/* Create a new category map from a string that can
either be a filename or a brief "inlined" category map, e.g.,
"NCATS = 3 ; CDS 1-3". Useful for command-line arguments. */
CategoryMap* cm_new_string_or_file(const char *optarg) {
int i;
char fname[STR_SHORT_LEN];
String *str = str_new_charstr(optarg);
FILE *F;
CategoryMap *retval = NULL;
str_double_trim(str);
if (str_starts_with_charstr(str, "NCATS")) {
/* replace semicolons with carriage returns */
for (i = 0; i < str->length; i++)
if (str->chars[i] == ';') str->chars[i] = '\n';
/* we'll just dump a little tmp file and read it with cm_read */
sprintf(fname, "cm.tmp.%d", getpid());
F = phast_fopen(fname, "w+");
fprintf(F, "%s", str->chars);
phast_fclose(F);
F = phast_fopen(fname, "r");
retval = cm_read(F);
phast_fclose(F);
unlink(fname);
}
else {
F = phast_fopen(str->chars, "r");
retval = cm_read(F);
phast_fclose(F);
}
str_free(str);
return retval;
}
/*
open a file with name out_root.name.maf, or returns it if already open.
This is a bit messy because in some cases (splitting by feature) there may
be more output files than the OS can handle. But it would be computationally
expensive to check and see which files are finished, assuming that the MAF is
sorted.
So, if it tries to open a file and fails, it the goes through the list of
filehandles, finds an open one, closes it, and tries to open the new one
again. Repeat until successful.
Then, if a filehandle needs to be re-opened, it is opened with append. Again,
if this is not successful, it looks for another file to close. If it can't
find one the program reports an error and dies.
Finally, close_outfiles below checks and makes sure that all files
are closed with mafBlock_close_file in the end, so that they get the #eof
closer.
*/
FILE *get_outfile(List *outfileList, Hashtable *outfileHash, String *name, char *out_root,
int argc, char *argv[]) {
int idx, i;
FILE *outfile;
char *fname = smalloc((strlen(out_root)+name->length+7)*sizeof(char));
sprintf(fname, "%s.%s.maf", out_root, name->chars);
idx = ptr_to_int(hsh_get(outfileHash, fname));
if (idx == -1) {
hsh_put(outfileHash, fname, int_to_ptr(lst_size(outfileList)));
outfile = mafBlock_open_outfile(fname, argc, argv);
while (outfile==NULL) { //too many files are open, close one first
for (i=0; i<lst_size(outfileList); i++) {
outfile = (FILE*)lst_get_ptr(outfileList, i);
if (outfile != NULL) break;
}
if (i == lst_size(outfileList)) {
die("ERROR: too many files open in maf_parse\n");
} else {
phast_fclose(outfile);
lst_set_ptr(outfileList, i, NULL);
}
outfile = mafBlock_open_outfile(fname, argc, argv);
}
lst_push_ptr(outfileList, (void*)outfile);
sfree(fname);
return outfile;
}
outfile = (FILE*)lst_get_ptr(outfileList, idx);
if (outfile == NULL) { //has already been opened but then closed.
outfile = phast_fopen_no_exit(fname, "a");
while (outfile == NULL) {
for (i=0; i<lst_size(outfileList); i++) {
outfile = (FILE*)lst_get_ptr(outfileList, i);
if (outfile != NULL) break;
}
if (i == lst_size(outfileList)) {
die("ERROR: too many files open in maf_parse\n");
} else {
phast_fclose(outfile);
lst_set_ptr(outfileList, i, NULL);
}
outfile = phast_fopen_no_exit(fname, "a");
}
lst_set_ptr(outfileList, idx, (void*)outfile);
}
sfree(fname);
return outfile;
}
SEXP rph_gff_read(SEXP filename) {
FILE *infile = phast_fopen(CHARACTER_VALUE(filename), "r");
SEXP rv;
PROTECT(rv = rph_gff_new_extptr(gff_read_set(infile)));
phast_fclose(infile);
UNPROTECT(1);
return rv;
}
/* simple wrapper for cm_read that opens specified filename and reads
a category map from it; aborts with error message if unable to open
