int
main(int argc, char **argv)
{
ESL_GETOPTS *go = NULL; /* application configuration */
ESL_ALPHABET *abc = NULL; /* biological alphabet */
char *alifile = NULL; /* alignment file name */
int fmt; /* format code for alifiles */
ESL_MSAFILE *afp = NULL; /* open alignment file */
ESL_MSA *msa = NULL; /* multiple sequence alignment */
int status; /* easel return code */
int do_info = TRUE; /* TRUE if -i */
int do_max = FALSE; /* TRUE if -x */
int do_ffreq = FALSE; /* TRUE if --ffreq */
int do_fmin = FALSE; /* TRUE if --fmin */
float fthresh = 0.; /* <x> from -f <x> */
int do_remove_bps = FALSE; /* TRUE if -r */
int do_consistent = FALSE; /* TRUE if -c */
int do_indi2cons = FALSE; /* TRUE if --indi <x> */
int have_cons; /* TRUE if first alignment has consensus sequence */
int do_newcons = FALSE; /* TRUE if we're creating a new consensus structure
* and outputing a new alignment (if -x -f -c or --indi)
*/
int do_a = FALSE; /* TRUE if -a */
char *indi; /* for <x> from --indi <x> */
int nindi_read; /* number of individual sequence SS lines we've read for current alignment */
int a; /* counter over seqs */
int i, i2; /* counter over residues */
int j, j2; /* counter over residues */
int nali; /* counter over alignments */
int **bp = NULL; /* bp[i][j] is number of individual bps exist between aln cols i and j */
int *cur_ct = NULL; /* ct array of basepairs for current sequence */
int *cons_ct = NULL; /* ct array of basepairs for SS_cons being created */
int *xcons_ct = NULL; /* ct array of basepairs for existing SS_cons */
int *ngaps = NULL; /* number of gaps in each alignment position */
FILE *ofp; /* output file (default is stdout) */
int be_verbose = FALSE; /* TRUE to print extra info */
int seqthresh; /* sequence number threshold for defining a bp as consensus (int) ((fthresh * nseq) + 0.5)*/
char *sscons = NULL; /* the new SS_cons line */
FILE *lfp = NULL; /* file to list sequences with conflicting bps to */
int nlist = 0; /* number of sequences listed to list file */
int *nconflictsA; /* number of conflicting bps in seq a's individual structure annotation */
int nconflicts_total = 0; /* total number of conflicts */
int nconflicts_list = 0; /* total number of conflicts in sequences listed to file <x> from -l <x> */
int noverlaps_total = 0; /* total number of overlaps */
int nconsistent_total = 0; /* total number of consistent bps */
int nbps_total = 0; /* total number of bps */
int *nconsistentA; /* number of consistent bps in seq a's individual structure annotation */
int *noverlapsA; /* number of bps in seq a's indi structure that overlap with consensus structure */
int *nbpsA; /* number of bps in seq a's indi structure that overlap with consensus structure */
int ncons_bps = 0; /* number of bps in consensus structure */
int max_noverlaps_aidx;
int max_nconsistent_aidx;
int max_nbps_aidx;
int *removebp; /* removebp[i] is TRUE remove consensus bp [i]:xcons_ct[i] */
int *has_conflict;
int *nmates_l2r; /* half matrix, nmate_l2r[i] = <x>, i < nmate_l2r[i], there are <x> different right mates j for i */
int *nmates_r2l; /* half matrix, nmate_r2l[j] = <x>, j < nmate_r2l[j], there are <x> different left mates i for j */
int lmax; /* with -l, maximum number of conflicts to allow */
int namewidth = 18; /* length of 'SS_cons(consensus)' */
char *namedashes = NULL; /* to store underline for seq name */
/* --fmin related variables */
int nbps = 0;
int prev_nbps = -1;
float fmin;
int inconsistent_flag;
int pknot_flag;
int k,l;
/***********************************************
* Parse command line
***********************************************/
go = esl_getopts_Create(options);
if (esl_opt_ProcessCmdline(go, argc, argv) != eslOK ||
esl_opt_VerifyConfig(go) != eslOK)
{
printf("Failed to parse command line: %s\n", go->errbuf);
esl_usage(stdout, argv[0], usage);
printf("\nTo see more help on available options, do %s -h\n\n", argv[0]);
exit(1);
}
if (esl_opt_GetBoolean(go, "-h") )
{
esl_banner(stdout, argv[0], banner);
esl_usage (stdout, argv[0], usage);
puts("\nwhere basic options are:");
esl_opt_DisplayHelp(stdout, go, 1, 2, 80);
puts("\noptions for defining a new consensus structure (all of these require -o):");
esl_opt_DisplayHelp(stdout, go, 2, 2, 80);
puts("\noptions for listing sequences based on structure:");
esl_opt_DisplayHelp(stdout, go, 3, 2, 80);
exit(0);
}
if (esl_opt_ArgNumber(go) != 1)
{
printf("Incorrect number of command line arguments.\n");
esl_usage(stdout, argv[0], usage);
printf("\nTo see more help on available options, do %s -h\n\n", argv[0]);
exit(1);
}
alifile = esl_opt_GetArg(go, 1);
fmt = eslMSAFILE_STOCKHOLM;
/***********************************************
* Open the MSA file; determine alphabet; set for digital input
***********************************************/
if (esl_opt_GetBoolean(go, "--dna")) abc = esl_alphabet_Create(eslDNA);
else if (esl_opt_GetBoolean(go, "--rna")) abc = esl_alphabet_Create(eslRNA);
if ( (status = esl_msafile_Open(&abc, alifile, NULL, fmt, NULL, &afp)) != eslOK)
esl_msafile_OpenFailure(afp, status);
/* open output file */
if (esl_opt_GetString(go, "-o") != NULL) {
if ((ofp = fopen(esl_opt_GetString(go, "-o"), "w")) == NULL)
esl_fatal("Failed to open -o output file %s\n", esl_opt_GetString(go, "-o"));
} else ofp = NULL;
if (esl_opt_GetString(go, "-l") != NULL) {
if ((lfp = fopen(esl_opt_GetString(go, "-l"), "w")) == NULL)
esl_fatal("Failed to open -l output file %s\n", esl_opt_GetString(go, "-l"));
}
/* determine if we're creating a structure */
do_max = esl_opt_GetBoolean(go, "-x");
if(!(esl_opt_IsDefault(go, "--ffreq"))) {
do_ffreq = TRUE;
fthresh = esl_opt_GetReal(go, "--ffreq");
}
if(!(esl_opt_IsDefault(go, "--fmin"))) {
do_fmin = TRUE;
}
do_remove_bps = esl_opt_GetBoolean(go, "-r");
do_consistent = esl_opt_GetBoolean(go, "-c");
if(!(esl_opt_IsDefault(go, "--indi"))) {
do_indi2cons = TRUE;
}
if(do_max || do_ffreq || do_fmin || do_remove_bps || do_consistent || do_indi2cons) {
do_newcons = TRUE;
}
do_a = esl_opt_GetBoolean(go, "-a");
if(do_a || do_max || do_ffreq || do_fmin || do_remove_bps || do_consistent || do_indi2cons) {
do_info = FALSE;
}
/***********************************************
* Read MSAs one at a time.
