/* Function: p7_Builder()
* Synopsis: Build a new HMM from an MSA.
*
* Purpose: Take the multiple sequence alignment <msa> and a build configuration <bld>,
* and build a new HMM.
*
* Effective sequence number determination and calibration steps require
* additionally providing a null model <bg>.
*
* Args: bld - build configuration
* msa - multiple sequence alignment
* bg - null model
* opt_hmm - optRETURN: new HMM
* opt_trarr - optRETURN: array of faux tracebacks, <0..nseq-1>
* opt_gm - optRETURN: profile corresponding to <hmm>
* opt_om - optRETURN: optimized profile corresponding to <gm>
* opt_postmsa - optRETURN: RF-annotated, possibly modified MSA
*
* Returns: <eslOK> on success. The new HMM is optionally returned in
* <*opt_hmm>, along with optional returns of an array of faux tracebacks
* for each sequence in <*opt_trarr>, the annotated MSA used to construct
* the model in <*opt_postmsa>, a configured search profile in
* <*opt_gm>, and an optimized search profile in <*opt_om>. These are
* all optional returns because the caller may, for example, be interested
* only in an optimized profile, or may only be interested in the HMM.
*
* Returns <eslENORESULT> if no consensus columns were annotated.
* Returns <eslEFORMAT> on MSA format problems, such as a missing RF annotation
* line in hand architecture construction. On any returned error,
* <bld->errbuf> contains an informative error message.
*
* Throws: <eslEMEM> on allocation error.
* <eslEINVAL> if relative weights couldn't be calculated from <msa>.
*
* Xref: J4/30.
*/
int
p7_Builder(P7_BUILDER *bld, ESL_MSA *msa, P7_BG *bg,
P7_HMM **opt_hmm, P7_TRACE ***opt_trarr, P7_PROFILE **opt_gm, P7_OPROFILE **opt_om,
ESL_MSA **opt_postmsa)
{
int i,j;
uint32_t checksum = 0; /* checksum calculated for the input MSA. hmmalign --mapali verifies against this. */
P7_HMM *hmm = NULL;
P7_TRACE **tr = NULL;
P7_TRACE ***tr_ptr = (opt_trarr != NULL || opt_postmsa != NULL) ? &tr : NULL;
int status;
if ((status = validate_msa (bld, msa)) != eslOK) goto ERROR;
if ((status = esl_msa_Checksum (msa, &checksum)) != eslOK) ESL_XFAIL(status, bld->errbuf, "Failed to calculate checksum");
if ((status = relative_weights (bld, msa)) != eslOK) goto ERROR;
if ((status = esl_msa_MarkFragments(msa, bld->fragthresh)) != eslOK) goto ERROR;
if ((status = build_model (bld, msa, &hmm, tr_ptr)) != eslOK) goto ERROR;
//Ensures that the weighted-average I->I count <= bld->max_insert_len
//(MI currently contains the number of observed insert-starts)
if (bld->max_insert_len>0)
for (i=1; i<hmm->M; i++ )
hmm->t[i][p7H_II] = ESL_MIN(hmm->t[i][p7H_II], bld->max_insert_len*hmm->t[i][p7H_MI]);
if ((status = effective_seqnumber (bld, msa, hmm, bg)) != eslOK) goto ERROR;
if ((status = parameterize (bld, hmm)) != eslOK) goto ERROR;
if ((status = annotate (bld, msa, hmm)) != eslOK) goto ERROR;
if ((status = calibrate (bld, hmm, bg, opt_gm, opt_om)) != eslOK) goto ERROR;
if ((status = make_post_msa (bld, msa, hmm, tr, opt_postmsa)) != eslOK) goto ERROR;
//force masked positions to background (it'll be close already, so no relevant impact on weighting)
if (hmm->mm != NULL)
for (i=1; i<hmm->M; i++ )
if (hmm->mm[i] == 'm')
for (j=0; j<hmm->abc->K; j++)
hmm->mat[i][j] = bg->f[j];
if ( bld->abc->type == eslDNA || bld->abc->type == eslRNA ) {
if (bld->w_len > 0) hmm->max_length = bld->w_len;
else if (bld->w_beta == 0.0) hmm->max_length = hmm->M *4;
else if ( (status = p7_Builder_MaxLength(hmm, bld->w_beta)) != eslOK) goto ERROR;
}
hmm->checksum = checksum;
hmm->flags |= p7H_CHKSUM;
if (opt_hmm != NULL) *opt_hmm = hmm; else p7_hmm_Destroy(hmm);
if (opt_trarr != NULL) *opt_trarr = tr; else p7_trace_DestroyArray(tr, msa->nseq);
return eslOK;
ERROR:
p7_hmm_Destroy(hmm);
p7_trace_DestroyArray(tr, msa->nseq);
if (opt_gm != NULL) p7_profile_Destroy(*opt_gm);
if (opt_om != NULL) p7_oprofile_Destroy(*opt_om);
return status;
}
int
main(int argc, char **argv)
{
int i,j;
ESL_GETOPTS *go = NULL; /* command line processing */
ESL_STOPWATCH *w = esl_stopwatch_Create();
int status;
ESL_MSA *msa = NULL;
FILE *ofp = NULL; /* output file (default is stdout) */
ESL_ALPHABET *abc = NULL; /* digital alphabet */
char *alifile; /* name of the alignment file we're building HMMs from */
ESLX_MSAFILE *afp = NULL; /* open alifile */
int infmt = eslMSAFILE_UNKNOWN; /* autodetect alignment format by default. */
int outfmt = eslMSAFILE_STOCKHOLM;
char *postmsafile; /* optional file to resave annotated, modified MSAs to */
FILE *postmsafp = NULL; /* open <postmsafile>, or NULL */
int mask_range_cnt = 0;
uint32_t mask_starts[100]; // over-the-top allocation.