the file. Saves typing in mains for command-line programs */
CategoryMap* cm_read_from_fname(char *fname) {
CategoryMap *cm = NULL;
FILE *F;
F = phast_fopen(fname, "r");
if ((cm = cm_read(F)) == NULL)
die("ERROR: cannot read category map from %s.\n", fname);
phast_fclose(F);
return cm;
}
SEXP rph_gff_print(SEXP filename, SEXP gff) {
FILE *outfile;
if (filename == R_NilValue)
outfile = stdout;
else outfile = phast_fopen(CHARACTER_VALUE(filename), "w");
gff_print_set(outfile, (GFF_Set*)EXTPTR_PTR(gff));
if (outfile != stdout) phast_fclose(outfile);
return R_NilValue;
}
/* read in a tree from a file. Return character string
representing tree */
SEXP rph_tree_read(SEXP filename) {
FILE *infile;
char c;
SEXP result;
char *currStr, **strvec;
int i, pos=0, currLen=10000, numparen=0, numtrees_alloc=1000, numtrees=0;
infile = phast_fopen(CHARACTER_VALUE(filename), "r");
currStr = smalloc((currLen+2)*sizeof(char));
strvec = smalloc(numtrees_alloc*sizeof(char*));
while (1) {
pos=0;
numparen=0;
while (';'!=(c=fgetc(infile))) {
if (c==EOF) {
if (pos==0) break;
die("unexpected EOF in tree file. Trees should be terminated by \";\"");
}
if (isspace(c)) continue;
if (c=='(') numparen++;
if (c==')') numparen--;
if (numparen < 0) die("bad tree in tree file");
if (pos==currLen) {
currLen += 10000;
currStr = srealloc(currStr, (currLen+2)*sizeof(char));
}
currStr[pos++] = c;
}
if (pos > 0) {
if (numparen != 0) die("unbalanced parenthesis in tree file");
currStr[pos++]=';';
currStr[pos]='\0';
if (numtrees == numtrees_alloc) {
numtrees_alloc += 1000;
strvec = srealloc(strvec, numtrees_alloc*sizeof(char*));
}
strvec[numtrees] = smalloc((strlen(currStr)+1)*sizeof(char));
strcpy(strvec[numtrees], currStr);
numtrees++;
}
else break;
}
phast_fclose(infile);
PROTECT(result = NEW_CHARACTER(numtrees));
for (i=0; i<numtrees; i++)
SET_STRING_ELT(result, i, mkChar(strvec[i]));
UNPROTECT(1);
return result;
}
SEXP rph_wig_print(SEXP gffP, SEXP filename, SEXP append) {
FILE *outfile;
char *mode;
GFF_Set *gff = (GFF_Set*)EXTPTR_PTR(gffP);
gff_register_protect(gff);
if (LOGICAL_VALUE(append)==TRUE) mode="a";
else mode="w";
if (filename == R_NilValue)
outfile = stdout;
else outfile = phast_fopen(CHARACTER_VALUE(filename), mode);
wig_print(outfile, gff);
if (outfile != stdout) phast_fclose(outfile);
return R_NilValue;
}
LocalPwAlignment *la_read_lav(FILE *F, int read_seqs) {
String *line = str_new(STR_MED_LEN);
int line_no=0;
LocalPwAlignment *lpwa = la_new();
List *fields = lst_new_ptr(6);
Regex *stanza_start_re = str_re_new("^([dshaxm])[[:space:]]*{");
AlignmentBlock *aln_block = NULL;
char stanza_type = '\0';
int i;
int done_with[256];
done_with[(int)'d'] = done_with[(int)'s'] = done_with[(int)'h'] =
done_with[(int)'x'] = done_with[(int)'m'] = 0;
while (str_readline(line, F) != EOF) {
str_trim(line);
if (line->length == 0) continue;
checkInterruptN(line_no, 1000);
line_no++;
if (line_no == 1) {
if (!str_equals_charstr(line, "#:lav")) {
die("ERROR: lav file missing header.\n");
}
}
else if (str_re_match(line, stanza_start_re, fields, 1) >= 0) {
String *tmpstr = lst_get_ptr(fields, 1);
stanza_type = tmpstr->chars[0];
str_free(tmpstr);
str_free(lst_get_ptr(fields, 0));
if (stanza_type != 'a' && done_with[(int)stanza_type]) {
die("ERROR: multiple '%c' stanzas in lav file.\n",
stanza_type);
}
if (stanza_type == 'a') {
aln_block = la_new_alignment_block(-1, -1, -1, -1, -1, NULL);
lst_push_ptr(lpwa->alignment_blocks, aln_block);
}
}
/* end current stanza */
else if (str_equals_charstr(line, "}")) {
if (stanza_type == '\0') {
die("ERROR: end stanza without matching begin.\n");
}
done_with[(int)stanza_type] = 1;
