp, &abc, &hmm)) != eslEOF)
{
if (status == eslEOD) p7_Fail("read failed, HMM file %s may be truncated?", hfp->fname);
else if (status == eslEFORMAT) p7_Fail("bad file format in HMM file %s", hfp->fname);
else if (status == eslEINCOMPAT) p7_Fail("HMM file %s contains different alphabets", hfp->fname);
else if (status != eslOK) p7_Fail("Unexpected error in reading HMMs from %s", hfp->fname);
if (strcmp(key, hmm->name) == 0 || (hmm->acc && strcmp(key, hmm->acc) == 0)) break;
p7_hmm_Destroy(hmm);
hmm = NULL;
}
if (status == eslOK)
{
p7_hmmfile_WriteASCII(ofp, -1, hmm);
p7_hmm_Destroy(hmm);
}
else p7_Fail("HMM %s not found in file %s\n", key, hfp->fname);
esl_alphabet_Destroy(abc);
}
/* recalibrate_model()
*
* Optionally, user can choose (with --recal) to replace the
* statistical parameters of the HMM. The calibrated params are used
* to generate "expected" distributions in output plots. see
* p7_Calibrate(), which this is a partial copy of. (p7_Calibrate()
* requires a P7_BUILDER object that we don't have.)
*
* On success, returns <eslOK> and the statistical parameters of the
* <hmm> have been recalibrated and replaced.
*/
static int
recalibrate_model(ESL_GETOPTS *go, struct cfg_s *cfg, char *errbuf, P7_HMM *hmm)
{
P7_PROFILE *gm = NULL;
P7_OPROFILE *om = NULL;
int EmL = esl_opt_GetInteger(go, "--EmL");
int EmN = esl_opt_GetInteger(go, "--EmN");
int EvL = esl_opt_GetInteger(go, "--EvL");
int EvN = esl_opt_GetInteger(go, "--EvN");
int EfL = esl_opt_GetInteger(go, "--EfL");
int EfN = esl_opt_GetInteger(go, "--EfN");
double Eft = esl_opt_GetReal (go, "--Eft");
double lambda, mmu, vmu, ftau;
gm = p7_profile_Create(hmm->M, cfg->abc);
p7_profile_Config(gm, hmm, cfg->bg); /* dual-mode multihit; L=0 (no length model needed right now) */
P7_HARDWARE *hw;
if ((hw = p7_hardware_Create ()) == NULL) p7_Fail("Couldn't get HW information data structure");
om = p7_oprofile_Create(gm->M, cfg->abc, hw->simd);
p7_oprofile_Convert(gm, om); /* om is now *local* multihit */
p7_Lambda(hmm, cfg->bg, &lambda);
p7_MSVMu (cfg->r, om, cfg->bg, EmL, EmN, lambda, &mmu);
p7_ViterbiMu(cfg->r, om, cfg->bg, EvL, EvN, lambda, &vmu);
p7_Tau (cfg->r, om, cfg->bg, EfL, EfN, lambda, Eft, &ftau);
hmm->evparam[p7_MLAMBDA] = lambda;
hmm->evparam[p7_VLAMBDA] = lambda;
hmm->evparam[p7_FLAMBDA] = lambda;
hmm->evparam[p7_MMU] = mmu;
hmm->evparam[p7_VMU] = vmu;
hmm->evparam[p7_FTAU] = ftau;
hmm->flags |= p7H_STATS;
p7_profile_Destroy(gm);
p7_oprofile_Destroy(om);
return eslOK;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_CreateDefaultApp(options, 1, argc, argv, banner, usage);
ESL_STOPWATCH *w = esl_stopwatch_Create();
ESL_ALPHABET *abc = NULL;
char *hmmfile = esl_opt_GetArg(go, 1);
P7_HMMFILE *hfp = NULL;
P7_OPROFILE *om = NULL;
int nmodel = 0;
uint64_t totM = 0;
int status;
char errbuf[eslERRBUFSIZE];
esl_stopwatch_Start(w);
status = p7_hmmfile_OpenE(hmmfile, NULL, &hfp, errbuf);
if (status == eslENOTFOUND) p7_Fail("File existence/permissions problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf);
else if (status == eslEFORMAT) p7_Fail("File format problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf);
else if (status != eslOK) p7_Fail("Unexpected error %d in opening HMM file %s.\n%s\n", status, hmmfile, errbuf);
while ((status = p7_oprofile_ReadMSV(hfp, &abc, &om)) == eslOK)
{
nmodel++;
totM += om->M;
p7_oprofile_Destroy(om);
}
if (status == eslEFORMAT) p7_Fail("bad file format in profile file %s", hmmfile);
else if (status == eslEINCOMPAT) p7_Fail("profile file %s contains different alphabets", hmmfile);
else if (status != eslEOF) p7_Fail("Unexpected error in reading profiles from %s", hmmfile);
esl_stopwatch_Stop(w);
esl_stopwatch_Display(stdout, w, "# CPU time: ");
printf("# number of models: %d\n", nmodel);
printf("# total M: %" PRId64 "\n", totM);
p7_hmmfile_Close(hfp);
esl_alphabet_Destroy(abc);
esl_stopwatch_Destroy(w);
esl_getopts_Destroy(go);
return 0;
}
int main(int argc, char **argv)
{
ESL_GETOPTS *go = esl_getopts_CreateDefaultApp(options, 2, argc, argv, banner, usage);
char *hmmfile = esl_opt_GetArg(go, 1);
