/* Function: p7_builder_SetScoreSystem()
* Synopsis: Initialize score system for single sequence queries.
*
* Purpose: Initialize the builder <bld> to be able to parameterize
* single sequence queries, using a substitution matrix
* from a file.
*
* Read a standard substitution score matrix from file
* <mxfile>. If <mxfile> is <NULL>, default to BLOSUM62
* scores. If <mxfile> is "-", read score matrix from
* <stdin> stream. If <env> is non-<NULL> and <mxfile> is
* not found in the current working directory, look for
* <mxfile> in colon-delimited directory list contained in
* environment variable <env>.
*
* Set the gap-open and gap-extend probabilities to
* <popen>, <pextend>, respectively.
*
* Use background residue frequencies in the null model
* <bg> to convert substitution matrix scores to
* conditional probability parameters.
*
* Args: bld - <P7_BUILDER> to initialize
* mxfile - score matrix file to use, or NULL for BLOSUM62 default
* env - env variable containing directory list where <mxfile> may reside
* popen - gap open probability
* pextend - gap extend probability
* bg - null model, containing background frequencies
*
* Returns: <eslOK> on success.
*
* <eslENOTFOUND> if <mxfile> can't be found or opened, even
* in any of the directories specified by the <env> variable.
*
* <eslEINVAL> if the score matrix can't be converted into
* conditional probabilities; for example, if it has no valid
* solution for <lambda>.
*
* On either error, <bld->errbuf> contains a useful error message
* for the user.
*
* Throws: <eslEMEM> on allocation failure.
*/
int
p7_builder_SetScoreSystem(P7_BUILDER *bld, const char *mxfile, const char *env, double popen, double pextend, P7_BG *bg)
{
ESL_FILEPARSER *efp = NULL;
double *f = NULL;
double slambda;
int status;
bld->errbuf[0] = '\0';
/* If a score system is already set, delete it. */
if (bld->S != NULL) esl_scorematrix_Destroy(bld->S);
if (bld->Q != NULL) esl_dmatrix_Destroy(bld->Q);
/* Get the scoring matrix */
if ((bld->S = esl_scorematrix_Create(bld->abc)) == NULL) { status = eslEMEM; goto ERROR; }
if (mxfile == NULL)
{
if (bld->abc->type == eslAMINO) {
if ((status = esl_scorematrix_Set("BLOSUM62", bld->S)) != eslOK) goto ERROR;
} else {
if ((status = esl_scorematrix_Set("DNA1", bld->S)) != eslOK) goto ERROR;
}
}
else
{
if ((status = esl_fileparser_Open(mxfile, env, &efp)) != eslOK) ESL_XFAIL(status, bld->errbuf, "Failed to find or open matrix file %s", mxfile);
if ((status = esl_scorematrix_Read(efp, bld->abc, &(bld->S))) != eslOK) ESL_XFAIL(status, bld->errbuf, "Failed to read matrix from %s:\n%s", mxfile, efp->errbuf);
esl_fileparser_Close(efp);
efp = NULL;
}
/* A wasteful conversion of the HMMER single-precision background probs to Easel double-prec */
ESL_ALLOC(f, sizeof(double) * bg->abc->K);
esl_vec_F2D(bg->f, bg->abc->K, f);
/* Backcalculate joint probability matrix Q, given scores S and background freqs bg->f. */
status = esl_scorematrix_ProbifyGivenBG(bld->S, f, f, &slambda, &(bld->Q));
if (status == eslEINVAL) ESL_XFAIL(eslEINVAL, bld->errbuf, "input score matrix %s has no valid solution for lambda", mxfile);
else if (status == eslENOHALT) ESL_XFAIL(eslEINVAL, bld->errbuf, "failed to solve input score matrix %s for lambda: are you sure it's valid?", mxfile);
else if (status != eslOK) ESL_XFAIL(eslEINVAL, bld->errbuf, "unexpected error in solving input score matrix %s for probability parameters", mxfile);
/* Convert joint probabilities P(ab) to conditionals P(b|a) */
esl_scorematrix_JointToConditionalOnQuery(bld->abc, bld->Q);
bld->popen = popen;
bld->pextend = pextend;
free(f);
return eslOK;
ERROR:
if (efp) esl_fileparser_Close(efp);
if (f) free(f);
return status;
}
/* Function: p7_builder_LoadScoreSystem()
* Synopsis: Load a standard score system for single sequence queries.