stanza_type = '\0';
}
else if (stanza_type == 'd') {
; /* do nothing for now */
}
else if (stanza_type == 's') {
int beg, end;
String *tmpstr, *fname, *seq=NULL;
FILE *F2;
str_double_trim(line);
str_split(line, NULL, fields);
if (lst_size(fields) != 3 ||
str_as_int(lst_get_ptr(fields, 1), &beg) != 0 ||
str_as_int(lst_get_ptr(fields, 2), &end) != 0) {
die("ERROR: bad line in 's' stanza in lav file.\n");
}
tmpstr = lst_get_ptr(fields, 0);
fname = str_new(tmpstr->length-2); /* remove quotes */
str_substring(fname, tmpstr, 1, tmpstr->length-2);
if (read_seqs) {
F2 = phast_fopen(fname->chars, "r");
seq = msa_read_seq_fasta(F2);
phast_fclose(F2);
}
for (i = 0; i < lst_size(fields); i++)
str_free(lst_get_ptr(fields, i));
if (beg != 1) {
die("ERROR: unexpected begin index in 's' stanza of lav file (begin index currently must be 1).\n");
}
if (lpwa->query_len == -1) {
lpwa->query_len = end;
if (read_seqs) lpwa->query_seq = seq;
}
else if (lpwa->target_len == -1) {
lpwa->target_len = end;
if (read_seqs) lpwa->target_seq = seq;
}
else {
die("ERROR: too many sequences listed in 's' stanza of lav file.\n");
}
str_free(fname);
}
else if (stanza_type == 'h') {
String *name;
str_double_trim(line);
name = str_new(line->length-3); /* get rid of quotes and
leading '>' */
str_substring(name, line, 2, line->length-3);
if (lpwa->query_name == NULL) lpwa->query_name = name;
else if (lpwa->target_name == NULL) lpwa->target_name = name;
else {
die("ERROR: too many entries in 'h' stanza of lav file.\n");
}
}
else if (stanza_type == 'a') {
String *type;
int val[6];
if (!done_with[(int)'s'] || !done_with[(int)'d'] ||
!done_with[(int)'h']) {
die("ERROR: 'a' stanza appears in lav file before 'd', 's', or 'h' stanza.\n");
}
str_double_trim(line);
str_split(line, NULL, fields);
type = lst_get_ptr(fields, 0);
if (lst_size(fields) > 6) {
die("ERROR: illegal line in 'a' stanza.\n");
}
for (i = 1; i < lst_size(fields); i++) {
str_as_int(lst_get_ptr(fields, i), &val[i]);
str_free(lst_get_ptr(fields, i));
}
if (type->chars[0] == 's')
aln_block->score = val[1];
else if (type->chars[0] == 'b') {
aln_block->query_beg = val[1];
aln_block->target_beg = val[2];
}
else if (type->chars[0] == 'e') {
aln_block->query_end = val[1];
aln_block->target_end = val[2];
}
else if (type->chars[0] == 'l')
lst_push_ptr(aln_block->gapless_alns,
la_new_gapless_aln(val[1], val[3], val[2], val[4]));
str_free(type);
}
}
str_free(line);
lst_free(fields);
str_re_free(stanza_start_re);
return lpwa;
}
int main(int argc, char *argv[]) {
int check_start = 0, check_stop = 0, check_splice = 0, check_nonsense = 0,
offset5 = 0, offset3 = 0, opt_idx, i, j, indel_strict = 0, no_output = 0,
check_alignment = 0, splice_strict = 0;
int ncons_tested, nkept, nconserved_exons;
int nce_gap_type[NGAP_TYPES], nconsid[NTYPES], nfail[NTYPES];
double Nfrac = 0.05;
char c;
MSA *msa;
GFF_Set *gff;
msa_format_type msa_format = UNKNOWN_FORMAT;
List *keepers, *problems = lst_new_ptr(10),
*ends_adjusted = lst_new_ptr(1), *starts_adjusted = lst_new_ptr(1),
*discards=NULL, *intron_splice = lst_new_ptr(10);
char *rseq_fname = NULL;
FILE *logf = NULL, *mlogf = NULL, *statsf = NULL, *discardf = NULL;
cds_gap_type fshift_mode = FSHIFT_BAD;
char *groupby = "transcript_id";
msa_coord_map *map;
int *countNs, *countCDSs;
FILE *infile;
char *msa_fname;
struct option long_opts[] = {
{"start", 0, 0, 's'},
{"stop", 0, 0, 't'},
{"splice", 0, 0, 'l'},
{"nonsense", 0, 0, 'n'},
{"fshift", 0, 0, 'f'},
{"conserved", 0, 0, 'c'},
{"N-limit", 1, 0, 'N'},
{"clean-gaps", 0, 0, 'e'},
{"indel-strict", 0, 0, 'I'},
{"splice-strict", 0, 0, 'C'},