char *seqfile = esl_opt_GetArg(go, 2);
ESL_STOPWATCH *w = esl_stopwatch_Create();
ESL_RANDOMNESS*r = esl_randomness_Create(esl_opt_GetInteger(go, "-s"));
ESL_ALPHABET*abc = NULL;
P7_HMMFILE *hfp = NULL;
P7_HMM *hmm = NULL;
P7_BG *bg = NULL;
P7_PROFILE *gm1, *gm2;
int L = 2000; // esl_opt_GetInteger(go, "-L");
int N = esl_opt_GetInteger(go, "-N");
int MaxPart = esl_opt_GetInteger(go, "-M");
__m128 resdata[10];
float *sc1 = (float*) (resdata+0); // uses 1 __m128s
ESL_DSQ *dsq = NULL;
int i, j;
ESL_SQFILE *sqfp = NULL;
DATA_STREAM *dstream;
struct timeb tbstart, tbend;
int sumlengths = 0;
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");
bg = p7_bg_Create(abc);
p7_bg_SetLength(bg, L);
gm1 = p7_profile_Create(hmm->M, abc);
gm2 = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm1, L, p7_UNILOCAL);
p7_ProfileConfig(hmm, bg, gm2, L, p7_UNILOCAL);
dstream = p7_ViterbiStream_Create(gm1);
p7_ViterbiStream_Setup(dstream, L+100, MaxPart); // use max L
dstream->L = L;
// No. of partitions computed without full parallelism ( == no. of threads active while some are idle)
int Niters_part = dstream->Npartitions % NTHREADS;
// No. of Model lines that could be computed but are wasted by idle threads waiting on the end
int Nwasted_threads = dstream->partition * ((NTHREADS-Niters_part) % NTHREADS);
// No. of lines in the last partition that go beyond M. It's wasted comp time by a single thread
int Nwasted_leftover= (dstream->partition - gm1->M % dstream->partition) % dstream->partition;
// Total number of wasted lines
int wastedcomp = Nwasted_threads + Nwasted_leftover;
// Total number of lines computed and waited for
int totalcomp = wastedcomp + gm1->M; // same as: roundtop(gm1->M, dstream->partition * NTHREADS);
printf("Viterbi Stream Word with %d Threads, model %s: Modelsize %d, #Segms: %d, SeqL: %d, Nseqs %d, Part %d, #Parts %d\n",
NTHREADS, hmmfile, gm1->M, (int) ceil(gm1->M/8.0), L, 8*N, dstream->partition, dstream->Npartitions);
printf("Total Comp Lines: %d | Wasted Comp Lines: %d\n", totalcomp, wastedcomp);
// for ViterbiFilter
P7_OPROFILE *om = p7_oprofile_Create(hmm->M, gm1->abc);
p7_oprofile_Convert(gm1, om);
P7_OMX *ox = p7_omx_Create(hmm->M, 0, 0);
dsq_cmp_t **seqsdb= calloc(8*N+64, sizeof(dsq_cmp_t*));
if(0)
{ ESL_SQ* sq = esl_sq_CreateDigital(abc);
if (esl_sqfile_OpenDigital(abc, seqfile, eslSQFILE_FASTA, NULL, &sqfp) != eslOK)
{ p7_Fail("Failed to open sequence file\n"); return -1; }
for (j = 0; j < 8*N; j++)
{
int res = esl_sqio_Read(sqfp, sq);
if (res != eslOK) { printf("ATENCAO: faltam sequencias\n"); break; }
int len = sq->n;
dsq = sq->dsq;
seqsdb[j] = malloc(sizeof(dsq_cmp_t));
seqsdb[j]->length = len;
seqsdb[j]->seq = malloc((len+4)*sizeof(ESL_DSQ));
memcpy(seqsdb[j]->seq, dsq, len+2);
sumlengths += len;
esl_sq_Reuse(sq);
}
ftime(&tbstart);
N = j/8;
printf("N = %d\n", N);
// Sort sequences by length
qsort(seqsdb, N*8, sizeof(dsq_cmp_t*), compare_seqs);
}
else if(0)
for (i = 0; i < N; i++)
{
for (j = 0; j < 8; j++)
{
int len = L - rand()%1000;
seqsdb[i*8+j] = malloc(sizeof(dsq_cmp_t));
seqsdb[i*8+j]->seq = malloc(len+4);
seqsdb[i*8+j]->length = len;
esl_rsq_xfIID(r, bg->f, abc->K, len, seqsdb[i*8+j]->seq);
sumlengths += len;
}
}
// double sumerrors = 0;
float* results = (float*) alloc_m128_aligned64(N*2+2);
ftime(&tbstart);
for (j = 0; j < N; j++)
for (i = 0; i < N; i++)
{
// if (i % 10000 == 0) printf("START %d\n", i);
p7_ViterbiStream(dstream, seqsdb+i*8, sc1);
// memcpy(results+i*8, sc1, 32);
}
ftime(&tbend);
double secs = TIMEDIFF(tbstart,tbend);
// printf("Qsort time: %6.3f | Viterbi time: %6.3f\n", TIMEDIFF(tbqsort,tbstart), secs);
w->elapsed = w->user = secs;
esl_stopwatch_Display(stdout, w, "# Opt CPU time: ");
printf("# %.0fM cells in %.1f Mc/s\n", (sumlengths * (double) gm1->M) / 1e6, (sumlengths * (double) gm1->M * 1e-6) / secs);
if(0) // compare results against base version
for (i = 0; i < 1000 && i < N; i++)
{
int maxll = 0; float sc2;
for (j = 0; j < 8; j++)
if (maxll < seqsdb[i*8+j]->length)
maxll = seqsdb[i*8+j]->length;
// for (j = 0; j < 8; j++) printf("%d ", seqsdb[i*8+j]->length); printf("\n");
