/* The "basic" utest is a minimal driver for making a small DNA profile and a small DNA sequence,
* then running Viterbi and Forward. It's useful for dumping DP matrices and profiles for debugging.
*/
static void
utest_basic(ESL_GETOPTS *go)
{
char *query= "# STOCKHOLM 1.0\n\nseq1 GAATTC\nseq2 GAATTC\n//\n";
int fmt = eslMSAFILE_STOCKHOLM;
char *targ = "GAATTC";
ESL_ALPHABET *abc = NULL;
ESL_MSA *msa = NULL;
P7_HMM *hmm = NULL;
P7_PROFILE *gm = NULL;
P7_BG *bg = NULL;
P7_PRIOR *pri = NULL;
ESL_DSQ *dsq = NULL;
P7_GMX *gx = NULL;
P7_TRACE *tr = NULL;
int L = strlen(targ);
float vsc, vsc2, fsc;
if ((abc = esl_alphabet_Create(eslDNA)) == NULL) esl_fatal("failed to create alphabet");
if ((pri = p7_prior_CreateNucleic()) == NULL) esl_fatal("failed to create prior");
if ((msa = esl_msa_CreateFromString(query, fmt)) == NULL) esl_fatal("failed to create MSA");
if (esl_msa_Digitize(abc, msa, NULL) != eslOK) esl_fatal("failed to digitize MSA");
if (p7_Fastmodelmaker(msa, 0.5, NULL, &hmm, NULL) != eslOK) esl_fatal("failed to create GAATTC model");
if (p7_ParameterEstimation(hmm, pri) != eslOK) esl_fatal("failed to parameterize GAATTC model");
if (p7_hmm_SetConsensus(hmm, NULL) != eslOK) esl_fatal("failed to make consensus");
if ((bg = p7_bg_Create(abc)) == NULL) esl_fatal("failed to create DNA null model");
if ((gm = p7_profile_Create(hmm->M, abc)) == NULL) esl_fatal("failed to create GAATTC profile");
if (p7_ProfileConfig(hmm, bg, gm, L, p7_UNILOCAL)!= eslOK) esl_fatal("failed to config profile");
if (p7_profile_Validate(gm, NULL, 0.0001) != eslOK) esl_fatal("whoops, profile is bad!");
if (esl_abc_CreateDsq(abc, targ, &dsq) != eslOK) esl_fatal("failed to create GAATTC digital sequence");
if ((gx = p7_gmx_Create(gm->M, L)) == NULL) esl_fatal("failed to create DP matrix");
if ((tr = p7_trace_Create()) == NULL) esl_fatal("trace creation failed");
p7_GViterbi (dsq, L, gm, gx, &vsc);
if (esl_opt_GetBoolean(go, "-v")) printf("Viterbi score: %.4f\n", vsc);
if (esl_opt_GetBoolean(go, "-v")) p7_gmx_Dump(stdout, gx, p7_DEFAULT);
p7_GTrace (dsq, L, gm, gx, tr);
p7_trace_Score(tr, dsq, gm, &vsc2);
if (esl_opt_GetBoolean(go, "-v")) p7_trace_Dump(stdout, tr, gm, dsq);
if (esl_FCompare(vsc, vsc2, 1e-5) != eslOK) esl_fatal("trace score and Viterbi score don't agree.");
p7_GForward (dsq, L, gm, gx, &fsc);
if (esl_opt_GetBoolean(go, "-v")) printf("Forward score: %.4f\n", fsc);
if (esl_opt_GetBoolean(go, "-v")) p7_gmx_Dump(stdout, gx, p7_DEFAULT);
p7_trace_Destroy(tr);
p7_gmx_Destroy(gx);
free(dsq);
p7_profile_Destroy(gm);
p7_bg_Destroy(bg);
p7_hmm_Destroy(hmm);
esl_msa_Destroy(msa);
p7_prior_Destroy(pri);
esl_alphabet_Destroy(abc);
return;
}
/* Evaluate fx = rel entropy - etarget, which we want to be = 0,
* for effective sequence number <x>.