***********************************************/
nali = 0;
have_cons = FALSE;
lmax = esl_opt_GetInteger(go, "--lmax");
if(esl_opt_GetBoolean(go, "-v")) be_verbose = TRUE;
while ((status = esl_msafile_Read(afp, &msa)) != eslEOF)
{
if (status != eslOK) esl_msafile_ReadFailure(afp, status);
nali++;
/* determine max length name */
namewidth = 18; /* length of 'SS_cons(consensus)' */
for(i = 0; i < msa->nseq; i++) namewidth = ESL_MAX(namewidth, strlen(msa->sqname[i]));
if(namedashes != NULL) { free(namedashes); }
ESL_ALLOC(namedashes, sizeof(char) * namewidth+1);
namedashes[namewidth] = '\0';
for(i = 0; i < namewidth; i++) namedashes[i] = '-';
ESL_ALLOC(sscons, sizeof(char) * (msa->alen+1));
ESL_ALLOC(cur_ct, sizeof(int) * (msa->alen+1));
ESL_ALLOC(cons_ct, sizeof(int) * (msa->alen+1));
ESL_ALLOC(xcons_ct, sizeof(int) * (msa->alen+1));
ESL_ALLOC(bp, sizeof(int *) * (msa->alen+1));
ESL_ALLOC(removebp, sizeof(int) * (msa->alen+1));
ESL_ALLOC(has_conflict, sizeof(int) * (msa->alen+1));
ESL_ALLOC(nmates_l2r, sizeof(int) * (msa->alen+1));
ESL_ALLOC(nmates_r2l, sizeof(int) * (msa->alen+1));
esl_vec_ISet(cur_ct, (msa->alen+1), 0);
esl_vec_ISet(cons_ct, (msa->alen+1), 0);
esl_vec_ISet(xcons_ct, (msa->alen+1), 0);
esl_vec_ISet(removebp, (msa->alen+1), FALSE);
esl_vec_ISet(has_conflict, (msa->alen+1), FALSE);
esl_vec_ISet(nmates_l2r, (msa->alen+1), 0);
esl_vec_ISet(nmates_r2l, (msa->alen+1), 0);
ESL_ALLOC(nconflictsA, sizeof(int) * msa->nseq);
ESL_ALLOC(noverlapsA, sizeof(int) * msa->nseq);
ESL_ALLOC(nconsistentA, sizeof(int) * msa->nseq);
ESL_ALLOC(nbpsA, sizeof(int) * msa->nseq);
esl_vec_ISet(nconflictsA, msa->nseq, 0);
esl_vec_ISet(noverlapsA, msa->nseq, 0);
esl_vec_ISet(nconsistentA, msa->nseq, 0);
esl_vec_ISet(nbpsA, msa->nseq, 0);
max_noverlaps_aidx = max_nconsistent_aidx = max_nbps_aidx = 0;
nconsistent_total = nbps_total = noverlaps_total = nconflicts_total = nconflicts_list = 0;
for(i = 1; i <= msa->alen; i++) {
ESL_ALLOC(bp[i], sizeof(int) * (msa->alen+1));
esl_vec_ISet(bp[i], (msa->alen+1), 0);
}
/* make sure we have ss_cons and indi ss if we need it */
if(msa->ss_cons == NULL && do_remove_bps) esl_fatal("-r requires all alignments have SS_cons annotation, alignment %d does not.", nali);
if(msa->ss == NULL && do_max) esl_fatal("-x requires all alignments have individual SS annotation, alignment %d does not.", nali);
if(msa->ss == NULL && do_consistent) esl_fatal("-c requires all alignments have individual SS annotation, alignment %d does not.", nali);
if(msa->ss == NULL && do_indi2cons) esl_fatal("--indi requires all alignments have individual SS annotation, alignment %d does not.", nali);
if(msa->ss == NULL && do_ffreq) esl_fatal("--ffreq requires all alignments have individual SS annotation, alignment %d does not.", nali);
if(msa->ss == NULL && do_fmin) esl_fatal("--fmin requires all alignments have individual SS annotation, alignment %d does not.", nali);
if(msa->ss_cons != NULL) {
if((status = esl_wuss2ct(msa->ss_cons, msa->alen, xcons_ct)) != eslOK) {
esl_fatal("Existing SS_cons for alignment %d is invalid.", nali);
}
ncons_bps = 0;
for(i = 1; i <= msa->alen; i++)
if(xcons_ct[i] != 0 && i < xcons_ct[i])
ncons_bps++;
if(nali > 1 && !have_cons)
esl_fatal("the first aln has SS_cons but aln %d lacks it, if one has it, they all must.", nali);
if(nali == 1) have_cons = TRUE;
}
else if (lfp != NULL) {
esl_fatal("the -l option requires existing SS_cons annotation, aln %d lacks it.", nali);
}
else if (do_remove_bps) {
esl_fatal("the -r option requires existing SS_cons annotation, aln %d lacks it.", nali);
}
else if (do_consistent) {
esl_fatal("the -c option requires existing SS_cons annotation, aln %d lacks it.", nali);
}
else {
if(nali > 1 && have_cons)
esl_fatal("the first aln does not have SS_cons but aln %d does, if one has it, they all must.", nali);
}
if(do_info) {
printf("# Per-sequence basepair information:\n");
printf("# Alignment file: %s\n", alifile);
printf("# Alignment idx: %d\n", nali);
if(msa->name != NULL) { printf("# Alignment name: %s\n", msa->name); }
if(have_cons) {
printf("#\n");
printf("# indibp: number of basepairs in the individual sequence SS annotation\n");
printf("# ovrlap: number of indibp basepairs that also exist as consensus basepairs\n");
printf("# cnsist: number of indibp basepairs that do not conflict with any consensus basepairs\n");
printf("# cnflct: number of indibp basepairs that conflict with >= 1 consensus basepairs\n");
printf("#\n");
printf("# A conflict exists between two basepairs in different structures, one between columns i and j\n");
printf("# and the other between columns k and l, if (i == k and j != l) or (j == l and i != k).\n");
printf("#\n");
printf("# %-*s %6s %6s %6s %6s\n", namewidth, "seqname", "indibp", "ovrlap", "cnsist", "cnflct");
printf("# %-*s %6s %6s %6s %6s\n", namewidth, namedashes, "------", "------", "-----", "------");
}
else {
printf("# %-*s %6s\n", namewidth, "seqname", "nbp");
printf("# %-*s %6s\n", namewidth, namedashes, "------");
}
}
nindi_read = 0;
for (a = 0; a < msa->nseq; a++) {
if(msa->ss != NULL && msa->ss[a] != NULL) {
if((status = esl_wuss2ct(msa->ss[a], msa->alen, cur_ct)) != eslOK) {
esl_fatal("SS annotation for sequence %d, aln %d is invalid.\n", (a+1), nali);
}
nindi_read++;
for(i = 1; i <= msa->alen; i++) {
if(i < cur_ct[i]) {
bp[i][cur_ct[i]]++;
if(bp[i][cur_ct[i]] == 1) {
nmates_l2r[i]++;
nmates_r2l[cur_ct[i]]++;
}
}
}
for(i = 1; i <= msa->alen; i++) {
if(cur_ct[i] != 0 && i < cur_ct[i]) {
if(xcons_ct[i] == cur_ct[i]) noverlapsA[a]++;
if((xcons_ct[i] != 0) && (xcons_ct[i] != cur_ct[i])) {
if(be_verbose) { printf("ali: %2d seq %3d (%s) bp %4d:%4d conflicts with consensus bp %4d:%4d\n", nali, a, msa->sqname[a], i, cur_ct[i], i, xcons_ct[i]); }
nconflictsA[a]++;
/* indi bp i:cur_ct[i] conflicts with i:xcons_ct[i] */
removebp[i] = TRUE;
removebp[xcons_ct[i]] = TRUE;
}
else if((xcons_ct[cur_ct[i]] != 0) && (xcons_ct[cur_ct[i]] != i) && (cur_ct[xcons_ct[cur_ct[i]]] == 0)) {
if(be_verbose) { printf("ali: %2d seq %3d (%s) bp %4d:%4d conflicts with consensus bp %4d:%4d\n", nali, a, msa->sqname[a], xcons_ct[i], cur_ct[xcons_ct[i]], xcons_ct[cur_ct[i]], cur_ct[i]); }
nconflictsA[a]++;
/* indi bp i:cur_ct[i] conflicts with xcons_ct[cur_ct[i]]:cur_ct[i] */
removebp[cur_ct[i]] = TRUE;
removebp[xcons_ct[cur_ct[i]]] = TRUE;
}
else nconsistentA[a]++;
}
}
if(nconflictsA[a] > lmax) {
if(lfp != NULL) fprintf(lfp, "%s\n", msa->sqname[a]);
nconflicts_list += nconflictsA[a];
nlist++;
}
nbpsA[a] = nconflictsA[a] + nconsistentA[a];
nconflicts_total += nconflictsA[a];
nconsistent_total += nconsistentA[a];
noverlaps_total += noverlapsA[a];
nbps_total += nbpsA[a];
if(do_info && have_cons) printf(" %-*s %6d %6d %6d %6d\n", namewidth, msa->sqname[a], nbpsA[a], noverlapsA[a], nconsistentA[a], nconflictsA[a]);
if(do_info && !have_cons) printf(" %-*s %6d\n", namewidth, msa->sqname[a], nbpsA[a]);
if(nbpsA[a] > nbpsA[max_nbps_aidx]) max_nbps_aidx = a;
if((noverlapsA[a] > noverlapsA[max_noverlaps_aidx]) || ((noverlapsA[a] == noverlapsA[max_noverlaps_aidx]) && (nbpsA[a] > nbpsA[max_noverlaps_aidx]))) max_noverlaps_aidx = a;
if((nconsistentA[a] > nconsistentA[max_nconsistent_aidx]) || ((nconsistentA[a] == nconsistentA[max_nconsistent_aidx]) && (nbpsA[a] > nbpsA[max_nconsistent_aidx]))) max_nconsistent_aidx = a;
}
else if(do_newcons || esl_opt_GetBoolean(go, "-a")) { esl_fatal("No SS annotation for sequence %d, aln %d.\n", (a+1), nali); }
}
if(do_info && have_cons) {
if(nindi_read > 0) printf("\n");
printf(" %-*s %6d %6d %6d %6d\n", namewidth, "SS_cons(consensus)", ncons_bps, ncons_bps, ncons_bps, 0);
if(nindi_read > 0) {
printf("\n# %6d/%6d (%.3f) overlap\n", noverlaps_total, nbps_total, nbps_total > 0 ? (float) noverlaps_total / (float) nbps_total : 0.);
printf("# %6d/%6d (%.3f) consistent\n", nconsistent_total, nbps_total, nbps_total > 0 ? (float) nconsistent_total / (float) nbps_total: 0.);
printf("# %6d/%6d (%.3f) conflict\n", nconflicts_total, nbps_total, nbps_total > 0 ? (float) nconflicts_total / (float) nbps_total: 0.);
}
else {
printf("# No sequences in the alignment have GR SS annotation.\n");
}
}
if(lfp != NULL) {