{"groupby", 1, 0, 'g'},
{"msa-format", 1, 0, 'i'},
{"refseq", 1, 0, 'r'},
{"offset5", 1, 0, 'o'},
{"offset3", 1, 0, 'p'},
{"no-output", 0, 0, 'x'},
{"discards", 1, 0, 'd'},
{"log", 1, 0, 'L'},
{"machine-log", 1, 0, 'M'},
{"stats", 1, 0, 'S'},
{"help", 0, 0, 'h'},
{0, 0, 0, 0}
};
while ((c = (char)getopt_long(argc, argv, "N:i:r:L:M:S:g:d:stlnfceICxh",
long_opts, &opt_idx)) != -1) {
switch(c) {
case 's':
check_alignment = check_start = 1;
break;
case 't':
check_alignment = check_stop = 1;
break;
case 'l':
check_alignment = check_splice = 1;
break;
case 'n':
check_alignment = check_nonsense = 1;
break;
case 'f':
check_alignment = 1;
fshift_mode = FSHIFT_OK;
break;
case 'c':
check_alignment = check_start = check_stop = check_splice = check_nonsense = 1;
if (fshift_mode < FSHIFT_OK) fshift_mode = FSHIFT_OK;
break;
case 'N':
Nfrac = get_arg_dbl_bounds(optarg, 0, 1);
break;
case 'e':
check_alignment = 1;
if (fshift_mode < CLN_GAPS) fshift_mode = CLN_GAPS;
break;
case 'I':
check_alignment = 1;
fshift_mode = NOVRLP_CLN_GAPS;
indel_strict = 1;
break;
case 'C':
check_alignment = check_splice = splice_strict = 1;
break;
case 'g':
groupby = optarg;
break;
case 'i':
msa_format = msa_str_to_format(optarg);
if (msa_format == UNKNOWN_FORMAT) die("Bad alignment format.\n");
break;
case 'r':
rseq_fname = optarg;
break;
case 'o':
offset5 = get_arg_int(optarg);
break;
case 'p':
offset3 = get_arg_int(optarg);
break;
case 'L':
logf = phast_fopen(optarg, "w+");
break;
case 'M':
mlogf = phast_fopen(optarg, "w+");
break;
case 'S':
statsf = phast_fopen(optarg, "w+");
break;
case 'd':
discardf = phast_fopen(optarg, "w+");
break;
case 'x':
no_output = 1;
break;
case 'h':
printf("%s", HELP);
exit(0);
case '?':
die("ERROR: Bad argument. Try the --help option.\n");
}
}
if (optind + 1 >= argc ) {
die("ERROR: Missing required arguments. Try the --help option.\n");
}
set_seed(-1);
gff = gff_read_set(phast_fopen(argv[optind], "r"));
msa_fname = argv[optind+1];
infile = phast_fopen(msa_fname, "r");
if (msa_format == UNKNOWN_FORMAT)
msa_format = msa_format_for_content(infile, 1);
if (msa_format == MAF) {
msa = maf_read(infile,
rseq_fname == NULL ? NULL : phast_fopen(rseq_fname, "r"),
1, NULL, NULL, NULL, -1, TRUE, NULL, NO_STRIP, FALSE);
}
else {
msa = msa_new_from_file_define_format(infile,
msa_format, NULL);
if (msa->ss == NULL)
ss_from_msas(msa, 1, 1, NULL, NULL, NULL, -1, 0);
}
if (!msa->ss->tuple_idx)
die("ERROR: need ordered tuples\n");
msa_remove_N_from_alph(msa); /* for backward compatibility (old SS files) */
if (msa->idx_offset != 0) { /* avoids offset problem */
for (i = 0; i < lst_size(gff->features); i++) {
GFF_Feature *f = lst_get_ptr(gff->features, i);
f->start -= msa->idx_offset;
f->end -= msa->idx_offset;
}
}
/* set up coordinate map; assume GFF is for sequence 1 */
map = msa_build_coord_map(msa, 1);
/* convert all features */
for (i = 0; i < lst_size(gff->features); i++) {
GFF_Feature *f = lst_get_ptr(gff->features, i);
int newstart, newend;
if (f->start < 0 || f->end < f->start)
die("ERROR: bad feature in GFF (start=%d, end=%d).\n",
f->start, f->end);
newstart = msa_map_seq_to_msa(map, f->start);
newend = msa_map_seq_to_msa(map, f->end);
if (newstart < 0 || newend < newstart)
die("ERROR: unable to map coordinates for feature (start=%d, end=%d).\n",
f->start, f->end);
f->start = newstart;
f->end = newend;
}
gff_group(gff, groupby); /* do this after coord conversion, or
group coords and feature coords
will be out of sync */
keepers = lst_new_ptr(lst_size(gff->features));
if (discardf != NULL) discards = lst_new_ptr(lst_size(gff->features));
ncons_tested = nkept = nconserved_exons = 0;