*
* Purpose: Initialize the builder <bld> to be able to parameterize
* single sequence queries, using the standard (built-in) score
* matrix named <mx>.
*
* Available score matrices <mx> include PAM30, 70, 120, and 240;
* and BLOSUM45, 50, 62, 80, and 90. See <esl_scorematrix.c>.
*
* Set the gap-open and gap-extend probabilities to
* <popen>, <pextend>, respectively.
*
* Use background residue frequencies in the null model
* <bg> to convert substitution matrix scores to
* conditional probability parameters.
*
* Args: bld - <P7_BUILDER> to initialize
* matrix - score matrix file to use
* popen - gap open probability
* pextend - gap extend probability
* bg - null model, containing background frequencies
*
* Returns: <eslOK> on success.
*
* <eslENOTFOUND> if <mxfile> can't be found or opened, even
* in any of the directories specified by the <env> variable.
*
* <eslEINVAL> if the score matrix can't be converted into
* conditional probabilities; for example, if it has no valid
* solution for <lambda>.
*
* On either error, <bld->errbuf> contains a useful error message
* for the user.
*
* Throws: <eslEMEM> on allocation failure.
*/
int
p7_builder_LoadScoreSystem(P7_BUILDER *bld, const char *matrix, double popen, double pextend, P7_BG *bg)
{
double *f = NULL;
double slambda;
int status;
bld->errbuf[0] = '\0';
/* If a score system is already set, delete it. */
if (bld->S != NULL) esl_scorematrix_Destroy(bld->S);
if (bld->Q != NULL) esl_dmatrix_Destroy(bld->Q);
/* Get the scoring matrix */
if ((bld->S = esl_scorematrix_Create(bld->abc)) == NULL) { status = eslEMEM; goto ERROR; }
status = esl_scorematrix_Set(matrix, bld->S);
if (status == eslENOTFOUND) ESL_XFAIL(status, bld->errbuf, "no matrix named %s is available as a built-in", matrix);
else if (status != eslOK) ESL_XFAIL(status, bld->errbuf, "failed to set score matrix %s as a built-in", matrix);
/* A wasteful conversion of the HMMER single-precision background probs to Easel double-prec */
ESL_ALLOC(f, sizeof(double) * bg->abc->K);
esl_vec_F2D(bg->f, bg->abc->K, f);
/* Backcalculate joint probability matrix Q, given scores S and background freqs bg->f. */
/* Failures shouldn't happen here: these are standard matrices. */
status = esl_scorematrix_ProbifyGivenBG(bld->S, f, f, &slambda, &(bld->Q));
if (status == eslEINVAL) ESL_XFAIL(eslEINVAL, bld->errbuf, "built-in score matrix %s has no valid solution for lambda", matrix);
else if (status == eslENOHALT) ESL_XFAIL(eslEINVAL, bld->errbuf, "failed to solve score matrix %s for lambda", matrix);
else if (status != eslOK) ESL_XFAIL(eslEINVAL, bld->errbuf, "unexpected error in solving score matrix %s for probability parameters", matrix);
/* Convert joint probabilities P(ab) to conditionals P(b|a) */
esl_scorematrix_JointToConditionalOnQuery(bld->abc, bld->Q);
bld->popen = popen;
bld->pextend = pextend;
free(f);
return eslOK;
ERROR:
if (f) free(f);
return status;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_CreateDefaultApp(options, 2, argc, argv, banner, usage);