printf("# %d/%d sequences with %.3f individual bps on avg that conflict with SS_cons written to %s\n", nlist, msa->nseq, (float) nconflicts_list / (float) nlist, esl_opt_GetString(go, "-l"));
}
/* determine number of gaps per alignment column */
if((status = get_gaps_per_column(msa, &ngaps)) != eslOK) goto ERROR;
/* -x: determine max bp structure OR
* -a: list all conflicts in individual structures */
if(do_max || do_a) {
for(i = 1; i <= msa->alen; i++) {
if(nmates_l2r[i] > 1) {/* list the conflicts */
has_conflict[i] = TRUE;
for(j = 1; j <= msa->alen; j++) {
if(bp[i][j] > 0) {
if(do_a) printf("More than 1 right mates for left mate %4d %4d:%4d bp exists in %4d/%4d seqs (%.3f)\n", i, i, j, bp[i][j], msa->nseq - ngaps[i], (float) bp[i][j] / (float) (msa->nseq - ngaps[i]));
has_conflict[j] = TRUE;
}
}
}
}
for(i = 1; i <= msa->alen; i++) {
if(nmates_r2l[i] > 1) {/* list the conflicts */
has_conflict[i] = TRUE;
for(j = 1; j <= msa->alen; j++) {
if(bp[j][i] > 0) {
if(do_a) printf("More than 1 left mates for right mate %4d %4d:%4d bp exists in %4d/%4d seqs (%.3f)\n", i, j, i, bp[j][i], msa->nseq - ngaps[i], (float) bp[j][i] / (float) (msa->nseq - ngaps[i]));
has_conflict[j] = TRUE;
}
}
}
}
for(i = 1; i <= msa->alen; i++) {
/*printf("conflict[%4d]: %d\n", i, has_conflict[i]);*/
if(nmates_l2r[i] == 1 && (!(has_conflict[i]))) {
j = i+1;
while(bp[i][j] == 0) j++;
cons_ct[i] = j;
cons_ct[j] = i;
}
}
/* remove pseudoknotted bps greedily */
for(i = 1; i <= msa->alen; i++) {
j = cons_ct[i];
if(j != 0 && i < j) {
for(i2 = i+1; i2 <= msa->alen; i2++) {
j2 = cons_ct[i2];
if(j2 != 0 && i2 < j2) {
if((i2 < j) && (j < j2)) {
/*printf("KNOT %4d:%4d (%4d) %4d:%4d (%4d)\n", i, j, bp[i][j], i2, j2, bp[i2][j2]);*/
/* note: remove both if they have equal number of sequences */
if(bp[i][j] <= bp[i2][j2]) {
/*printf("rm %4d:%4d\n", i, j);*/
cons_ct[cons_ct[i]] = 0;
cons_ct[i] = 0;
}
if(bp[i][j] >= bp[i2][j2]) {
/*printf("rm %4d:%4d\n", i2, j2);*/
cons_ct[cons_ct[i2]] = 0;
cons_ct[i2] = 0;
}
}
}
}
}
}
}
/***************************************/
/*PARANOID, second check for knots
for(i = 1; i <= msa->alen; i++) {
j = cons_ct[i];
if(j != 0 && i < j) {
printf("BP: %4d:%4d\n", i, j);
for(i2 = 1; i2 <= msa->alen; i2++) {
j2 = cons_ct[i2];
if(j2 != 0 && i2 < j2) {
if((i2 < j) && (j < j2)) {
if((i < i2)) {
printf("KNOT %4d:%4d (%4d) %4d:%4d (%4d)\n", i, j, bp[i][j], i2, j2, bp[i2][j2]);
}
}
}
}
}
}
******************************************/
/***************************************/
/*PARANOID, check cons_ct for consistency
for(i = 1; i <= msa->alen; i++) {
if(cons_ct[i] != 0) {
if(cons_ct[cons_ct[i]] != i) { printf("ERROR: i: %4d cons_ct[i]: %4d cons_ct[cons_ct[i]]: %4d\n", i, cons_ct[i], cons_ct[cons_ct[i]]); }
}
}
*/
/*PARANOID, write out SS_cons
for(i = 1; i <= msa->alen; i++) {
if(i < cons_ct[i]) printf("<");
else if(cons_ct[i] != 0) { printf(">"); }
else printf(".");
}
printf("\n");
*/
/***************************************/
/* textize alignment */
if((status = esl_msa_Textize(msa)) != eslOK) esl_fatal("ERROR textizing alignment %d\n", nali);
/* --fmin */
if(do_fmin) {
/* define ss_cons */
prev_nbps = -1;
fthresh = 0.99;
inconsistent_flag = pknot_flag = FALSE;
printf("# Defining consensus structure:\n");
printf("# indi SS basepair aln columns i:j (from at least 1 indi SS) will become consensus basepair\n");
printf("# if > <x> individual SS contain i:j as a pair\n");
printf("# We'll search for minimal <x> that gives a consistent consensus structure.\n");
printf("# A consistent structure has each position involved in 0 or 1 basepairs.\n");
printf("#\n");
printf("# Alignment file: %s\n", alifile);
printf("# Alignment idx: %d\n", nali);
printf("# Number of seqs: %d\n", msa->nseq);
printf("#\n");
printf("# %5s %23s %6s\n", "<x>", "nseq-required-with-bp", "numbps");
printf("# %5s %23s %6s\n", "-----", "-----------------------", "------");
while(fthresh >= 0.00 && (inconsistent_flag == FALSE) && (pknot_flag == FALSE)) {
nbps = 0;
seqthresh = (int) (fthresh * msa->nseq);
/*printf("fthresh: %f seqthresh: %d nseq: %d\n", fthresh, seqthresh, msa->nseq);*/
esl_vec_ISet(cons_ct, msa->alen+1, 0);
for(i = 1; i <= msa->alen; i++) {
for(j = i+1; j <= msa->alen; j++) {
if(bp[i][j] > seqthresh) {
if(cons_ct[i] != 0 || cons_ct[j] != 0) {
inconsistent_flag = TRUE;
}
/* check for pseudoknots */
for(k = i+1; k < j; k++) {
l = cons_ct[k];
if((k < l) && (l > j)) {
pknot_flag = TRUE;
}
if((k > l) && (l != 0) && (l < i)) {
pknot_flag = TRUE;
}
}
cons_ct[i] = j;
cons_ct[j] = i;
nbps++;
}
}
}
if(inconsistent_flag)
printf(" %.3f %23d %s\n", fthresh, seqthresh+1, "inconsistent");
else if(pknot_flag)
printf(" %.3f %23d %s\n", fthresh, seqthresh+1, "pseudoknotted");
else {
if(nbps != prev_nbps) {
printf(" %.3f %23d %6d\n", fthresh, seqthresh+1, nbps);
}
fmin = fthresh;
}
fthresh -= 0.01;
prev_nbps = nbps;
}
fthresh = fmin;
esl_vec_ISet(cons_ct, msa->alen+1, 0);
}
/* --ffreq: determine structure by defining consensus bps that occur in <x> fraction of indi structures */
if(do_ffreq || do_fmin) {
if(do_fmin) { printf("#\n# <x> determined to be %.3f\n", fthresh); }
if(do_ffreq) {
printf("# Defining consensus structure:\n");
printf("# indi SS basepair aln columns i:j (from at least 1 indi SS) will become consensus basepair\n");
printf("# if > %f individual SS contain i:j as a pair\n", fthresh);
}
esl_vec_ISet(cons_ct, msa->alen+1, 0);
/* define ss_cons */
seqthresh = (int) (fthresh * msa->nseq);
/*printf("fthresh: %f seqthresh: %d nseq: %d\n", fthresh, seqthresh, msa->nseq);*/
for(i = 1; i <= msa->alen; i++) {
for(j = i+1; j <= msa->alen; j++) {
if(bp[i][j] > seqthresh) {
if(cons_ct[i] != 0) {
esl_fatal("ERROR, two base pairs including position %d satisfy threshold (%d:%d and %d:%d)!\n", i, i, cons_ct[i], i, j);
}
if(cons_ct[j] != 0) {
esl_fatal("ERROR, two base pairs including position %d satisfy threshold (%d:%d and %d:%d)!\n", j, j, cons_ct[j], i, j);
}
cons_ct[i] = j;
cons_ct[j] = i;
}
}
}
}
/* -r: redefine consensus struct by removing any bps that conflict with individual structures */
if(do_remove_bps) {
for(i = 1; i <= msa->alen; i++) {
if(!(removebp[i])) {
cons_ct[i] = xcons_ct[i];
cons_ct[cons_ct[i]] = i;
}
else {
printf("# Removing consensus bp: %d:%d\n", i, xcons_ct[i]);
cons_ct[xcons_ct[i]] = 0;
cons_ct[i] = 0;
}
}
}
/* -c: define consensus structure as indi sequence with highest number of consistent bps with structure OR */
/* --indi: define consensus structure as indi sequence <x> from --indi <x> */
if(do_consistent || do_indi2cons) {
if(do_indi2cons) {
indi = esl_opt_GetString(go, "--indi");
for(a = 0; a < msa->nseq; a++) {
if(strcmp(indi, msa->sqname[a]) == 0) break;
}
if(a == msa->nseq) esl_fatal("ERROR, could not find a sequence named %s in the alignment.\n", indi);
}
else { /* do_consistent */
a = max_nconsistent_aidx;
}
if(msa->ss == NULL || msa->ss[a] == NULL) esl_fatal("ERROR, no individual SS annotation for %s in the alignment.\n", msa->sqname[a]);
if((status = esl_wuss2ct(msa->ss[a], msa->alen, cons_ct)) != eslOK) {
esl_fatal("Second pass... SS annotation for sequence %d, aln %d is invalid.\n", (a), nali);
}
printf("# Defined new SS_cons as SS annotation for %s (%d basepairs)\n", msa->sqname[a], nbpsA[a]);
if(esl_opt_GetBoolean(go, "--rfc") || esl_opt_GetBoolean(go, "--rfindi")) {
if(msa->rf != NULL) { free(msa->rf); msa->rf = NULL; }
if((status = esl_strcat(&(msa->rf), -1, msa->aseq[a], msa->alen)) != eslOK) goto ERROR;
printf("# Defined new RF as %s sequence\n", msa->sqname[a]);
}
}
/* write out alignment with new SS_cons */
if(do_newcons) {
if((status = esl_ct2wuss(cons_ct, msa->alen, sscons)) != eslOK) goto ERROR;
if(msa->ss_cons != NULL) { free(msa->ss_cons); msa->ss_cons = NULL; }
if((status = esl_strcat(&(msa->ss_cons), -1, sscons, msa->alen)) != eslOK) goto ERROR;
status = esl_msafile_Write(ofp, msa, (esl_opt_GetBoolean(go, "--pfam") ? eslMSAFILE_PFAM : eslMSAFILE_STOCKHOLM));
if (status == eslEMEM) esl_fatal("Memory error when outputting alignment\n");
else if (status != eslOK) esl_fatal("Writing alignment file failed with error %d\n", status);
}
free(sscons);
free(cur_ct);
free(cons_ct);
free(xcons_ct);