for (i = 0; i < NTYPES; i++) nconsid[i] = 0;
for (i = 0; i < NTYPES; i++) nfail[i] = 0;
for (i = 0; i < NGAP_TYPES; i++) nce_gap_type[i] = 0;
countNs = smalloc(msa->nseqs * sizeof(int));
countCDSs = smalloc(msa->nseqs * sizeof(int));
for (i = 0; i < lst_size(gff->groups); i++) {
GFF_FeatureGroup *group = lst_get_ptr(gff->groups, i);
List *gfeatures = group->features;
GFF_Feature *feat;
status_type status = OKAY;
cds_gap_type gt = FSHIFT_BAD;
problems_clear(problems);
/* make sure have frame info for CDSs */
for (j = 0; j < lst_size(gfeatures); j++) {
feat = lst_get_ptr(gfeatures, j);
if (str_equals_charstr(feat->feature, GFF_CDS_TYPE) &&
feat->frame == GFF_NULL_FRAME)
die("ERROR: Missing frame info for CDS.\n");
}
/* First, exclude stop codons from cds's, if necessary (simplifies
the detection of nonsense mutations). */
exclude_stops(group, starts_adjusted, ends_adjusted);
/* In all cases, discard any group for which the reference sequence
doesn't have valid splice sites or start/stop codons, or has a
premature stop codon */
if (!ref_seq_okay(gfeatures, msa, offset3, indel_strict, splice_strict,
problems)) {
status = BAD_REF;
nfail[BAD_REF]++;
}
else
/* Everything else counts as a potentially valid group */
ncons_tested++;
if (status == OKAY && check_alignment) {
/* only bother with below if
interested in cross-species
conservation */
/* Check first to make sure there's alignment across species in
the cds; if not, there's no need to look at individual
features. */
for (j = 0; j < lst_size(gfeatures); j++) {
feat = lst_get_ptr(gfeatures, j);
if (str_equals_charstr(feat->feature, GFF_CDS_TYPE) &&
is_incomplete_alignment(feat, msa)) {
status = NO_ALN;
nfail[NO_ALN]++;
problem_add(problems, feat, NO_ALN, -1, -1);
break;
}
}
if (status == OKAY) { /* we have alignment and agreement
with the ref seq; now check feature
by feature */
lst_clear(intron_splice);
for (j = 0; j < msa->nseqs; j++) countNs[j] = countCDSs[j] = 0;
for (j = 0; j < lst_size(gfeatures); j++) {
feat = lst_get_ptr(gfeatures, j);
if (feat->end - 1 >= msa->length)
die("ERROR: feature extends beyond alignment (%d >= %d).\n",
feat->end - 1, msa->length);
if (check_start && str_equals_charstr(feat->feature, GFF_START_TYPE)) {
nconsid[BAD_START]++;
if (!is_conserved_start(feat, msa)) {
status = BAD_START;
problem_add(problems, feat, BAD_START, -1, -1);
}
}
else if (check_stop && str_equals_charstr(feat->feature, GFF_STOP_TYPE)) {
nconsid[BAD_STOP]++;
if (!is_conserved_stop(feat, msa)) {
status = BAD_STOP;
problem_add(problems, feat, BAD_STOP, -1, -1);
}
}
else if (check_splice &&
str_equals_charstr(feat->feature, SPLICE_5)) {
nconsid[BAD_5_SPLICE]++;
if (!is_conserved_5splice(feat, msa, offset5, splice_strict)) {
status = BAD_5_SPLICE;
problem_add(problems, feat, BAD_5_SPLICE, -1, -1);
}
else lst_push_ptr(intron_splice, feat);
}
else if (check_splice &&
str_equals_charstr(feat->feature, SPLICE_5_UTR)) {
nconsid[BAD_5_SPLICE_UTR]++;
if (!is_conserved_5splice(feat, msa, offset5, splice_strict)) {
status = BAD_5_SPLICE_UTR;
problem_add(problems, feat, BAD_5_SPLICE_UTR, -1, -1);
}
else lst_push_ptr(intron_splice, feat);
}
else if (check_splice && str_equals_charstr(feat->feature, SPLICE_3)) {
nconsid[BAD_3_SPLICE]++;
if (!is_conserved_3splice(feat, msa, offset3, splice_strict)) {
status = BAD_3_SPLICE;
problem_add(problems, feat, BAD_3_SPLICE, -1, -1);
}
else lst_push_ptr(intron_splice, feat);
}
else if (check_splice && str_equals_charstr(feat->feature, SPLICE_3)) {
nconsid[BAD_3_SPLICE_UTR]++;
if (!is_conserved_3splice(feat, msa, offset3, splice_strict)) {
status = BAD_3_SPLICE_UTR;