ESL_ALPHABET *abc = esl_alphabet_Create(eslAMINO);
char *qfile = esl_opt_GetArg(go, 1);
char *tfile = esl_opt_GetArg(go, 2);
ESL_SQFILE *qfp = NULL;
ESL_SQFILE *tfp = NULL;
ESL_SQ *qsq = esl_sq_CreateDigital(abc);
ESL_SQ *tsq = esl_sq_CreateDigital(abc);
ESL_SCOREMATRIX *S = esl_scorematrix_Create(abc);
ESL_DMATRIX *Q = NULL;
P7_BG *bg = p7_bg_Create(abc);
P7_HMM *hmm = NULL;
P7_PROFILE *gm = NULL;
P7_REFMX *vit = p7_refmx_Create(200, 400); /* will grow as needed */
double *fa = malloc(sizeof(double) * abc->K);
double popen = 0.02;
double pextend = 0.4;
double lambda;
float vsc;
float nullsc;
int status;
esl_composition_BL62(fa);
esl_vec_D2F(fa, abc->K, bg->f);
esl_scorematrix_Set("BLOSUM62", S);
esl_scorematrix_ProbifyGivenBG(S, fa, fa, &lambda, &Q);
esl_scorematrix_JointToConditionalOnQuery(abc, Q);
if (esl_sqfile_OpenDigital(abc, qfile, eslSQFILE_UNKNOWN, NULL, &qfp) != eslOK) esl_fatal("failed to open %s", qfile);
if (esl_sqio_Read(qfp, qsq) != eslOK) esl_fatal("failed to read query seq");
p7_Seqmodel(abc, qsq->dsq, qsq->n, qsq->name, Q, bg->f, popen, pextend, &hmm);
p7_hmm_SetComposition(hmm);
p7_hmm_SetConsensus(hmm, qsq);
gm = p7_profile_Create(hmm->M, abc);
p7_profile_ConfigUnilocal(gm, hmm, bg, 400);
if (esl_sqfile_OpenDigital(abc, tfile, eslSQFILE_UNKNOWN, NULL, &tfp) != eslOK) esl_fatal("failed to open %s", tfile);
while ((status = esl_sqio_Read(tfp, tsq)) == eslOK)
{
p7_bg_SetLength (bg, tsq->n);
p7_profile_SetLength(gm, tsq->n);
p7_ReferenceViterbi(tsq->dsq, tsq->n, gm, vit, NULL, &vsc);
p7_bg_NullOne(bg, tsq->dsq, tsq->n, &nullsc);
printf("%.4f %-25s %-25s\n", (vsc - nullsc) / eslCONST_LOG2, tsq->name, gm->name);
esl_sq_Reuse(tsq);
p7_refmx_Reuse(vit);
}
p7_refmx_Destroy(vit);
p7_profile_Destroy(gm);
p7_hmm_Destroy(hmm);
p7_bg_Destroy(bg);
esl_dmatrix_Destroy(Q);
esl_scorematrix_Destroy(S);
free(fa);
esl_sq_Destroy(qsq);
esl_sq_Destroy(tsq);
esl_sqfile_Close(qfp);
esl_sqfile_Close(tfp);
esl_alphabet_Destroy(abc);
esl_getopts_Destroy(go);
return 0;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = p7_CreateDefaultApp(options, 2, argc, argv, banner, usage);
ESL_ALPHABET *abc = esl_alphabet_Create(eslAMINO);
char *hmmfile = esl_opt_GetArg(go, 1);
char *qfile = esl_opt_GetArg(go, 2);
ESL_SQ *qsq = esl_sq_CreateDigital(abc);
ESL_SQFILE *qfp = NULL;
FILE *hmmfp = NULL;
ESL_SCOREMATRIX *S = esl_scorematrix_Create(abc);
ESL_DMATRIX *Q = NULL;
P7_BG *bg = p7_bg_Create(abc);
P7_HMM *hmm = NULL;
double *fa = NULL;
double popen = esl_opt_GetReal (go, "-q");
double pextend = esl_opt_GetReal (go, "-r");
char *mxfile = esl_opt_GetString(go, "-m");
char errbuf[eslERRBUFSIZE];
double slambda;
int a,b;
int status;
/* Reverse engineer a scoring matrix to obtain conditional prob's
* that we'll use for the single-seq query HMM. Because score mx is
* symmetric, we can set up P[a][b] = P(b | a), so we can use the
* matrix rows as HMM match emission vectors. This means dividing
* the joint probs through by f_a.