for(i = 1; i <= msa->alen; i++) free(bp[i]);
free(bp);
esl_msa_Destroy(msa);
}
if (nali == 0) esl_fatal("No alignments found in file %s\n", alifile);
/* Cleanup, normal return
*/
if(lfp != NULL) fclose(lfp);
if(ofp != NULL) {
printf("# Alignment(s) saved to file %s\n", esl_opt_GetString(go, "-o"));
fclose(ofp);
}
esl_msafile_Close(afp);
esl_getopts_Destroy(go);
return 0;
ERROR:
if(afp) esl_msafile_Close(afp);
if(go) esl_getopts_Destroy(go);
if(msa) esl_msa_Destroy(msa);
if(lfp) fclose(lfp);
if(ofp) fclose(ofp);
esl_fatal("ERROR\n");
return 1;
}
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 = esl_opt_GetInteger(go, "-L");
int N = esl_opt_GetInteger(go, "-N") / SSE16_NVALS;
int MaxPart = esl_opt_GetInteger(go, "-M");
int NROUNDS = esl_opt_GetInteger(go, "-R");
int check = esl_opt_GetBoolean(go, "-c");
__m128 resdata[10];
int i, j;
float *sc1 = (float*) resdata;
ESL_SQFILE *sqfp = NULL;
DATA_COPS16 *dcops;
struct timeb tbstart, tbend;
int sumlengths = 0;
float* results = NULL;
srand(time(NULL));
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);
dcops = p7_ViterbiCOPSw_Create(gm1);
p7_ViterbiCOPSW_Setup(dcops, L+100, MaxPart); // use max L
dcops->L = L;
int dbsize = SSE16_NVALS*N;
SEQ **seqsdb= calloc(dbsize+1, sizeof(SEQ*));
int equallength = 1;
if (esl_sqfile_OpenDigital(abc, seqfile, eslSQFILE_FASTA, NULL, &sqfp) == eslOK)
{ // Use Sequence file
ESL_SQ* sq = esl_sq_CreateDigital(abc);
int maxseqs, len=0;
if (esl_opt_IsDefault(go, "-N")) // N not specified in cmdline
maxseqs = INT_MAX; // no limit
else
maxseqs = SSE16_NVALS*N; // use cmdline limit
for (j = 0; j < maxseqs && esl_sqio_Read(sqfp, sq) == eslOK; j++)
{
if (equallength && sq->n != len && j > 0)
equallength = 0;
len = sq->n;
if (j > dbsize)
{ seqsdb = realloc(seqsdb, 2*(dbsize+1)*sizeof(SEQ*));
dbsize *= 2;
}
ESL_DSQ* dsq = sq->dsq;
seqsdb[j] = malloc(sizeof(SEQ));
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);
}
N = j/SSE16_NVALS;
}
else // Not found database. Generate random sequences
for (i = 0; i < N; i++)
for (j = 0; j < SSE16_NVALS; j++)
{
int len = L; // - rand()%1000;
seqsdb[i*SSE16_NVALS+j] = malloc(sizeof(SEQ));
seqsdb[i*SSE16_NVALS+j]->seq = malloc(len+4);
seqsdb[i*SSE16_NVALS+j]->length = len;
esl_rsq_xfIID(r, bg->f, abc->K, len, seqsdb[i*SSE16_NVALS+j]->seq);
sumlengths += len;
}
printf("Viterbi COPS Word with %d threads, model %s. ModelLen: %d, #Segms: %d, SeqL.: %d, #seqs: %d, Partition: %d, #parts: %d\n",
NTHREADS, hmmfile, gm1->M, (int) ceil(gm1->M/SSE16_NVALS), L, SSE16_NVALS*N*NROUNDS, dcops->partition, dcops->Npartitions);
/* // No. of partitions computed without full parallelism ( == no. of threads active while some are idle)
int Niters_part = dcops->Npartitions % NTHREADS;
// No. of Model lines that could be computed but are wasted by idle threads waiting on the end
int Nwasted_threads = dcops->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= (dcops->partition - gm1->M % dcops->partition) % dcops->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, dcops->partition * NTHREADS);
printf("Total Comp Lines: %d | Wasted Comp Lines: %d\n", totalcomp, wastedcomp);
*/
if (check) results = (float*) alloc_m128_aligned64((N+1)*2);
ftime(&tbstart);
if (!equallength)
{ // Sort sequences by length
qsort(seqsdb, N*SSE16_NVALS, sizeof(SEQ*), compare_seqs);
}
for (j = 0; j < NROUNDS; j++)
for (i = 0; i < N; i++)
{
// if (i % 1000 == 0) printf("Seq %d\n", i);
p7_ViterbiCOPSw_run(dcops, seqsdb+i*SSE16_NVALS, sc1);
if (check) memcpy(results+i*SSE16_NVALS, sc1, 32); // 32 bytes indeed! SSE16_NVALS floats
}
ftime(&tbend);
double secs = TIMEDIFF(tbstart,tbend);
w->elapsed = w->user = secs;
esl_stopwatch_Display(stdout, w, "# Opt CPU time: ");
double compmillioncells = NROUNDS * (double) sumlengths * (double) hmm->M * 1e-6;
printf("# %.0fM cells in %.1f Mc/s\n", compmillioncells, compmillioncells / secs);
if (check)
{ 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);
printf("Compare results against base version\n");
for (i = 0; i < N; i++)
{
int maxll = 0; float sc2;
for (j = 0; j < SSE16_NVALS; j++)
if (maxll < seqsdb[i*SSE16_NVALS+j]->length)
maxll = seqsdb[i*SSE16_NVALS+j]->length;
p7_oprofile_ReconfigRestLength(om, maxll);
// p7_ReconfigLength(gm2, maxll); // emulate the lock-step inter-sequence reconfigs
for (j = 0; j < SSE16_NVALS; j++)
{
// p7_ReconfigLength(gm2, seqsdb[i*SSE16_NVALS+j]->length);
// p7_Viterbi_unilocal(seqsdb[i*SSE16_NVALS+j]->seq, seqsdb[i*SSE16_NVALS+j]->length, gm2, &sc3);
// p7_Viterbi_unilocal_word(seqsdb[i*SSE16_NVALS+j]->seq, seqsdb[i*SSE16_NVALS+j]->length, gm2, &sc2);
// p7_oprofile_ReconfigLength(om, seqsdb[i*SSE16_NVALS+j]->length);
p7_ViterbiFilter(seqsdb[i*SSE16_NVALS+j]->seq, seqsdb[i*SSE16_NVALS+j]->length, om, ox, &sc2);
sc2 += 1.0; // -2.0nat optimization, Local to Unilocal mode
if (fabs(results[i*SSE16_NVALS+j] - sc2) > 0.0001)
{ printf("Seq %d Len %4d: %f - %f\tdiff: %f\n", i*SSE16_NVALS+j, seqsdb[i*SSE16_NVALS+j]->length,
results[i*SSE16_NVALS+j], sc2, fabs(results[i*SSE16_NVALS+j] - sc2));
}
}
}
}
return 0;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = NULL; /* command line configuration */
struct cfg_s cfg; /* application configuration */
char *basename= NULL; /* base of the output file names */
char *alifile = NULL; /* alignment file name */
char *dbfile = NULL; /* name of seq db file */
char outfile[256]; /* name of an output file */
int alifmt; /* format code for alifile */
int dbfmt; /* format code for dbfile */
ESL_MSAFILE *afp = NULL; /* open alignment file */
ESL_MSA *origmsa = NULL; /* one multiple sequence alignment */
ESL_MSA *msa = NULL; /* MSA after frags are removed */
ESL_MSA *trainmsa= NULL; /* training set, aligned */
ESL_STACK *teststack=NULL; /* test set: stack of ESL_SQ ptrs */
int status; /* easel return code */
int nfrags; /* # of fragments removed */
int ntestdom; /* # of test domains */
int ntest; /* # of test sequences created */
int nali; /* number of alignments read */
double avgid;
/* Parse command line */
go = esl_getopts_Create(options);
if (esl_opt_ProcessCmdline(go, argc, argv) != eslOK) cmdline_failure(argv[0], "Failed to parse command line: %s\n", go->errbuf);
if (esl_opt_VerifyConfig(go) != eslOK) cmdline_failure(argv[0], "Error in app configuration: %s\n", go->errbuf);
if (esl_opt_GetBoolean(go, "-h")) cmdline_help(argv[0], go);
if (esl_opt_ArgNumber(go) != 3) cmdline_failure(argv[0], "Incorrect number of command line arguments\n");
basename = esl_opt_GetArg(go, 1);
alifile = esl_opt_GetArg(go, 2);
dbfile = esl_opt_GetArg(go, 3);
alifmt = eslMSAFILE_STOCKHOLM;
dbfmt = eslSQFILE_FASTA;
/* Set up the configuration structure shared amongst functions here */
if (esl_opt_IsDefault(go, "--seed")) cfg.r = esl_randomness_CreateTimeseeded();
else cfg.r = esl_randomness_Create(esl_opt_GetInteger(go, "--seed"));
cfg.abc = NULL; /* until we open the MSA file, below */
cfg.fragfrac = esl_opt_GetReal(go, "-F");
cfg.idthresh1 = esl_opt_GetReal(go, "-1");
cfg.idthresh2 = esl_opt_GetReal(go, "-2");
cfg.test_lens = NULL;
cfg.ntest = 0;
/* Open the output files */
if (snprintf(outfile, 256, "%s.msa", basename) >= 256) esl_fatal("Failed to construct output MSA file name");
if ((cfg.out_msafp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open MSA output file %s\n", outfile);
if (snprintf(outfile, 256, "%s.fa", basename) >= 256) esl_fatal("Failed to construct output FASTA file name");
if ((cfg.out_seqfp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open FASTA output file %s\n", outfile);
if (snprintf(outfile, 256, "%s.pos", basename) >= 256) esl_fatal("Failed to construct pos test set summary file name");
if ((cfg.possummfp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open pos test set summary file %s\n", outfile);
if (snprintf(outfile, 256, "%s.neg", basename) >= 256) esl_fatal("Failed to construct neg test set summary file name");
if ((cfg.negsummfp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open neg test set summary file %s\n", outfile);
if (snprintf(outfile, 256, "%s.tbl", basename) >= 256) esl_fatal("Failed to construct benchmark table file name");