problem_add(problems, feat, BAD_3_SPLICE_UTR, -1, -1);
}
else lst_push_ptr(intron_splice, feat);
}
else if (str_equals_charstr(feat->feature, GFF_CDS_TYPE)) {
if (fshift_mode > FSHIFT_BAD
&& (gt = get_cds_gap_type(feat, msa, problems)) < fshift_mode) {
if (status == OKAY || status == NONSENSE) status = FSHIFT;
}
if (check_nonsense && !is_nonsense_clean(feat, msa, problems)) {
if (status == OKAY) status = NONSENSE;
}
if (Nfrac < 1)
get_N_counts(countNs, countCDSs, feat, msa);
}
} /* end loop through features in group */
/* still have to make sure splice sites are paired correctly
(GT-AG, GC-AG, AT-AC) */
if (status == OKAY && !splice_strict && lst_size(intron_splice) >= 2 &&
!are_introns_okay(intron_splice, msa, problems, offset5, offset3))
status = BAD_INTRON;
/* also check fraction of Ns */
if (Nfrac < 1) {
enum {MY_OKAY, MY_FAIL, MY_WARN} Nstatus = MY_OKAY;
for (j = 0; j < msa->nseqs; j++) {
if ((double)countNs[j] / countCDSs[j] > Nfrac) Nstatus = MY_FAIL;
if (Nstatus == MY_OKAY && countNs[j] > 0) Nstatus = MY_WARN;
}
if (Nstatus == MY_FAIL) {
problem_add(problems, NULL, TOO_MANY_Ns, -1, -1);
if (status == OKAY) status = TOO_MANY_Ns;
}
else if (Nstatus == MY_WARN)
problem_add(problems, NULL, WARN_Ns, -1, -1);
}
/* if collecting stats, record counts for failures */
if (statsf != NULL) {
if (status != OKAY) {
for (j = 0; j < lst_size(problems); j++) {
struct Problem *problem = lst_get_ptr(problems, j);
status_type ftype = problem->status;
if ((ftype == FSHIFT || ftype == NONSENSE) &&
status != FSHIFT && status != NONSENSE)
continue; /* don't count secondary frame shifts
and nonsense mutations */
if (ftype == BAD_INTRON && j % 2 == 0)
continue; /* only count one of every pair of these */
nfail[ftype]++;
}
}
/* also keep track of the total number of "conserved exons", and
the number having each kind of gap */
if ((status == OKAY || (status == FSHIFT && gt >= FSHIFT_OK))) {
nconserved_exons++;
nce_gap_type[gt]++; /* number of conserved exons having
given type of gaps */
}
}
} /* end if (status == OKAY) [checks for conserved features] */
} /* end if (status == OKAY && check_alignment) [all cross-species
checks] */
/* now we have looked at the whole group; we just need to do some
final accounting and logging */
if (status == OKAY) {
nkept++;
if (!no_output) {
restore_stops(group, starts_adjusted, ends_adjusted);
for (j = 0; j < lst_size(gfeatures); j++)
lst_push_ptr(keepers, lst_get_ptr(gfeatures, j));
}
if (logf != NULL && lst_size(problems) > 0) /* warnings only */
write_log(logf, group, status, problems, msa, map);
if (mlogf != NULL) {
/* no problem, need to add an okay status to log */
problem_add(problems, NULL, OKAY, -1, -1);
write_machine_log(mlogf, group, problems, map); /* may include
warnings */
}
}
else {
if (discardf != NULL) {
restore_stops(group, starts_adjusted, ends_adjusted);
for (j = 0; j < lst_size(gfeatures); j++)
lst_push_ptr(discards, lst_get_ptr(gfeatures, j));
}
if (logf != NULL)
write_log(logf, group, status, problems, msa, map);
if (mlogf != NULL)
write_machine_log(mlogf, group, problems, map);
}
} /* end loop over groups */
/* write main output and discards */
if (!no_output || discardf != NULL) {
/* first map features back to coord frame of reference seq. */
for (i = 0; i < lst_size(gff->features); i++) {
GFF_Feature *f = lst_get_ptr(gff->features, i);
f->start = msa_map_msa_to_seq(map, f->start) + msa->idx_offset;
f->end = msa_map_msa_to_seq(map, f->end) + msa->idx_offset;
}
if (!no_output) {
gff->features = keepers;
gff_print_set(stdout, gff);
}
if (discardf != NULL) {
gff->features = discards;
gff_print_set(discardf, gff);