uint32_t mask_ends[100];
char *rangestr;
char *range;
int *map = NULL; /* map[i]=j, means model position i comes from column j of the alignment; 1..alen */
int keep_mm;
/* Set processor specific flags */
impl_Init();
alifile = NULL;
postmsafile = NULL;
/* Parse the command line
*/
process_commandline(argc, argv, &go, &alifile, &postmsafile);
keep_mm = esl_opt_IsUsed(go, "--apendmask");
/* Initialize what we can in the config structure (without knowing the alphabet yet).
* Fields controlled by masters are set up in usual_master() or mpi_master()
* Fields used by workers are set up in mpi_worker()
*/
ofp = NULL;
infmt = eslMSAFILE_UNKNOWN;
afp = NULL;
abc = NULL;
if (esl_opt_IsOn(go, "--informat")) {
infmt = eslx_msafile_EncodeFormat(esl_opt_GetString(go, "--informat"));
if (infmt == eslMSAFILE_UNKNOWN) p7_Fail("%s is not a recognized input sequence file format\n", esl_opt_GetString(go, "--informat"));
}
/* Determine output alignment file format */
outfmt = eslx_msafile_EncodeFormat(esl_opt_GetString(go, "--outformat"));
if (outfmt == eslMSAFILE_UNKNOWN) p7_Fail(argv[0], "%s is not a recognized output MSA file format\n", esl_opt_GetString(go, "--outformat"));
/* Parse the ranges */
if (esl_opt_IsUsed(go, "--alirange")) {
esl_strdup(esl_opt_GetString(go, "--alirange"), -1, &rangestr) ;
} else if (esl_opt_IsUsed(go, "--modelrange")) {
esl_strdup(esl_opt_GetString(go, "--modelrange"), -1, &rangestr) ;
} else if (esl_opt_IsUsed(go, "--model2ali")) {
esl_strdup(esl_opt_GetString(go, "--model2ali"), -1, &rangestr) ;
} else if (esl_opt_IsUsed(go, "--ali2model")) {
esl_strdup(esl_opt_GetString(go, "--ali2model"), -1, &rangestr) ;
} else {
if (puts("Must specify mask range with --modelrange, --alirange, --model2ali, or --ali2model\n") < 0) ESL_XEXCEPTION_SYS(eslEWRITE, "write failed"); goto ERROR;
}
while ( (status = esl_strtok(&rangestr, ",", &range) ) == eslOK) {
status = esl_regexp_ParseCoordString(range, mask_starts + mask_range_cnt, mask_ends + mask_range_cnt );
if (status == eslESYNTAX) esl_fatal("range flags take coords <from>..<to>; %s not recognized", range);
if (status == eslFAIL) esl_fatal("Failed to find <from> or <to> coord in %s", range);
mask_range_cnt++;
}
/* Start timing. */
esl_stopwatch_Start(w);
/* Open files, set alphabet.
* afp - open alignment file for input
* abc - alphabet expected or guessed in ali file
* postmsafp - open MSA output file
* ofp - optional open output file, or stdout
*/
if (esl_opt_GetBoolean(go, "--amino")) abc = esl_alphabet_Create(eslAMINO);
else if (esl_opt_GetBoolean(go, "--dna")) abc = esl_alphabet_Create(eslDNA);
else if (esl_opt_GetBoolean(go, "--rna")) abc = esl_alphabet_Create(eslRNA);
else abc = NULL;
status = eslx_msafile_Open(&abc, alifile, NULL, infmt, NULL, &afp);
if (status != eslOK) eslx_msafile_OpenFailure(afp, status);
if (esl_opt_IsUsed(go, "--alirange") || esl_opt_IsUsed(go, "--modelrange") ) {
postmsafp = fopen(postmsafile, "w");
if (postmsafp == NULL) p7_Fail("Failed to MSA output file %s for writing", postmsafile);
}
if (esl_opt_IsUsed(go, "-o"))
{
ofp = fopen(esl_opt_GetString(go, "-o"), "w");
if (ofp == NULL) p7_Fail("Failed to open -o output file %s\n", esl_opt_GetString(go, "-o"));
}
else ofp = stdout;
/* Looks like the i/o is set up successfully...