*/
static int
eweight_target_f(double Neff, void *params, double *ret_fx)
{
struct ew_param_s *p = (struct ew_param_s *) params;
p7_hmm_CopyParameters(p->hmm, p->h2);
p7_hmm_Scale(p->h2, Neff / (double) p->h2->nseq);
p7_ParameterEstimation(p->h2, p->pri);
*ret_fx = p7_MeanMatchRelativeEntropy(p->h2, p->bg) - p->etarget;
return eslOK;
}
/* parameterize()
* Converts counts to probability parameters.
*/
static int
parameterize(P7_BUILDER *bld, P7_HMM *hmm)
{
int status;
if ((status = p7_ParameterEstimation(hmm, bld->prior)) != eslOK) ESL_XFAIL(status, bld->errbuf, "parameter estimation failed");
return eslOK;
ERROR:
return status;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_CreateDefaultApp(options, 1, argc, argv, banner, usage);
char *msafile = esl_opt_GetArg(go, 1);
int fmt = eslMSAFILE_UNKNOWN;
ESL_ALPHABET *abc = NULL;
ESL_MSAFILE *afp = NULL;
ESL_MSA *msa = NULL;
P7_HMM *hmm = NULL;
P7_PRIOR *prior = NULL;
P7_TRACE **trarr = NULL;
P7_BG *bg = NULL;
P7_PROFILE *gm = NULL;
ESL_MSA *postmsa = NULL;
int i;
int status;
/* Standard idioms for opening and reading a digital MSA. (See esl_msafile.c example). */
if (esl_opt_GetBoolean(go, "--rna")) abc = esl_alphabet_Create(eslRNA);
else if (esl_opt_GetBoolean(go, "--dna")) abc = esl_alphabet_Create(eslDNA);
else if (esl_opt_GetBoolean(go, "--amino")) abc = esl_alphabet_Create(eslAMINO);
if ((status = esl_msafile_Open(&abc, msafile, NULL, fmt, NULL, &afp)) != eslOK)
esl_msafile_OpenFailure(afp, status);
bg = p7_bg_Create(abc);
switch (abc->type) {
case eslAMINO: prior = p7_prior_CreateAmino(); break;
case eslDNA: prior = p7_prior_CreateNucleic(); break;
case eslRNA: prior = p7_prior_CreateNucleic(); break;
default: prior = p7_prior_CreateLaplace(abc); break;
}
if (prior == NULL) esl_fatal("Failed to initialize prior");
while ((status = esl_msafile_Read(afp, &msa)) != eslEOF)
{
if (status != eslOK) esl_msafile_ReadFailure(afp, status);
/* The modelmakers collect counts in an HMM structure */
status = p7_Handmodelmaker(msa, NULL, &hmm, &trarr);
if (status == eslENORESULT) esl_fatal("no consensus columns in alignment %s\n", msa->name);
else if (status != eslOK) esl_fatal("failed to build HMM from alignment %s\n", msa->name);
printf("COUNTS:\n");
p7_hmm_Dump(stdout, hmm);
/* These counts, in combination with a prior, are converted to probability parameters */
status = p7_ParameterEstimation(hmm, prior);
if (status != eslOK) esl_fatal("failed to parameterize HMM for %s", msa->name);
printf("PROBABILITIES:\n");
p7_hmm_Dump(stdout, hmm);
/* Just so we can dump a more informatively annotated trace - build a profile */
gm = p7_profile_Create(hmm->M, abc);
p7_profile_Config (gm, hmm, bg);
p7_profile_SetLength(gm, 400);
/* Dump the individual traces */
for (i = 0; i < msa->nseq; i++)
{
printf("Trace %d: %s\n", i+1, msa->sqname[i]);
p7_trace_DumpAnnotated(stdout, trarr[i], gm, msa->ax[i]);
}
/* Create an MSA from the individual traces */
status = p7_tracealign_MSA(msa, trarr, hmm->M, p7_DEFAULT, &postmsa);
if (status != eslOK) esl_fatal("failed to create new MSA from traces\n");
esl_msafile_Write(stdout, postmsa, eslMSAFILE_PFAM);
p7_profile_Destroy(gm);
p7_hmm_Destroy(hmm);
p7_trace_DestroyArray(trarr, msa->nseq);
esl_msa_Destroy(postmsa);
esl_msa_Destroy(msa);
}
esl_msafile_Close(afp);
p7_bg_Destroy(bg);
esl_alphabet_Destroy(abc);
esl_getopts_Destroy(go);
return 0;
}