}
}
/* dump counts to stats file */
if (statsf != NULL) {
fprintf(statsf, "#%11s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s %12s\n",
"total", "nbad_ref", "nconsid", "nkept", "nno_aln",
"nbad_starts", "(out of)", "nbad_stops", "(out of)",
"nbad_5spl", "(out of)", "nbad_3spl", "(out of)",
"nbad_5utr", "(out of)", "nbad_3utr", "(out of)",
"nbad_intron", "nnons", "nfshifts", "nNs", "ncons_exons",
"nce_ngaps", "nce_nov_cln", "nce_clean", "nce_fshftok");
fprintf(statsf, "%12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d %12d\n",
nfail[BAD_REF]+ncons_tested, nfail[BAD_REF], ncons_tested, nkept,
nfail[NO_ALN], nfail[BAD_START], nconsid[BAD_START],
nfail[BAD_STOP], nconsid[BAD_STOP], nfail[BAD_5_SPLICE],
nconsid[BAD_5_SPLICE], nfail[BAD_3_SPLICE], nconsid[BAD_3_SPLICE],
nfail[BAD_5_SPLICE_UTR], nconsid[BAD_5_SPLICE_UTR],
nfail[BAD_3_SPLICE_UTR], nconsid[BAD_3_SPLICE_UTR],
nfail[BAD_INTRON], nfail[NONSENSE], nfail[FSHIFT],
nfail[TOO_MANY_Ns], nconserved_exons, nce_gap_type[NGAPS],
nce_gap_type[NOVRLP_CLN_GAPS], nce_gap_type[CLN_GAPS],
nce_gap_type[FSHIFT_OK]);
fprintf(statsf, "%s", STATS_DESCRIPTION);
}
if (logf != NULL) phast_fclose(logf);
if (mlogf != NULL) phast_fclose(mlogf);
if (statsf != NULL) phast_fclose(statsf);
if (discardf != NULL) phast_fclose(discardf);
return 0;
}
int main(int argc, char *argv[]) {
char c;
int opt_idx, node;
FILE *out_f = NULL, *msa_f, *mod_f;
char *out_root;
TreeModel *mod;
MSA *msa;
char out_fname[STR_MED_LEN];
struct option long_opts[] = {
{"refseq", 1, 0, 'r'},
{"msa-format", 1, 0, 'i'},
{"seqs", 1, 0, 's'},
{"exclude", 0, 0, 'x'},
{"no-probs", 0, 0, 'n'},
{"suff-stats", 0, 0, 'S'},
{"encode", 1, 0, 'e'},
{"keep-gaps", 0, 0, 'k'},
{"gibbs", 1, 0, 'G'},
{"help", 0, 0, 'h'},
{0, 0, 0, 0}
};
/* arguments and defaults for options */
FILE *refseq_f = NULL;
msa_format_type msa_format = UNKNOWN_FORMAT;
int suff_stats = FALSE, exclude = FALSE, keep_gaps = FALSE, do_probs = TRUE;
List *seqlist = NULL;
PbsCode *code = NULL;
int gibbs_nsamples = -1;
while ((c = (char)getopt_long(argc, argv, "r:i:s:e:knxSh", long_opts, &opt_idx)) != -1) {
switch (c) {
case 'r':
refseq_f = phast_fopen(optarg, "r");
break;
case 'i':
msa_format = msa_str_to_format(optarg);
if (msa_format == UNKNOWN_FORMAT)
die("ERROR: unrecognized alignment format.\n");
break;
case 'S':
suff_stats = TRUE;
break;
case 'e':
code = pbs_new_from_file(phast_fopen(optarg, "r"));
break;
case 's':
seqlist = get_arg_list(optarg);
break;
case 'x':
exclude = TRUE;
break;
case 'n':
do_probs = FALSE;
break;
case 'k':
keep_gaps = TRUE;
break;
case 'G':
gibbs_nsamples = get_arg_int_bounds(optarg, 1, INFTY);
break;
case 'h':
printf("%s", HELP);
exit(0);
case '?':
die("Bad argument. Try 'prequel -h'.\n");
}
}
if (optind != argc - 3)
die("Three arguments required. Try 'prequel -h'.\n");
set_seed(-1);
if (!do_probs && (suff_stats || code != NULL))
die("ERROR: --no-probs can't be used with --suff-stats or --encode.\n");
msa_f = phast_fopen(argv[optind], "r");
if (msa_format == UNKNOWN_FORMAT)
msa_format = msa_format_for_content(msa_f, 1);
fprintf(stderr, "Reading alignment from %s...\n", argv[optind]);
if (msa_format == MAF) {
msa = maf_read(msa_f, refseq_f, 1, NULL, NULL, NULL, -1, !suff_stats, NULL,
NO_STRIP, FALSE);
/* (no need to store order if suff_stats mode) */
}
else
msa = msa_new_from_file_define_format(msa_f, msa_format, NULL);
if (msa->ss == NULL) {
fprintf(stderr, "Extracting sufficient statistics...\n");
ss_from_msas(msa, 1, TRUE, NULL, NULL, NULL, -1, 0);
}
else if (msa->ss->tuple_idx == NULL && !suff_stats)