* Initial output to the user
*/
output_header(go, ofp, alifile, postmsafile); /* cheery output header */
/* read the alignment */
if ((status = eslx_msafile_Read(afp, &msa)) != eslOK) eslx_msafile_ReadFailure(afp, status);
if (esl_opt_IsUsed(go, "--alirange") || esl_opt_IsUsed(go, "--modelrange") ) {
/* add/modify mmline for the mask */
if (msa->mm == NULL) {
ESL_ALLOC(msa->mm, msa->alen);
keep_mm = FALSE;
}
if (!keep_mm)
for (i=0; i<msa->alen; i++) msa->mm[i] = '.';
}
// convert model coordinates to alignment coordinates, if necessary
if (esl_opt_IsUsed(go, "--modelrange") || esl_opt_IsUsed(go, "--model2ali") || esl_opt_IsUsed(go, "--ali2model") ) {
float symfrac = esl_opt_GetReal(go, "--symfrac");
int do_hand = esl_opt_IsOn(go, "--hand");
int L;
//same as p7_builder relative_weights
if (esl_opt_IsOn(go, "--wnone") ) { esl_vec_DSet(msa->wgt, msa->nseq, 1.); }
else if (esl_opt_IsOn(go, "--wgiven") ) ;
else if (esl_opt_IsOn(go, "--wpb") ) status = esl_msaweight_PB(msa);
else if (esl_opt_IsOn(go, "--wgsc") ) status = esl_msaweight_GSC(msa);
else if (esl_opt_IsOn(go, "--wblosum")) status = esl_msaweight_BLOSUM(msa, esl_opt_GetReal(go, "--wid"));
if ((status = esl_msa_MarkFragments(msa, esl_opt_GetReal(go, "--fragthresh"))) != eslOK) goto ERROR;
//build a map of model mask coordinates to alignment coords
ESL_ALLOC(map, sizeof(int) * (msa->alen+1));
L = p7_Alimask_MakeModel2AliMap(msa, do_hand, symfrac, map );
if ( esl_opt_IsUsed(go, "--model2ali") ) {
//print mapping
printf ("model coordinates alignment coordinates\n");
for (i=0; i<mask_range_cnt; i++)
printf ("%8d..%-8d -> %8d..%-8d\n", mask_starts[i], mask_ends[i], map[mask_starts[i]-1], map[mask_ends[i]-1]);
/* If I wanted to, I could print all the map values independently:
printf("\n\n-----------\n");
printf("Map\n");
printf("---\n");
for (i=0; i<L; i++)
printf("%d -> %d\n", i+1, map[i]);
*/
} else if ( esl_opt_IsUsed(go, "--ali2model") ) {
//print mapping (requires scanning the inverted map
int alistart = 0;
int aliend = 0;
printf ("alignment coordinates model coordinates\n");
for (i=0; i<mask_range_cnt; i++) {
//find j for ali positions
while (map[alistart] < mask_starts[i] )
alistart++;
aliend = alistart;
while (map[aliend] < mask_ends[i] )
aliend++;
printf (" %8d..%-8d -> %8d..%-8d\n", map[alistart], map[aliend], alistart+1, aliend+1);
}
} else {
//convert the mask coords based on map
for (i=0; i<mask_range_cnt; i++) {
mask_starts[i] = map[mask_starts[i]-1]; //-1 because mmline is offset by one relative to the 1-base alignment
mask_ends[i] = map[mask_ends[i]-1];
}
}
}
if (esl_opt_IsUsed(go, "--alirange") || esl_opt_IsUsed(go, "--modelrange") ) {
//overwrite '.' with 'm' everywhere the range says to do it
for (i=0; i<mask_range_cnt; i++)
for (j=mask_starts[i]; j<=mask_ends[i]; j++)
msa->mm[j-1] = 'm';
if ((status = eslx_msafile_Write(postmsafp, msa, outfmt)) != eslOK) ESL_XEXCEPTION_SYS(eslEWRITE, "write failed");
}
esl_stopwatch_Stop(w);
if (esl_opt_IsOn(go, "-o")) fclose(ofp);
if (postmsafp) fclose(postmsafp);
if (afp) eslx_msafile_Close(afp);
if (abc) esl_alphabet_Destroy(abc);
esl_getopts_Destroy(go);
esl_stopwatch_Destroy(w);
return 0;
ERROR:
return eslFAIL;
}