*/
if (mxfile == NULL) {
if (esl_scorematrix_Set("BLOSUM62", S) != eslOK) esl_fatal("failed to set BLOSUM62 scores");
} else {
ESL_FILEPARSER *efp = NULL;
if ( esl_fileparser_Open(mxfile, NULL, &efp) != eslOK) esl_fatal("failed to open score file %s", mxfile);
if ( esl_scorematrix_Read(efp, abc, &S) != eslOK) esl_fatal("failed to read matrix from %s", mxfile);
esl_fileparser_Close(efp);
}
/* A wasteful conversion of the HMMER single-precision background probs to Easel double-prec */
ESL_ALLOC(fa, sizeof(double) * bg->abc->K);
esl_vec_F2D(bg->f, bg->abc->K, fa);
/* Backcalculate joint probabilities Q, given score matrix S and background frequencies fa */
status = esl_scorematrix_ProbifyGivenBG(S, fa, fa, &slambda, &Q);
if (status == eslEINVAL) esl_fatal("built-in score matrix %s has no valid solution for lambda", matrix);
else if (status == eslENOHALT) esl_fatal("failed to solve score matrix %s for lambda", matrix);
else if (status != eslOK) esl_fatal("unexpected error in solving score matrix %s for probability parameters", matrix);
esl_scorematrix_JointToConditionalOnQuery(abc, Q);
/* Open the query sequence file in FASTA format */
status = esl_sqfile_Open(qfile, eslSQFILE_FASTA, NULL, &qfp);
if (status == eslENOTFOUND) esl_fatal("No such file %s.", qfile);
else if (status == eslEFORMAT) esl_fatal("Format of %s unrecognized.", qfile);
else if (status == eslEINVAL) esl_fatal("Can't autodetect stdin or .gz.");
else if (status != eslOK) esl_fatal("Open of %s failed, code %d.", qfile, status);
/* Open the output HMM file */
if ((hmmfp = fopen(hmmfile, "w")) == NULL) esl_fatal("Failed to open output HMM file %s", hmmfile);
/* For each sequence, build a model and save it.
*/
while ((status = esl_sqio_Read(qfp, qsq)) == eslOK)
{
p7_Seqmodel(abc, qsq->dsq, qsq->n, qsq->name, Q, bg->f, popen, pextend, &hmm);
if ( p7_hmm_Validate(hmm, errbuf, 1e-5) != eslOK) esl_fatal("HMM validation failed: %s\n", errbuf);
if ( p7_hmmfile_WriteASCII(hmmfp, -1, hmm) != eslOK) esl_fatal("HMM save failed");
p7_hmm_Destroy(hmm);
}
if (status == eslEFORMAT) esl_fatal("Parse failed (sequence file %s line %" PRId64 "):\n%s\n",
qfp->filename, qfp->linenumber, qfp->errbuf);
else if (status != eslEOF) esl_fatal("Unexpected error %d reading sequence file %s",
status, qfp->filename);
esl_dmatrix_Destroy(Q);
esl_scorematrix_Destroy(S);
free(fa);
free(fb);
esl_sq_Destroy(qsq);
esl_sqfile_Close(qfp);
fclose(hmmfp);
esl_alphabet_Destroy(abc);
esl_getopts_Destroy(go);
return 0;
}