// if (i % 10 == 0) printf("i %d\n", i);
p7_oprofile_ReconfigRestLength(om, maxll);
p7_ReconfigLength(gm2, maxll); // fazer Reconfig aqui para emular compl o VitStream
for (j = 0; j < 8; j++)
{
// p7_ReconfigLength(gm2, seqsdb[i*8+j]->length);
// p7_Viterbi_unilocal(seqsdb[i*8+j]->seq, seqsdb[i*8+j]->length, gm2, &sc3);
// p7_Viterbi_unilocal_word(seqsdb[i*8+j]->seq, seqsdb[i*8+j]->length, gm2, &sc2);
// p7_oprofile_ReconfigLength(om, seqsdb[i*8+j]->length);
p7_ViterbiFilter(seqsdb[i*8+j]->seq, seqsdb[i*8+j]->length, om, ox, &sc2);
//sumerrors += fabs(sc1[j]- sc2);
if (fabs(results[i*8+j] - sc2) > 0.00001)
{ printf("%3d-%d L %4d: VS %f d %f\t| Base SerI %f\n", i, j, seqsdb[i*8+j]->length,
results[i*8+j], fabs(results[i*8+j] - sc2), sc2);
getc(stdin);
}
}
}
return 0;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = esl_getopts_CreateDefaultApp(options, 2, argc, argv, banner, usage);
char *hmmfile = esl_opt_GetArg(go, 1);
char *seqfile = esl_opt_GetArg(go, 2);
ESL_ALPHABET *abc = NULL;
P7_HMMFILE *hfp = NULL;
P7_HMM *hmm = NULL;
P7_BG *bg = NULL;
P7_PROFILE *gm = NULL;
P7_GMX *fwd = NULL;
ESL_SQ *sq = NULL;
ESL_SQFILE *sqfp = NULL;
P7_TRACE *tr = NULL;
int format = eslSQFILE_UNKNOWN;
char errbuf[eslERRBUFSIZE];
float sc;
int d;
int status;
/* Read in one HMM */
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);
/* Read in one sequence */
sq = esl_sq_CreateDigital(abc);
status = esl_sqfile_Open(seqfile, format, NULL, &sqfp);
if (status == eslENOTFOUND) p7_Fail("No such file.");
else if (status == eslEFORMAT) p7_Fail("Format unrecognized.");
else if (status == eslEINVAL) p7_Fail("Can't autodetect stdin or .gz.");
else if (status != eslOK) p7_Fail("Open failed, code %d.", status);
if (esl_sqio_Read(sqfp, sq) != eslOK) p7_Fail("Failed to read sequence");
esl_sqfile_Close(sqfp);
/* Configure a profile from the HMM */
bg = p7_bg_Create(abc);
p7_bg_SetLength(bg, sq->n);
gm = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, sq->n, p7_LOCAL);
/* Allocate matrix and a trace */
fwd = p7_gmx_Create(gm->M, sq->n);
tr = p7_trace_Create();
/* Run Viterbi; do traceback */
p7_GViterbi (sq->dsq, sq->n, gm, fwd, &sc);
p7_GTrace (sq->dsq, sq->n, gm, fwd, tr);
/* Dump and validate the trace. */
p7_trace_Dump(stdout, tr, gm, sq->dsq);
if (p7_trace_Validate(tr, abc, sq->dsq, errbuf) != eslOK) p7_Die("trace fails validation:\n%s\n", errbuf);
/* Domain info in the trace. */
p7_trace_Index(tr);
printf("# Viterbi: %d domains : ", tr->ndom);
for (d = 0; d < tr->ndom; d++) printf("%6d %6d %6d %6d ", tr->sqfrom[d], tr->sqto[d], tr->hmmfrom[d], tr->hmmto[d]);
printf("\n");
/* Cleanup */
p7_trace_Destroy(tr);
p7_gmx_Destroy(fwd);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
esl_sq_Destroy(sq);
esl_alphabet_Destroy(abc);
esl_getopts_Destroy(go);
return 0;
}
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);
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_GMX *gx = NULL;
int L = esl_opt_GetInteger(go, "-L");
int N = esl_opt_GetInteger(go, "-N");
ESL_DSQ *dsq = malloc(sizeof(ESL_DSQ) * (L+2));
int i;
float sc;
double base_time, bench_time, Mcs;
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");
bg = p7_bg_Create(abc);
p7_bg_SetLength(bg, L);
gm = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, L, p7_UNILOCAL);
gx = p7_gmx_Create(gm->M, L);
/* Baseline time. */
esl_stopwatch_Start(w);
for (i = 0; i < N; i++) esl_rsq_xfIID(r, bg->f, abc->K, L, dsq);
esl_stopwatch_Stop(w);
base_time = w->user;
/* Benchmark time. */
esl_stopwatch_Start(w);
for (i = 0; i < N; i++)
{
esl_rsq_xfIID(r, bg->f, abc->K, L, dsq);
p7_GViterbi (dsq, L, gm, gx, &sc);
}
esl_stopwatch_Stop(w);
bench_time = w->user - base_time;
Mcs = (double) N * (double) L * (double) gm->M * 1e-6 / (double) bench_time;
esl_stopwatch_Display(stdout, w, "# CPU time: ");
printf("# M = %d\n", gm->M);
printf("# %.1f Mc/s\n", Mcs);
free(dsq);
p7_gmx_Destroy(gx);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
p7_hmmfile_Close(hfp);
esl_alphabet_Destroy(abc);
esl_stopwatch_Destroy(w);
esl_randomness_Destroy(r);