if ((cfg.tblfp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open benchmark table file %s\n", outfile);
/* Open the MSA file; determine alphabet */
status = esl_msafile_Open(alifile, alifmt, NULL, &afp);
if (status == eslENOTFOUND) esl_fatal("Alignment file %s doesn't exist or is not readable\n", alifile);
else if (status == eslEFORMAT) esl_fatal("Couldn't determine format of alignment %s\n", alifile);
else if (status != eslOK) esl_fatal("Alignment file open failed with error %d\n", status);
if (esl_opt_GetBoolean(go, "--amino")) cfg.abc = esl_alphabet_Create(eslAMINO);
else if (esl_opt_GetBoolean(go, "--dna")) cfg.abc = esl_alphabet_Create(eslDNA);
else if (esl_opt_GetBoolean(go, "--rna")) cfg.abc = esl_alphabet_Create(eslRNA);
else {
int type;
status = esl_msafile_GuessAlphabet(afp, &type);
if (status == eslEAMBIGUOUS) esl_fatal("Failed to guess the bio alphabet used in %s.\nUse --dna, --rna, or --amino option to specify it.", alifile);
else if (status == eslEFORMAT) esl_fatal("Alignment file parse failed: %s\n", afp->errbuf);
else if (status == eslENODATA) esl_fatal("Alignment file %s is empty\n", alifile);
else if (status != eslOK) esl_fatal("Failed to read alignment file %s\n", alifile);
cfg.abc = esl_alphabet_Create(type);
}
esl_msafile_SetDigital(afp, cfg.abc);
if (cfg.abc->type == eslAMINO) esl_composition_SW34(cfg.fq);
else esl_vec_DSet(cfg.fq, cfg.abc->K, 1.0 / (double) cfg.abc->K);
/* Open and process the dbfile; make sure it's in the same alphabet */
process_dbfile(&cfg, dbfile, dbfmt);
/* Read and process MSAs one at a time */
nali = 0;
while ((status = esl_msa_Read(afp, &origmsa)) == eslOK)
{
remove_fragments(&cfg, origmsa, &msa, &nfrags);
separate_sets (&cfg, msa, &trainmsa, &teststack);
ntestdom = esl_stack_ObjectCount(teststack);
if (ntestdom >= 2)
{
esl_stack_Shuffle(cfg.r, teststack);
synthesize_positives(go, &cfg, msa->name, teststack, &ntest);
esl_msa_MinimGaps(trainmsa, NULL, NULL);
esl_msa_Write(cfg.out_msafp, trainmsa, eslMSAFILE_STOCKHOLM);
esl_dst_XAverageId(cfg.abc, trainmsa->ax, trainmsa->nseq, 10000, &avgid); /* 10000 is max_comparisons, before sampling kicks in */
fprintf(cfg.tblfp, "%-20s %3.0f%% %6d %6d %6d %6d %6d %6d\n", msa->name, 100.*avgid, (int) trainmsa->alen, msa->nseq, nfrags, trainmsa->nseq, ntestdom, ntest);
nali++;
}
esl_msa_Destroy(trainmsa);
esl_msa_Destroy(origmsa);
esl_msa_Destroy(msa);
}
if (status == eslEFORMAT) esl_fatal("Alignment file parse error, line %d of file %s:\n%s\nOffending line is:\n%s\n",
afp->linenumber, afp->fname, afp->errbuf, afp->buf);
else if (status != eslEOF) esl_fatal("Alignment file read failed with error code %d\n", status);
else if (nali == 0) esl_fatal("No alignments found in file %s\n", alifile);
if (nali > 0)
synthesize_negatives(go, &cfg, esl_opt_GetInteger(go, "-N"));
fclose(cfg.out_msafp);
fclose(cfg.out_seqfp);
fclose(cfg.possummfp);
fclose(cfg.negsummfp);
fclose(cfg.tblfp);
esl_randomness_Destroy(cfg.r);
esl_alphabet_Destroy(cfg.abc);
esl_msafile_Close(afp);
esl_getopts_Destroy(go);
return 0;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go; /* application configuration */
int kstatus, tstatus;/* return code from Easel routine */
int fmt; /* expected format of kfile, tfile */
char *kfile, *tfile; /* known, test structure file */
ESL_MSAFILE *kfp, *tfp; /* open kfile, tfile */
ESL_MSA *ka, *ta; /* known, trusted alignment */
int64_t klen, tlen; /* lengths of dealigned seqs */
int i; /* counter over sequences */
int apos; /* counter over alignment columns */
int rfpos; /* counter over consensus (non-gap RF) columns */
int is_rfpos; /* TRUE if current apos is a consensus pos, FALSE if not */
int uapos; /* counter over unaligned residue positions */
int nali; /* number of alignment we're on in each file */
int **kp; /* [0..i..nseq-1][1..r..sq->n] = x known non-gap RF position of residue r in sequence i */
int **tp; /* [0..i..nseq-1][1..r..sq->n] = x predicted non-gap RF position of residue r in sequence i */
/* for both kp and pp, if x <= 0, residue r for seq i is not aligned to a non-gap RF position, but rather as an 'insert'
* after non-gap RF position (x * -1)
*/
int *km_pos; /* [0..rflen] = x, in known aln, number of residues aligned to non-gap RF column x; special case: mct[0] = 0 */
int *ki_pos; /* [0..rflen] = x, in known aln, number of residues inserted after non-gap RF column x */
int *tm_pos; /* [0..rflen] = x, in predicted aln, number of residues aligned to non-gap RF column x; special case: mct[0] = 0 */
int *ti_pos; /* [0..rflen] = x, in predicted aln, number of residues inserted after non-gap RF column x */
int *cor_tm_pos; /* [0..rflen] = x, in predicted aln, number of correctly predicted residues aligned to non-gap RF column x; special case: mct[0] = 0 */
int *cor_ti_pos; /* [0..rflen] = x, in predicted aln, number of correctly predicted residues inserted after non-gap RF column x */
int *km_seq; /* [0..i..nseq-1] = x, in known aln, number of residues aligned to non-gap RF columns in seq i; */
int *ki_seq; /* [0..i..nseq-1] = x, in known aln, number of residues inserted in seq i */
int *tm_seq; /* [0..i..nseq-1] = x, in predicted aln, number of residues aligned to non-gap RF columns in seq i; */
int *ti_seq; /* [0..i..nseq-1] = x, in predicted aln, number of residues inserted in seq i */
int *cor_tm_seq; /* [0..i..nseq-1] = x, in predicted aln, number of correctly predicted residues aligned to non-gap RF columns in seq i */
int *cor_ti_seq; /* [0..i..nseq-1] = x, in predicted aln, number of correctly predicted residues inserted in seq i */
int *seqlen; /* [0..i..nseq-1] = x, unaligned seq i has length x */
ESL_ALPHABET *abc = NULL; /* alphabet for all alignments */
int rflen, t_rflen; /* non-gap RF length (consensus lengths) */
int status;
char *namedashes;
int ni;
int namewidth = 8; /* length of 'seq name' */
int cor_tm, cor_ti, km, ki; /* correct predicted match, correct predicted insert, total match, total insert */
char *mask = NULL;
int masklen;
ESL_DSQ *ks;
ESL_DSQ *ts;
FILE *dfp = NULL; /* for --c2dfile */
/* variables needed for -p and related options */
int do_post = FALSE; /* TRUE if -p enabled */
int do_post_for_this_rfpos = FALSE; /* set for each consensus position, always TRUE unless --mask-p2xm */
int p; /* counter over integerized posteriors */
int *ptm = NULL; /* [0..p..10] number of total matches with posterior value p (10="*")*/
int *pti = NULL; /* [0..p..10] number of total inserts with posterior value p */
int *cor_ptm = NULL; /* [0..p..10] number of correct matches with posterior value p */
int *cor_pti = NULL; /* [0..p..10] number of correct inserts with posterior value p */
int npostvals = 11; /* number of posterior values 0-9, * */
int ppidx; /* index of PP */
char ppchars[11] = "0123456789*";
int cm_cor_ptm, cm_cor_pti, cm_ptm, cm_pti, cm_incor_ptm, cm_incor_pti; /* cumulative counts of posteriors */
// int tot_cor_ptm, tot_cor_pti, tot_ptm, tot_pti; /* total counts of posteriors */
// int tot_incor_ptm,tot_incor_pti; // SRE: commented out; don't seem to be used; need to silence compiler warning
char errbuf[eslERRBUFSIZE];
/***********************************************
* Parse command line
***********************************************/
go = esl_getopts_Create(options);
if (esl_opt_ProcessCmdline(go, argc, argv) != eslOK ||
esl_opt_VerifyConfig(go) != eslOK)
{
printf("Failed to parse command line: %s\n", go->errbuf);
esl_usage(stdout, argv[0], usage);
printf("\nTo see more help on available options, do %s -h\n\n", argv[0]);
exit(1);
}
if (esl_opt_GetBoolean(go, "-h") )
{
esl_banner(stdout, argv[0], banner);
esl_usage (stdout, argv[0], usage);
puts("\n where options are:");
esl_opt_DisplayHelp(stdout, go, 1, 2, 80);
esl_opt_DisplayHelp(stdout, go, 2, 2, 80);
exit(EXIT_SUCCESS);
}
if (esl_opt_ArgNumber(go) != 2)
{
printf("Incorrect number of command line arguments.\n");
esl_usage(stdout, argv[0], usage);
printf("\nTo see more help on available options, do %s -h\n\n", argv[0]);
exit(1);
}
kfile = esl_opt_GetArg(go, 1);
tfile = esl_opt_GetArg(go, 2);
fmt = eslMSAFILE_STOCKHOLM;
/***********************************************
* Open the two Stockholm files.