esl_getopts_Destroy(go);
return 0;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = NULL;
int i, j;
int status = eslOK;
P7_HMMFILE *hfp = NULL; /* open input HMM file */
P7_HMM *hmm = NULL; /* one HMM query */
ESL_ALPHABET *abc = NULL; /* digital alphabet */
P7_BG *bg = NULL;
char errbuf[eslERRBUFSIZE];
char* hmmfile;
float *rel_ents = NULL;
float **heights = NULL;
float **probs = NULL;
float *ins_P = NULL;
float *ins_expL = NULL;
float *occupancy = NULL;
int mode = HMMLOGO_RELENT_ALL; //default
go = esl_getopts_Create(options);
if (esl_opt_ProcessCmdline(go, argc, argv) != eslOK) esl_fatal(argv[0], "Failed to parse command line: %s\n", go->errbuf);
if (esl_opt_VerifyConfig(go) != eslOK) esl_fatal(argv[0], "Error in configuration: %s\n", go->errbuf);
if (esl_opt_GetBoolean(go, "-h") ) {
p7_banner (stdout, argv[0], banner);
esl_usage (stdout, argv[0], usage);
puts("\nOptions:");
esl_opt_DisplayHelp(stdout, go, 1, 2, 100);
exit(0);
}
if (esl_opt_ArgNumber(go) != 1) esl_fatal(argv[0], "Incorrect number of command line arguments.\n");
hmmfile = esl_opt_GetArg(go, 1);
if (esl_opt_IsOn(go, "--height_relent_all"))
mode = HMMLOGO_RELENT_ALL;
else if (esl_opt_IsOn(go, "--height_relent_abovebg"))
mode = HMMLOGO_RELENT_ABOVEBG;
else if (esl_opt_IsOn(go, "--height_score"))
mode = HMMLOGO_SCORE;
else
mode = HMMLOGO_RELENT_ALL; //default
/* Open the query profile HMM file */
status = p7_hmmfile_OpenE(hmmfile, NULL, &hfp, errbuf);
if (status == eslENOTFOUND) p7_Fail("File existence/permissions problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf);
else if (status == eslEFORMAT) p7_Fail("File format problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf);
else if (status != eslOK) p7_Fail("Unexpected error %d in opening HMM file %s.\n%s\n", status, hmmfile, errbuf);
status = p7_hmmfile_Read(hfp, &abc, &hmm);
bg = p7_bg_Create(abc);
ESL_ALLOC(rel_ents, (hmm->M+1) * sizeof(float));
ESL_ALLOC(heights, (hmm->M+1) * sizeof(float*));
ESL_ALLOC(probs, (hmm->M+1) * sizeof(float*));
for (i = 1; i <= hmm->M; i++) {
ESL_ALLOC(heights[i], abc->K * sizeof(float));
ESL_ALLOC(probs[i], abc->K * sizeof(float));
}
/* residue heights */
if (mode == HMMLOGO_RELENT_ALL) {
printf ("max expected height = %.2f\n", hmmlogo_maxHeight(bg) );
hmmlogo_RelativeEntropy_all(hmm, bg, rel_ents, probs, heights);
} else if (mode == HMMLOGO_RELENT_ABOVEBG) {
printf ("max expected height = %.2f\n", hmmlogo_maxHeight(bg) );
hmmlogo_RelativeEntropy_above_bg(hmm, bg, rel_ents, probs, heights);
} else if (mode == HMMLOGO_SCORE) {
hmmlogo_ScoreHeights(hmm, bg, heights );
}
printf ("Residue heights\n");
for (i = 1; i <= hmm->M; i++) {
printf("%d: ", i);
for (j=0; j<abc->K; j++)
printf("%6.3f ", heights[i][j] );
if (mode != HMMLOGO_SCORE)
printf(" (%6.3f)", rel_ents[i]);
printf("\n");
}
if (rel_ents != NULL) free(rel_ents);
if (heights != NULL) {
for (i = 1; i <= hmm->M; i++)
if (heights[i] != NULL) free(heights[i]);
free(heights);
}
/* indel values */
if (! esl_opt_IsOn(go, "--no_indel")) {
ESL_ALLOC(ins_P, (hmm->M+1) * sizeof(float));
ESL_ALLOC(ins_expL, (hmm->M+1) * sizeof(float));
ESL_ALLOC(occupancy, (hmm->M+1) * sizeof(float));
hmmlogo_IndelValues(hmm, ins_P, ins_expL, occupancy);
printf ("Indel values\n");
for (i = 1; i <= hmm->M; i++)
printf("%d: %6.3f %6.3f %6.3f\n", i, ins_P[i], ins_expL[i], occupancy[i] );
free(ins_P);
free(ins_expL);
free(occupancy);
}
p7_hmmfile_Close(hfp);
esl_alphabet_Destroy(abc);
p7_bg_Destroy(bg);
exit(0);
ERROR:
if (rel_ents != NULL) free(rel_ents);
if (heights != NULL) {
for (i = 1; i <= hmm->M; i++)
if (heights[i] != NULL) free(heights[i]);
free(heights);
}
if (hfp != NULL) p7_hmmfile_Close(hfp);
if (abc != NULL) esl_alphabet_Destroy(abc);
if (ins_P != NULL) free(ins_P);
if (ins_expL != NULL) free(ins_expL);
if (occupancy != NULL) free(occupancy);
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_CreateDefaultApp(options, 1, argc, argv, banner, usage);
char *hmmfile = esl_opt_GetArg(go, 1);
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;
double lambda = 0.0;
double mmu = 0.0;