***********************************************/
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);
if ( (kstatus = esl_msafile_Open(&abc, kfile, NULL, fmt, NULL, &kfp)) != eslOK) esl_msafile_OpenFailure(kfp, kstatus);
if ( (tstatus = esl_msafile_Open(&abc, tfile, NULL, fmt, NULL, &tfp)) != eslOK) esl_msafile_OpenFailure(tfp, tstatus);
do_post = esl_opt_GetBoolean(go, "-p");
/* read the mask file if --p-mask is enabled */
if(! esl_opt_IsDefault(go, "--p-mask")) {
if((status = read_mask_file(esl_opt_GetString(go, "--p-mask"), errbuf, &mask, &masklen)) != eslOK) esl_fatal(errbuf);
}
/* open the c2dfile for output, if nec */
if (esl_opt_IsOn(go, "--c2dfile")) {
if ((dfp = fopen(esl_opt_GetString(go, "--c2dfile"), "w")) == NULL) esl_fatal("Failed to open --c2dfile output file %s\n", esl_opt_GetString(go, "--c2dfile"));
}
/***********************************************
* Do alignment comparisons, one seq at a time;
* this means looping over all seqs in all alignments.
***********************************************/
nali = 0;
while ( (kstatus = esl_msafile_Read(kfp, &ka)) != eslEOF)
{
if ( kstatus != eslOK) esl_msafile_ReadFailure(kfp, kstatus);
if ( (tstatus = esl_msafile_Read(tfp, &ta)) != eslOK) esl_msafile_ReadFailure(tfp, tstatus);
nali++;
if((nali > 1) && (esl_opt_IsOn(go, "--c2dfile"))) esl_fatal("--c2dfile is only meant for msafiles with single alignments");
/* Sanity check on alignment
*/
if (ka->nseq != ta->nseq)
esl_fatal("trusted, test alignments don't have same seq #\n");
if (ka->rf == NULL)
esl_fatal("trusted alignment has no reference annotation\n");
if (ta->rf == NULL)
esl_fatal("test alignment has no reference annotation\n");
/* make sure the sequences are all identical */
ESL_ALLOC(seqlen, sizeof(int) * ka->nseq);
for(i = 0; i < ka->nseq; i++) {
if(strcmp(ka->sqname[i], ta->sqname[i]) != 0) esl_fatal("sequence %d of trusted alignment %s has different name than seq %d of predicted alignment %s\n", (i+1), ka->sqname[i], (i+1), ta->sqname[i]);
ESL_ALLOC(ks, sizeof(ESL_DSQ) * (ka->alen+2));
memcpy(ks, ka->ax[i], (ka->alen+2) * sizeof(ESL_DSQ));
esl_abc_XDealign(ka->abc, ks, ka->ax[i], &klen);
ESL_ALLOC(ts, sizeof(ESL_DSQ) * (ta->alen+2));
memcpy(ts, ta->ax[i], (ta->alen+2) * sizeof(ESL_DSQ));
esl_abc_XDealign(ta->abc, ts, ta->ax[i], &tlen);
if (tlen != klen)
esl_fatal("dealigned sequence mismatch, seq %d, when dealigned, is %d residues in the known alignment, but %d residues in the trusted alignment.", (i+1), klen, tlen);
if (memcmp(ks, ts, sizeof(ESL_DSQ) * klen) != 0)
esl_fatal("dealigned sequence mismatch, seq %d %s, when dealigned, are not identical.", (i+1), ka->sqname[i]);
seqlen[i] = tlen;
free(ks);
free(ts);
}
/* determine non-gap RF length */
rflen = 0;
for(apos = 1; apos <= ka->alen; apos++) {
if((! esl_abc_CIsGap (ka->abc, ka->rf[apos-1])) &&
(! esl_abc_CIsMissing(ka->abc, ka->rf[apos-1]))) rflen++;
}
t_rflen = 0;
for(apos = 1; apos <= ta->alen; apos++) {
if((! esl_abc_CIsGap (ta->abc, ta->rf[apos-1])) &&
(! esl_abc_CIsMissing (ta->abc, ta->rf[apos-1]))) t_rflen++;
}
if(t_rflen != rflen) esl_fatal("Trusted alignment non-gap RF length (%d) != predicted alignment non-gap RF length (%d).\n", rflen, t_rflen);
/* if -p, make sure the test alignment has posterior probabilities, and allocate our counters for correct/incorrect per post value */
if(do_post) {
if(! esl_opt_IsDefault(go, "--p-mask")) {
if(masklen != rflen) {
esl_fatal("Length of mask in %s (%d) not equal to non-gap RF len of alignments (%d)\n", esl_opt_GetString(go, "--p-mask"), masklen, rflen);
}
}
if(ta->pp == NULL) esl_fatal("-p requires \"#=GR PP\" annotation in the test alignment, but none exists");
ESL_ALLOC(ptm, sizeof(int) * npostvals);
ESL_ALLOC(pti, sizeof(int) * npostvals);
ESL_ALLOC(cor_ptm, sizeof(int) * npostvals);
ESL_ALLOC(cor_pti, sizeof(int) * npostvals);
esl_vec_ISet(ptm, npostvals, 0);
esl_vec_ISet(pti, npostvals, 0);
esl_vec_ISet(cor_ptm, npostvals, 0);
esl_vec_ISet(cor_pti, npostvals, 0);
}
/* allocate and initialize our counters */
ESL_ALLOC(kp, sizeof(int *) * ka->nseq);
ESL_ALLOC(tp, sizeof(int *) * ta->nseq);
for(i = 0; i < ka->nseq; i++) {
ESL_ALLOC(kp[i], sizeof(int) * (seqlen[i]+1));
ESL_ALLOC(tp[i], sizeof(int) * (seqlen[i]+1));
esl_vec_ISet(kp[i], seqlen[i]+1, -987654321);
esl_vec_ISet(tp[i], seqlen[i]+1, -987654321);
}
ESL_ALLOC(km_pos, sizeof(int) * (rflen+1));
ESL_ALLOC(ki_pos, sizeof(int) * (rflen+1));
ESL_ALLOC(tm_pos, sizeof(int) * (rflen+1));
ESL_ALLOC(ti_pos, sizeof(int) * (rflen+1));
ESL_ALLOC(cor_tm_pos, sizeof(int) * (rflen+1));
ESL_ALLOC(cor_ti_pos, sizeof(int) * (rflen+1));
esl_vec_ISet(km_pos, rflen+1, 0);
esl_vec_ISet(ki_pos, rflen+1, 0);
esl_vec_ISet(tm_pos, rflen+1, 0);
esl_vec_ISet(ti_pos, rflen+1, 0);
esl_vec_ISet(cor_tm_pos, rflen+1, 0);
esl_vec_ISet(cor_ti_pos, rflen+1, 0);
ESL_ALLOC(km_seq, sizeof(int) * ka->nseq);
ESL_ALLOC(ki_seq, sizeof(int) * ka->nseq);
ESL_ALLOC(tm_seq, sizeof(int) * ka->nseq);
ESL_ALLOC(ti_seq, sizeof(int) * ka->nseq);
ESL_ALLOC(cor_tm_seq, sizeof(int) * ka->nseq);
ESL_ALLOC(cor_ti_seq, sizeof(int) * ka->nseq);
esl_vec_ISet(km_seq, ka->nseq, 0);
esl_vec_ISet(ki_seq, ka->nseq, 0);
esl_vec_ISet(tm_seq, ka->nseq, 0);
esl_vec_ISet(ti_seq, ka->nseq, 0);
esl_vec_ISet(cor_tm_seq, ka->nseq, 0);
esl_vec_ISet(cor_ti_seq, ka->nseq, 0);
/* determine non-gap RF location of each residue in known alignment */
for(i = 0; i < ka->nseq; i++) {
uapos = rfpos = 0;
for(apos = 1; apos <= ka->alen; apos++) {
is_rfpos = FALSE;
if((! esl_abc_CIsGap (ka->abc, ka->rf[apos-1])) &&
(! esl_abc_CIsMissing (ka->abc, ka->rf[apos-1]))) {
rfpos++; is_rfpos = TRUE;
}
if(esl_abc_XIsResidue(ka->abc, ka->ax[i][apos])) {
uapos++;
kp[i][uapos] = (is_rfpos) ? rfpos : (-1 * rfpos);
if(is_rfpos) { km_pos[rfpos]++; km_seq[i]++; }
else { ki_pos[rfpos]++; ki_seq[i]++; }
}
}
}
/* determine non-gap RF location of each residue in predicted alignment */
for(i = 0; i < ta->nseq; i++) {
uapos = rfpos = 0;
for(apos = 1; apos <= ta->alen; apos++) {
is_rfpos = FALSE;
if((! esl_abc_CIsGap (abc, ta->rf[apos-1])) &&
(! esl_abc_CIsMissing (abc, ta->rf[apos-1]))) {
rfpos++; is_rfpos = TRUE;
if(do_post) {
do_post_for_this_rfpos = (mask != NULL && mask[rfpos-1] == '0') ? FALSE : TRUE;
}
}
if(esl_abc_XIsResidue(ta->abc, ta->ax[i][apos])) {
uapos++;
tp[i][uapos] = (is_rfpos) ? rfpos : (-1 * rfpos);
if(do_post) {
if(esl_abc_CIsGap(abc, ta->pp[i][(apos-1)])) esl_fatal("gap PP value for nongap residue: ali: %d seq: %d apos: %d\n", nali, i, apos);
ppidx = get_pp_idx(abc, ta->pp[i][(apos-1)]);