double vmu = 0.0;
double ftau = 0.0;
int Z = esl_opt_GetInteger(go, "-Z");
int EmL = esl_opt_GetInteger(go, "--EmL");
int EmN = esl_opt_GetInteger(go, "--EmN");
int EvL = esl_opt_GetInteger(go, "--EvL");
int EvN = esl_opt_GetInteger(go, "--EvN");
int EfL = esl_opt_GetInteger(go, "--EfL");
int EfN = esl_opt_GetInteger(go, "--EfN");
int Eft = esl_opt_GetReal (go, "--Eft");
int iteration;
int do_msv, do_vit, do_fwd;
int status;
if (esl_opt_GetBoolean(go, "--msvonly") == TRUE) { do_msv = TRUE; do_vit = FALSE; do_fwd = FALSE; }
else if (esl_opt_GetBoolean(go, "--vitonly") == TRUE) { do_msv = FALSE; do_vit = TRUE; do_fwd = FALSE; }
else if (esl_opt_GetBoolean(go, "--fwdonly") == TRUE) { do_msv = FALSE; do_vit = FALSE; do_fwd = TRUE; }
else { do_msv = TRUE; do_vit = TRUE; do_fwd = TRUE; }
if (p7_hmmfile_OpenE(hmmfile, NULL, &hfp, NULL) != eslOK) p7_Fail("Failed to open HMM file %s", hmmfile);
while ((status = p7_hmmfile_Read(hfp, &abc, &hmm)) != eslEOF)
{
if (bg == NULL) bg = p7_bg_Create(abc);
gm = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, EvL, p7_LOCAL); /* the EvL doesn't matter */
om = p7_oprofile_Create(hmm->M, abc);
p7_oprofile_Convert(gm, om);
if (esl_opt_IsOn(go, "--lambda")) lambda = esl_opt_GetReal(go, "--lambda");
else p7_Lambda(hmm, bg, &lambda);
for (iteration = 0; iteration < Z; iteration++)
{
if (do_msv) p7_MSVMu (r, om, bg, EmL, EmN, lambda, &mmu);
if (do_vit) p7_ViterbiMu (r, om, bg, EvL, EvN, lambda, &vmu);
if (do_fwd) p7_Tau (r, om, bg, EfL, EfN, lambda, Eft, &ftau);
printf("%s %.4f %.4f %.4f %.4f\n", hmm->name, lambda, mmu, vmu, ftau);
}
p7_hmm_Destroy(hmm);
p7_profile_Destroy(gm);
p7_oprofile_Destroy(om);
}
p7_hmmfile_Close(hfp);
p7_bg_Destroy(bg);
esl_alphabet_Destroy(abc);
esl_randomness_Destroy(r);
esl_getopts_Destroy(go);
return eslOK;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = esl_getopts_CreateDefaultApp(options, 2, argc, argv, banner, usage);
char *hmmfile = esl_opt_GetArg(go, 1);
char *seqfile = esl_opt_GetArg(go, 2);
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_OMX *ox = NULL;
P7_GMX *gx = NULL;
ESL_SQ *sq = NULL;
ESL_SQFILE *sqfp = NULL;
int format = eslSQFILE_UNKNOWN;
float vfraw, nullsc, vfscore;
float graw, gscore;
double P, gP;
int status;
/* Read in one HMM */
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");
/* Read in one sequence */
sq = esl_sq_CreateDigital(abc);
status = esl_sqfile_Open(seqfile, format, NULL, &sqfp);
if (status == eslENOTFOUND) p7_Fail("No such file.");
else if (status == eslEFORMAT) p7_Fail("Format unrecognized.");
else if (status == eslEINVAL) p7_Fail("Can't autodetect stdin or .gz.");
else if (status != eslOK) p7_Fail("Open failed, code %d.", status);
/* create default null model, then create and optimize profile */
bg = p7_bg_Create(abc);
p7_bg_SetLength(bg, sq->n);
gm = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, sq->n, p7_LOCAL);
om = p7_oprofile_Create(gm->M, abc);
p7_oprofile_Convert(gm, om);
/* allocate DP matrices, both a generic and an optimized one */
ox = p7_omx_Create(gm->M, 0, sq->n);
gx = p7_gmx_Create(gm->M, sq->n);
/* Useful to place and compile in for debugging:
p7_oprofile_Dump(stdout, om); dumps the optimized profile
p7_omx_SetDumpMode(ox, TRUE); makes the fast DP algorithms dump their matrices
p7_gmx_Dump(stdout, gx); dumps a generic DP matrix
*/
while ((status = esl_sqio_Read(sqfp, sq)) == eslOK)
{
p7_oprofile_ReconfigLength(om, sq->n);
p7_ReconfigLength(gm, sq->n);
p7_bg_SetLength(bg, sq->n);
p7_omx_GrowTo(ox, om->M, 0, sq->n);
p7_gmx_GrowTo(gx, gm->M, sq->n);
p7_ViterbiFilter (sq->dsq, sq->n, om, ox, &vfraw);
p7_bg_NullOne (bg, sq->dsq, sq->n, &nullsc);
vfscore = (vfraw - nullsc) / eslCONST_LOG2;
P = esl_gumbel_surv(vfscore, om->evparam[p7_VMU], om->evparam[p7_VLAMBDA]);
p7_GViterbi (sq->dsq, sq->n, gm, gx, &graw);
gscore = (graw - nullsc) / eslCONST_LOG2;
gP = esl_gumbel_surv(gscore, gm->evparam[p7_VMU], gm->evparam[p7_VLAMBDA]);
if (esl_opt_GetBoolean(go, "-1"))
{
printf("%-30s\t%-20s\t%9.2g\t%7.2f\t%9.2g\t%7.2f\n", sq->name, hmm->name, P, vfscore, gP, gscore);
}