if(ppidx == -1) esl_fatal("unrecognized PP value (%c) for nongap residue: ali: %d seq: %d apos: %d\n", ta->pp[i][(apos-1)], nali, i, apos);
}
if(is_rfpos) {
tm_pos[rfpos]++; tm_seq[i]++;
if(do_post_for_this_rfpos) ptm[ppidx]++;
}
else {
ti_pos[rfpos]++; ti_seq[i]++;
if(do_post) pti[ppidx]++;
}
if(kp[i][uapos] == tp[i][uapos]) { /* correctly predicted this residue */
if(is_rfpos) {
cor_tm_seq[i]++; cor_tm_pos[rfpos]++;
if(do_post_for_this_rfpos) cor_ptm[ppidx]++;
}
else {
cor_ti_seq[i]++; cor_ti_pos[rfpos]++;
if(do_post) cor_pti[ppidx]++;
}
}
}
}
}
if((! (esl_opt_GetBoolean(go, "-c"))) && (! esl_opt_GetBoolean(go, "-p"))) {
/* print per sequence statistics */
/* determine the longest name in msa */
for(ni = 0; ni < ka->nseq; ni++) namewidth = ESL_MAX(namewidth, strlen(ka->sqname[ni]));
ESL_ALLOC(namedashes, sizeof(char) * namewidth+1);
namedashes[namewidth] = '\0';
for(ni = 0; ni < namewidth; ni++) namedashes[ni] = '-';
printf("# %-*s %6s %28s %28s %28s\n", namewidth, "seq name", "len", "match columns", "insert columns", "all columns");
printf("# %-*s %6s %28s %28s %28s\n", namewidth, namedashes, "------", "----------------------------", "----------------------------", "----------------------------");
for(i = 0; i < ta->nseq; i++) {
printf(" %-*s %6d %8d / %8d (%.3f) %8d / %8d (%.3f) %8d / %8d (%.3f)\n", namewidth, ka->sqname[i], seqlen[i],
cor_tm_seq[i], km_seq[i], (km_seq[i] == 0) ? 0. : ((float) cor_tm_seq[i] / (float) km_seq[i]),
cor_ti_seq[i], ki_seq[i], (ki_seq[i] == 0) ? 0. : ((float) cor_ti_seq[i] / (float) ki_seq[i]),
(cor_tm_seq[i] + cor_ti_seq[i]), (km_seq[i] + ki_seq[i]), ((float) (cor_tm_seq[i] + cor_ti_seq[i]) / ((float) km_seq[i] + ki_seq[i])));
}
cor_tm = esl_vec_ISum(cor_tm_seq, ka->nseq);
cor_ti = esl_vec_ISum(cor_ti_seq, ka->nseq);
km = esl_vec_ISum(km_seq, ka->nseq);
ki = esl_vec_ISum(ki_seq, ka->nseq);
printf("# %-*s %6s %28s %28s %28s\n", namewidth, namedashes, "-----", "----------------------------", "----------------------------", "----------------------------");
printf("# %-*s %6s %8d / %8d (%.3f) %8d / %8d (%.3f) %8d / %8d (%.3f)\n",
namewidth, "*all*", "-",
cor_tm, km, ((float) cor_tm / (float) km),
cor_ti, ki, ((float) cor_ti / (float) ki),
(cor_tm+cor_ti), (km+ki), (((float) (cor_tm + cor_ti))/ ((float) (km + ki))));
free(namedashes);
for(i = 0; i < ka->nseq; i++) {
free(kp[i]);
free(tp[i]);
}
}
else if(esl_opt_GetBoolean(go, "-c")) { /* print per column statistics */
printf("# %5s %20s %20s %20s\n", "rfpos", "match", "insert", "both");
printf("# %5s %20s %20s %20s\n", "-----", "--------------------", "--------------------", "--------------------");
for(rfpos = 0; rfpos <= rflen; rfpos++) {
printf(" %5d %4d / %4d (%.3f) %4d / %4d (%.3f) %4d / %4d (%.3f)\n", rfpos,
cor_tm_pos[rfpos], km_pos[rfpos], (km_pos[rfpos] == 0) ? 0. : ((float) cor_tm_pos[rfpos] / (float) km_pos[rfpos]),
cor_ti_pos[rfpos], ki_pos[rfpos], (ki_pos[rfpos] == 0) ? 0. : ((float) cor_ti_pos[rfpos] / (float) ki_pos[rfpos]),
(cor_tm_pos[rfpos] + cor_ti_pos[rfpos]), (km_pos[rfpos] + ki_pos[rfpos]), ((float) (cor_tm_pos[rfpos] + cor_ti_pos[rfpos]) / ((float) km_pos[rfpos] + ki_pos[rfpos])));
}
}
else if(do_post) { /* do posterior output */
if(mask == NULL) {
printf("# %2s %29s %29s\n", "", " match columns ", " insert columns ");
printf("# %2s %29s %29s\n", "", "-----------------------------", "-----------------------------") ;
printf("# %2s %8s %8s %9s %8s %8s %9s\n", "PP", "ncorrect", "ntotal", "fractcor", "ncorrect", "ntotal", "fractcor");
printf("# %2s %8s %8s %9s %8s %8s %9s\n", "--", "--------", "--------", "---------", "--------", "--------", "---------");
}
else {
printf("# %2s %29s %29s\n", "", " match columns within mask ", " insert columns ");
printf("# %2s %29s %29s\n", "", "-----------------------------", "-----------------------------") ;
printf("# %2s %8s %8s %9s %8s %8s %9s\n", "PP", "ncorrect", "ntotal", "fractcor", "ncorrect", "ntotal", "fractcor");
printf("# %2s %8s %8s %9s %8s %8s %9s\n", "--", "--------", "--------", "---------", "--------", "--------", "---------");
}
cm_ptm = cm_pti = cm_cor_ptm = cm_cor_pti = cm_incor_ptm = cm_incor_pti = 0;
//tot_ptm = esl_vec_ISum(ptm, npostvals);
//tot_pti = esl_vec_ISum(pti, npostvals);
//tot_cor_ptm = esl_vec_ISum(cor_ptm, npostvals);
//tot_cor_pti = esl_vec_ISum(cor_pti, npostvals);
//tot_incor_ptm = tot_ptm - tot_cor_ptm;
//tot_incor_pti = tot_pti - tot_cor_pti;
for(p = (npostvals-1); p >= 0; p--) {
cm_cor_ptm += cor_ptm[p];
cm_cor_pti += cor_pti[p];
cm_ptm += ptm[p];
cm_pti += pti[p];
cm_incor_ptm += ptm[p] - cor_ptm[p];
cm_incor_pti += pti[p] - cor_pti[p];
printf(" %2c %8d / %8d (%.5f) %8d / %8d (%.5f)\n",
ppchars[p], cor_ptm[p], ptm[p],
(ptm[p] == 0) ? 0. : (float) cor_ptm[p] / (float) ptm[p],
cor_pti[p], pti[p],
(pti[p] == 0) ? 0. : (float) cor_pti[p] / (float) pti[p]);
}
}
/* handle --c2dfile */
if (dfp != NULL) {
/* match stats, 4 fields, CMYK color values */
for(rfpos = 1; rfpos <= rflen; rfpos++) {
if(km_pos[rfpos] == 0) { /* special case, no known alignment residues, a blank position */
fprintf(dfp, "%.3f %.3f %.3f %.3f\n", 0., 0., 0., 0.);
}
else {
fprintf(dfp, "%.3f %.3f %.3f %.3f\n",
0., /* cyan */
1. - ((float) cor_tm_pos[rfpos] / (float) km_pos[rfpos]), /* magenta, fraction incorrect */
1. - ((float) km_pos[rfpos] / ta->nseq), /* yellow, 1 - fraction of seqs with residue in column */
0.);
}
}
fprintf(dfp, "//\n");
/* insert stats, 4 fields, CMYK color values */
rfpos = 0; /* special case, combine insert posn 0 and 1 together */
if(ki_pos[rfpos] == 0) { /* special case, no known alignment residues, a blank position */
fprintf(dfp, "%.3f %.3f %.3f %.3f\n", 0., 0., 0., 0.);
}
else {
fprintf(dfp, "%.3f %.3f %.3f %.3f\n",
0., /* cyan */
1. - ((float) (cor_ti_pos[0] + cor_ti_pos[1]) / ((float) (ki_pos[0] + ki_pos[1]))), /* magenta, fraction correct */
0.,
0.);
}
/* insert stats posn 2..rflen */
for(rfpos = 2; rfpos <= rflen; rfpos++) {
if(ki_pos[rfpos] == 0) { /* special case, no known alignment residues, a blank position */
fprintf(dfp, "%.3f %.3f %.3f %.3f\n", 0., 0., 0., 0.);
}
else {
fprintf(dfp, "%.3f %.3f %.3f %.3f\n",
0., /* cyan */
1. - ((float) cor_ti_pos[rfpos] / (float) ki_pos[rfpos]), /* magenta, fraction correct */
0.,
0.);
}
}
fprintf(dfp, "//\n");
}
if(ptm != NULL) free(ptm);
if(pti != NULL) free(pti);
if(cor_ptm != NULL) free(cor_ptm);
if(cor_ptm != NULL) free(cor_pti);
free(kp);
free(tp);
free(km_seq);
free(ki_seq);
free(tm_seq);
free(ti_seq);
free(cor_tm_seq);
free(cor_ti_seq);
free(km_pos);
free(ki_pos);
free(tm_pos);
free(ti_pos);