else if (esl_opt_GetBoolean(go, "-P"))
{ /* output suitable for direct use in profmark benchmark postprocessors: */
printf("%g\t%.2f\t%s\t%s\n", P, vfscore, sq->name, hmm->name);
}
else
{
printf("target sequence: %s\n", sq->name);
printf("vit filter raw score: %.2f nats\n", vfraw);
printf("null score: %.2f nats\n", nullsc);
printf("per-seq score: %.2f bits\n", vfscore);
printf("P-value: %g\n", P);
printf("GViterbi raw score: %.2f nats\n", graw);
printf("GViterbi seq score: %.2f bits\n", gscore);
printf("GViterbi P-value: %g\n", gP);
}
esl_sq_Reuse(sq);
}
/* cleanup */
esl_sq_Destroy(sq);
esl_sqfile_Close(sqfp);
p7_omx_Destroy(ox);
p7_gmx_Destroy(gx);
p7_oprofile_Destroy(om);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
p7_hmmfile_Close(hfp);
esl_alphabet_Destroy(abc);
esl_getopts_Destroy(go);
return 0;
}
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);
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_OMX *ox = NULL;
P7_GMX *gx = NULL;
int L = esl_opt_GetInteger(go, "-L");
int N = esl_opt_GetInteger(go, "-N");
ESL_DSQ *dsq = malloc(sizeof(ESL_DSQ) * (L+2));
int i;
float sc1, sc2;
double base_time, bench_time, Mcs;
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");
bg = p7_bg_Create(abc);
p7_bg_SetLength(bg, L);
gm = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, L, p7_LOCAL);
om = p7_oprofile_Create(gm->M, abc);
p7_oprofile_Convert(gm, om);
p7_oprofile_ReconfigLength(om, L);
if (esl_opt_GetBoolean(go, "-x")) p7_profile_SameAsVF(om, gm);
ox = p7_omx_Create(gm->M, 0, 0);
gx = p7_gmx_Create(gm->M, L);
/* Get a baseline time: how long it takes just to generate the sequences */
esl_stopwatch_Start(w);
for (i = 0; i < N; i++)
esl_rsq_xfIID(r, bg->f, abc->K, L, dsq);
esl_stopwatch_Stop(w);
base_time = w->user;
/* Run the benchmark */
esl_stopwatch_Start(w);
for (i = 0; i < N; i++)
{
esl_rsq_xfIID(r, bg->f, abc->K, L, dsq);
p7_ViterbiFilter(dsq, L, om, ox, &sc1);
if (esl_opt_GetBoolean(go, "-c"))
{
p7_GViterbi(dsq, L, gm, gx, &sc2);
printf("%.4f %.4f\n", sc1, sc2);
}
if (esl_opt_GetBoolean(go, "-x"))
{
p7_GViterbi(dsq, L, gm, gx, &sc2);
sc2 /= om->scale_w;
if (om->mode == p7_UNILOCAL) sc2 -= 2.0; /* that's ~ L \log \frac{L}{L+2}, for our NN,CC,JJ */
else if (om->mode == p7_LOCAL) sc2 -= 3.0; /* that's ~ L \log \frac{L}{L+3}, for our NN,CC,JJ */
printf("%.4f %.4f\n", sc1, sc2);
}
}
esl_stopwatch_Stop(w);
bench_time = w->user - base_time;
Mcs = (double) N * (double) L * (double) gm->M * 1e-6 / (double) bench_time;
esl_stopwatch_Display(stdout, w, "# CPU time: ");
printf("# M = %d\n", gm->M);
printf("# %.1f Mc/s\n", Mcs);
free(dsq);
p7_omx_Destroy(ox);
p7_gmx_Destroy(gx);
p7_oprofile_Destroy(om);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
p7_hmmfile_Close(hfp);
esl_alphabet_Destroy(abc);
esl_stopwatch_Destroy(w);
esl_randomness_Destroy(r);
esl_getopts_Destroy(go);
return 0;
}
int
main(int argc, char **argv)
{
ESL_ALPHABET *abc = NULL; /* sequence alphabet */
ESL_GETOPTS *go = NULL; /* command line processing */
ESL_RANDOMNESS *r = NULL; /* source of randomness */
P7_HMM *hmm = NULL; /* sampled HMM to emit from */
P7_HMM *core = NULL; /* safe copy of the HMM, before config */
P7_BG *bg = NULL; /* null model */
ESL_SQ *sq = NULL; /* sampled sequence */
P7_TRACE *tr = NULL; /* sampled trace */
P7_PROFILE *gm = NULL; /* profile */
int i,j;
int i1,i2;
int k1,k2;
int iseq;
FILE *fp = NULL;
double expected;
int do_ilocal;
char *hmmfile = NULL;
int nseq;
int do_swlike;
int do_ungapped;
int L;
int M;
int do_h2;
char *ipsfile = NULL;
char *kpsfile = NULL;
ESL_DMATRIX *imx = NULL;
ESL_DMATRIX *kmx = NULL;
ESL_DMATRIX *iref = NULL; /* reference matrix: expected i distribution under ideality */
int Lbins;
int status;
char errbuf[eslERRBUFSIZE];
/*****************************************************************
* Parse the command line
*****************************************************************/
go = esl_getopts_Create(options);
if (esl_opt_ProcessCmdline(go, argc, argv) != eslOK) esl_fatal("Failed to parse command line: %s\n", go->errbuf);
if (esl_opt_VerifyConfig(go) != eslOK) esl_fatal("Failed to parse command line: %s\n", go->errbuf);
if (esl_opt_GetBoolean(go, "-h") == TRUE) {
puts(usage);
puts("\n where options are:\n");
esl_opt_DisplayHelp(stdout, go, 0, 2, 80); /* 0=all docgroups; 2 = indentation; 80=textwidth*/
return eslOK;
}
do_ilocal = esl_opt_GetBoolean(go, "-i");
hmmfile = esl_opt_GetString (go, "-m");
nseq = esl_opt_GetInteger(go, "-n");
do_swlike = esl_opt_GetBoolean(go, "-s");
do_ungapped = esl_opt_GetBoolean(go, "-u");
L = esl_opt_GetInteger(go, "-L");
M = esl_opt_GetInteger(go, "-M");
do_h2 = esl_opt_GetBoolean(go, "-2");
ipsfile = esl_opt_GetString (go, "--ips");
kpsfile = esl_opt_GetString (go, "--kps");
if (esl_opt_ArgNumber(go) != 0) {
puts("Incorrect number of command line arguments.");
printf("Usage: %s [options]\n", argv[0]);
return eslFAIL;
}
r = esl_randomness_CreateFast(0);
if (hmmfile != NULL)
{ /* Read the HMM (and get alphabet from it) */
P7_HMMFILE *hfp = NULL;
status = p7_hmmfile_OpenE(hmmfile, NULL, &hfp, errbuf);
if (status == eslENOTFOUND) p7_Fail("File existence/permissions problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf);
else if (status == eslEFORMAT) p7_Fail("File format problem in trying to open HMM file %s.\n%s\n", hmmfile, errbuf);
else if (status != eslOK) p7_Fail("Unexpected error %d in opening HMM file %s.\n%s\n", status, hmmfile, errbuf);
if ((status = p7_hmmfile_Read(hfp, &abc, &hmm)) != eslOK) {
if (status == eslEOD) esl_fatal("read failed, HMM file %s may be truncated?", hmmfile);
else if (status == eslEFORMAT) esl_fatal("bad file format in HMM file %s", hmmfile);
else if (status == eslEINCOMPAT) esl_fatal("HMM file %s contains different alphabets", hmmfile);
else esl_fatal("Unexpected error in reading HMMs");
}
M = hmm->M;
p7_hmmfile_Close(hfp);
}
else
{ /* Or sample the HMM (create alphabet first) */
abc = esl_alphabet_Create(eslAMINO);
if (do_ungapped) p7_hmm_SampleUngapped(r, M, abc, &hmm);
else if (do_swlike) p7_hmm_SampleUniform (r, M, abc, 0.05, 0.5, 0.05, 0.2, &hmm); /* tmi, tii, tmd, tdd */
else p7_hmm_Sample (r, M, abc, &hmm);
}
Lbins = M;
imx = esl_dmatrix_Create(Lbins, Lbins);
iref = esl_dmatrix_Create(Lbins, Lbins);
kmx = esl_dmatrix_Create(M, M);
esl_dmatrix_SetZero(imx);
esl_dmatrix_SetZero(iref);
esl_dmatrix_SetZero(kmx);
tr = p7_trace_Create();
sq = esl_sq_CreateDigital(abc);
bg = p7_bg_Create(abc);
core = p7_hmm_Clone(hmm);
if (do_h2) {
gm = p7_profile_Create(hmm->M, abc);
p7_H2_ProfileConfig(hmm, bg, gm, p7_UNILOCAL);
} else {
gm = p7_profile_Create(hmm->M, abc);
p7_ProfileConfig(hmm, bg, gm, L, p7_UNILOCAL);
if (p7_hmm_Validate (hmm, NULL, 0.0001) != eslOK) esl_fatal("whoops, HMM is bad!");
if (p7_profile_Validate(gm, NULL, 0.0001) != eslOK) esl_fatal("whoops, profile is bad!");
}
/* Sample endpoints.
* Also sample an ideal reference distribution for i endpoints. i
* endpoints are prone to discretization artifacts, when emitted
* sequences have varying lengths. Taking log odds w.r.t. an ideal
* reference that is subject to the same discretization artifacts
* cancels out the effect.
*/
for (iseq = 0; iseq < nseq; iseq++)
{
if (do_ilocal) ideal_local_endpoints (r, core, sq, tr, Lbins, &i1, &i2, &k1, &k2);
else profile_local_endpoints(r, core, gm, sq, tr, Lbins, &i1, &i2, &k1, &k2);
imx->mx[i1-1][i2-1] += 1.;
kmx->mx[k1-1][k2-1] += 1.;
/* reference distribution for i */
ideal_local_endpoints (r, core, sq, tr, Lbins, &i1, &i2, &k1, &k2);
iref->mx[i1-1][i2-1] += 1.;
}
/* Adjust both mx's to log_2(obs/exp) ratio */
printf("Before normalization/log-odds:\n");
printf(" i matrix values range from %f to %f\n", dmx_upper_min(imx), dmx_upper_max(imx));
printf(" k matrix values range from %f to %f\n", dmx_upper_min(kmx), dmx_upper_max(kmx));
printf("iref matrix values range from %f to %f\n", dmx_upper_min(iref), dmx_upper_max(iref));
expected = (double) nseq * 2. / (double) (M*(M+1));
for (i = 0; i < kmx->m; i++)
for (j = i; j < kmx->n; j++)
kmx->mx[i][j] = log(kmx->mx[i][j] / expected) / log(2.0);
for (i = 0; i < imx->m; i++)
for (j = i; j < imx->m; j++)
if (iref->mx[i][j] == 0. && imx->mx[i][j] == 0.)