free(cor_tm_pos);
free(cor_ti_pos);
free(seqlen);
esl_msa_Destroy(ka);
esl_msa_Destroy(ta);
}
if(mask != NULL) free(mask);
if(dfp != NULL) {
fclose(dfp);
printf("# Draw file of per-column stats saved to file: %s\n", esl_opt_GetString(go, "--c2dfile"));
}
if(abc) esl_alphabet_Destroy(abc);
esl_getopts_Destroy(go);
esl_msafile_Close(tfp);
esl_msafile_Close(kfp);
return 0;
ERROR:
return status;
}
int
main(int argc, char **argv)
{
ESL_GETOPTS *go = NULL; /* command line configuration */
struct cfg_s cfg; /* application configuration */
char *basename= NULL; /* base of the output file names */
char *alifile = NULL; /* alignment file name */
char *dbfile = NULL; /* name of seq db file */
char outfile[256]; /* name of an output file */
int alifmt; /* format code for alifile */
int dbfmt; /* format code for dbfile */
ESLX_MSAFILE *afp = NULL; /* open alignment file */
ESL_MSA *origmsa = NULL; /* one multiple sequence alignment */
ESL_MSA *msa = NULL; /* MSA after frags are removed */
ESL_MSA *trainmsa= NULL; /* training set, aligned */
ESL_STACK *teststack=NULL; /* test set: stack of ESL_SQ ptrs */
int status; /* easel return code */
int nfrags; /* # of fragments removed */
int ntestdom; /* # of test domains */
int ntest; /* # of test sequences created */
int nali; /* number of alignments read */
double avgid;
/* Parse command line */
go = esl_getopts_Create(options);
if (esl_opt_ProcessCmdline(go, argc, argv) != eslOK) cmdline_failure(argv[0], "Failed to parse command line: %s\n", go->errbuf);
if (esl_opt_VerifyConfig(go) != eslOK) cmdline_failure(argv[0], "Error in app configuration: %s\n", go->errbuf);
if (esl_opt_GetBoolean(go, "-h")) cmdline_help(argv[0], go);
if (esl_opt_ArgNumber(go) != 3) cmdline_failure(argv[0], "Incorrect number of command line arguments\n");
basename = esl_opt_GetArg(go, 1);
alifile = esl_opt_GetArg(go, 2);
dbfile = esl_opt_GetArg(go, 3);
alifmt = eslMSAFILE_STOCKHOLM;
dbfmt = eslSQFILE_FASTA;
/* Set up the configuration structure shared amongst functions here */
if (esl_opt_IsDefault(go, "--seed")) cfg.r = esl_randomness_CreateTimeseeded();
else cfg.r = esl_randomness_Create(esl_opt_GetInteger(go, "--seed"));
cfg.abc = NULL; /* until we open the MSA file, below */
cfg.fragfrac = esl_opt_GetReal(go, "-F");
cfg.idthresh1 = esl_opt_GetReal(go, "-1");
cfg.idthresh2 = esl_opt_GetReal(go, "-2");
cfg.test_lens = NULL;
cfg.ntest = 0;
cfg.max_ntest = (esl_opt_IsOn(go, "--maxtest") ? esl_opt_GetInteger(go, "--maxtest") : 0);
cfg.max_ntrain = (esl_opt_IsOn(go, "--maxtrain") ? esl_opt_GetInteger(go, "--maxtrain") : 0);
/* Open the output files */
if (snprintf(outfile, 256, "%s.msa", basename) >= 256) esl_fatal("Failed to construct output MSA file name");
if ((cfg.out_msafp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open MSA output file %s\n", outfile);
if (snprintf(outfile, 256, "%s.fa", basename) >= 256) esl_fatal("Failed to construct output FASTA file name");
if ((cfg.out_seqfp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open FASTA output file %s\n", outfile);
if (snprintf(outfile, 256, "%s.pos", basename) >= 256) esl_fatal("Failed to construct pos test set summary file name");
if ((cfg.possummfp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open pos test set summary file %s\n", outfile);
if (snprintf(outfile, 256, "%s.neg", basename) >= 256) esl_fatal("Failed to construct neg test set summary file name");
if ((cfg.negsummfp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open neg test set summary file %s\n", outfile);
if (snprintf(outfile, 256, "%s.tbl", basename) >= 256) esl_fatal("Failed to construct benchmark table file name");
if ((cfg.tblfp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open benchmark table file %s\n", outfile);
if (esl_opt_GetBoolean(go, "--pid")) {
if (snprintf(outfile, 256, "%s.pid", basename) >= 256) esl_fatal("Failed to construct %%id table file name");
if ((cfg.pidfp = fopen(outfile, "w")) == NULL) esl_fatal("Failed to open %%id table file %s\n", outfile);
} else cfg.pidfp = NULL;
/* Open the MSA file, digital mode; determine alphabet */
if (esl_opt_GetBoolean(go, "--amino")) cfg.abc = esl_alphabet_Create(eslAMINO);
else if (esl_opt_GetBoolean(go, "--dna")) cfg.abc = esl_alphabet_Create(eslDNA);
else if (esl_opt_GetBoolean(go, "--rna")) cfg.abc = esl_alphabet_Create(eslRNA);
status = eslx_msafile_Open(&(cfg.abc), alifile, NULL, alifmt, NULL, &afp);
if (status != eslOK) eslx_msafile_OpenFailure(afp, status);
if (cfg.abc->type == eslAMINO) esl_composition_SW34(cfg.fq);
else esl_vec_DSet(cfg.fq, cfg.abc->K, 1.0 / (double) cfg.abc->K);
/* Open and process the dbfile; make sure it's in the same alphabet */
process_dbfile(&cfg, dbfile, dbfmt);
/* Read and process MSAs one at a time */
nali = 0;
while ((status = eslx_msafile_Read(afp, &origmsa)) != eslEOF)
{
if (status != eslOK) eslx_msafile_ReadFailure(afp, status);
esl_msa_ConvertDegen2X(origmsa);
esl_msa_Hash(origmsa);
remove_fragments(&cfg, origmsa, &msa, &nfrags);
separate_sets (&cfg, msa, &trainmsa, &teststack);
if ( esl_stack_ObjectCount(teststack) >= 2)
{
/* randomize test domain order, and apply size limit if any */
esl_stack_Shuffle(cfg.r, teststack);
if (cfg.max_ntest) pstack_select_topn(&teststack, cfg.max_ntest);
ntestdom = esl_stack_ObjectCount(teststack);
/* randomize training set alignment order, and apply size limit if any */
esl_msashuffle_PermuteSequenceOrder(cfg.r, trainmsa);
if (cfg.max_ntrain) msa_select_topn(&trainmsa, cfg.max_ntrain);
esl_msa_MinimGaps(trainmsa, NULL, NULL, FALSE);
if (esl_opt_GetBoolean(go, "--pid")) write_pids(cfg.pidfp, origmsa, trainmsa, teststack);
synthesize_positives(go, &cfg, msa->name, teststack, &ntest);
eslx_msafile_Write(cfg.out_msafp, trainmsa, eslMSAFILE_STOCKHOLM);
esl_dst_XAverageId(cfg.abc, trainmsa->ax, trainmsa->nseq, 10000, &avgid); /* 10000 is max_comparisons, before sampling kicks in */
fprintf(cfg.tblfp, "%-20s %3.0f%% %6d %6d %6d %6d %6d %6d\n", msa->name, 100.*avgid, (int) trainmsa->alen, msa->nseq, nfrags, trainmsa->nseq, ntestdom, ntest);
nali++;
}
esl_msa_Destroy(trainmsa);
esl_msa_Destroy(origmsa);
esl_msa_Destroy(msa);
}
if (nali == 0) esl_fatal("No alignments found in file %s\n", alifile);
synthesize_negatives(go, &cfg, esl_opt_GetInteger(go, "-N"));
fclose(cfg.out_msafp);
fclose(cfg.out_seqfp);
fclose(cfg.possummfp);
fclose(cfg.negsummfp);
fclose(cfg.tblfp);
if (cfg.pidfp) fclose(cfg.pidfp);
esl_randomness_Destroy(cfg.r);
esl_alphabet_Destroy(cfg.abc);
eslx_msafile_Close(afp);
esl_getopts_Destroy(go);
return 0;
}