imx->mx[i][j] = 0.;
else if (iref->mx[i][j] == 0.)
imx->mx[i][j] = eslINFINITY;
else if (imx->mx[i][j] == 0.)
imx->mx[i][j] = -eslINFINITY;
else
imx->mx[i][j] = log(imx->mx[i][j] / iref->mx[i][j]) / log(2.0);
/* Print ps files */
if (kpsfile != NULL) {
if ((fp = fopen(kpsfile, "w")) == NULL) esl_fatal("Failed to open output postscript file %s", kpsfile);
dmx_Visualize(fp, kmx, -4., 5.);
fclose(fp);
}
if (ipsfile != NULL) {
if ((fp = fopen(ipsfile, "w")) == NULL) esl_fatal("Failed to open output postscript file %s", ipsfile);
dmx_Visualize(fp, imx, -4., 5.);
/* dmx_Visualize(fp, imx, dmx_upper_min(imx), dmx_upper_max(imx)); */
fclose(fp);
}
printf("After normalization/log-odds:\n");
printf("i matrix values range from %f to %f\n", dmx_upper_min(imx), dmx_upper_max(imx));
printf("k matrix values range from %f to %f\n", dmx_upper_min(kmx), dmx_upper_max(kmx));
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(core);
p7_hmm_Destroy(hmm);
p7_trace_Destroy(tr);
esl_sq_Destroy(sq);
esl_dmatrix_Destroy(imx);
esl_dmatrix_Destroy(kmx);
esl_alphabet_Destroy(abc);
esl_randomness_Destroy(r);
esl_getopts_Destroy(go);
return eslOK;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_CreateDefaultApp(options, 1, argc, argv, banner, usage);
char *hmmfile = esl_opt_GetArg(go, 1);
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_OMX *ox1 = NULL;
P7_OMX *ox2 = NULL;
int L = esl_opt_GetInteger(go, "-L");
int N = esl_opt_GetInteger(go, "-N");
ESL_DSQ *dsq = malloc(sizeof(ESL_DSQ) * (L+2));
float null2[p7_MAXCODE];
int i,j,d,pos;
int nsamples = 200;
float fsc, bsc;
double Mcs;
if (p7_hmmfile_OpenE(hmmfile, NULL, &hfp, NULL) != eslOK) p7_Fail("Failed to open HMM file %s", hmmfile);
if (p7_hmmfile_Read(hfp, &abc, &hmm) != eslOK) p7_Fail("Failed to read HMM");
bg = p7_bg_Create(abc); p7_bg_SetLength(bg, L);
gm = p7_profile_Create(hmm->M, abc); p7_ProfileConfig(hmm, bg, gm, L, p7_LOCAL);
om = p7_oprofile_Create(gm->M, abc); p7_oprofile_Convert(gm, om);
p7_oprofile_ReconfigLength(om, L);
ox1 = p7_omx_Create(gm->M, L, L);
ox2 = p7_omx_Create(gm->M, L, L);
esl_rsq_xfIID(r, bg->f, abc->K, L, dsq);
p7_Forward (dsq, L, om, ox1, &fsc);
if (esl_opt_GetBoolean(go, "-t"))
{
P7_TRACE *tr = p7_trace_Create();
float *n2sc = malloc(sizeof(float) * (L+1));
esl_stopwatch_Start(w);
for (i = 0; i < N; i++)
{ /* This is approximately what p7_domaindef.c::region_trace_ensemble() is doing: */
for (j = 0; j < nsamples; j++)
{
p7_StochasticTrace(r, dsq, L, om, ox1, tr);
p7_trace_Index(tr);
pos = 1;
for (d = 0; d < tr->ndom; d++)
{
p7_Null2_ByTrace(om, tr, tr->tfrom[d], tr->tto[d], ox2, null2);
for (; pos <= tr->sqfrom[d]; pos++) n2sc[pos] += 1.0;
for (; pos < tr->sqto[d]; pos++) n2sc[pos] += null2[dsq[pos]];
}
for (; pos <= L; pos++) n2sc[pos] += 1.0;
p7_trace_Reuse(tr);
}
for (pos = 1; pos <= L; pos++)
n2sc[pos] = logf(n2sc[pos] / nsamples);
}
esl_stopwatch_Stop(w);
free(n2sc);
p7_trace_Destroy(tr);
}
else
{
p7_Backward(dsq, L, om, ox1, ox2, &bsc);
p7_Decoding(om, ox1, ox2, ox2);
esl_stopwatch_Start(w);
for (i = 0; i < N; i++)
p7_Null2_ByExpectation(om, ox2, null2);
esl_stopwatch_Stop(w);
}
Mcs = (double) N * (double) L * (double) gm->M * 1e-6 / (double) w->user;
esl_stopwatch_Display(stdout, w, "# CPU time: ");
printf("# M = %d\n", gm->M);
printf("# %.1f Mc/s\n", Mcs);
free(dsq);
p7_omx_Destroy(ox1);
p7_omx_Destroy(ox2);
p7_oprofile_Destroy(om);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
p7_hmmfile_Close(hfp);
esl_alphabet_Destroy(abc);
esl_stopwatch_Destroy(w);
esl_randomness_Destroy(r);
esl_getopts_Destroy(go);
return 0;
}