die("ERROR: ordered representation of alignment required unless --suff-stats.\n");
mod_f = phast_fopen(argv[optind+1], "r");
out_root = argv[optind+2];
mod = tm_new_from_file(mod_f, 1);
/* MH prune just like in phastcons */
int old_nnodes = mod->tree->nnodes;
List *pruned_names = lst_new_ptr(msa->nseqs);
tm_prune(mod, msa, pruned_names);
if (lst_size(pruned_names) == (old_nnodes + 1) / 2)
die("ERROR: no match for leaves of tree in alignment (leaf names must match alignment names).\n");
if (lst_size(pruned_names) > 0) {
fprintf(stderr, "WARNING: pruned away leaves of tree with no match in alignment (");
int j;
for (j = 0; j < lst_size(pruned_names); j++)
fprintf(stderr, "%s%s", ((String*)lst_get_ptr(pruned_names, j))->chars, j < lst_size(pruned_names) - 1 ? ", " : ").\n");
}
lst_free_strings(pruned_names);
tr_name_ancestors(mod->tree);
if (mod->order != 0)
die("ERROR: Only single nucleotide models are supported.\n");
if (mod->nratecats > 1)
die("ERROR: Rate variation not supported.\n");
mod->tree_posteriors = tl_new_tree_posteriors(mod, msa, TRUE, FALSE,
FALSE, FALSE, FALSE, FALSE,
FALSE);
fprintf(stderr, "Computing posterior probabilities...\n");
if (gibbs_nsamples > 0)
die("ERROR: --gibbs not implemented yet.");
/* gb_sample_ancestral_seqs(mod, msa, mod->tree_posteriors, gibbs_nsamples); */
else
tl_compute_log_likelihood(mod, msa, NULL, NULL, -1, mod->tree_posteriors);
fprintf(stderr, "Reconstructing indels by parsimony...\n");
do_indels(msa, mod);
for (node = 0; node < mod->tree->nnodes; node++) {
int i, j;
TreeNode *n = lst_get_ptr(mod->tree->nodes, node);
if (n->lchild == NULL || n->rchild == NULL) continue;
if (seqlist != NULL) {
int in_list = str_in_list_charstr(n->name, seqlist);
if ((in_list && exclude) || (!in_list && !exclude))
continue;
}
fprintf(stderr, "Writing output for ancestral node '%s'...\n",
n->name);
if (suff_stats) {
if (out_f == NULL) {
sprintf(out_fname, "%s.stats", out_root);
out_f = phast_fopen(out_fname, "w+");
fprintf(out_f, "#count\t");
for (j = 0; j < mod->rate_matrix->size; j++)
fprintf(out_f, "p(%c)%c", mod->rate_matrix->states[j],
j == mod->rate_matrix->size - 1 ? '\n' : '\t');
}
for (i = 0; i < msa->ss->ntuples; i++) {
if (mod->tree_posteriors->base_probs[0][0][node][i] == -1)
continue; /* no base this node */
fprintf(out_f, "%.0f\t", msa->ss->counts[i]);
for (j = 0; j < mod->rate_matrix->size; j++) {
fprintf(out_f, "%f%c",
mod->tree_posteriors->base_probs[0][j][node][i],
j == mod->rate_matrix->size - 1 ? '\n' : '\t');
}
}
}
else if (code == NULL && do_probs) { /* ordinary sequence-by-sequence
output */
sprintf(out_fname, "%s.%s.probs", out_root, n->name);
out_f = phast_fopen(out_fname, "w+");
fprintf(out_f, "#");
for (j = 0; j < mod->rate_matrix->size; j++)
fprintf(out_f, "p(%c)%c", mod->rate_matrix->states[j],
j == mod->rate_matrix->size - 1 ? '\n' : '\t');
for (i = 0; i < msa->length; i++) {
if (mod->tree_posteriors->base_probs[0][0][node][msa->ss->tuple_idx[i]] == -1) {
/* no base */
if (keep_gaps) fprintf(out_f, "-\n");
/* otherwise do nothing */
}
else
for (j = 0; j < mod->rate_matrix->size; j++)
fprintf(out_f, "%f%c",
mod->tree_posteriors->base_probs[0][j][node][msa->ss->tuple_idx[i]],
j == mod->rate_matrix->size - 1 ? '\n' : '\t');
}
phast_fclose(out_f);
}
else if (code == NULL && !do_probs) { /* write point estimates
to FASTA file */
char *outseq = smalloc((msa->length + 1) * sizeof(char));
int len = 0;
for (i = 0; i < msa->length; i++) {
if (mod->tree_posteriors->base_probs[0][0][node][msa->ss->tuple_idx[i]] == -1) {
/* no base */
if (keep_gaps) outseq[len++] = GAP_CHAR;
/* otherwise do nothing */
}
else {
double maxprob = 0;
int maxidx = -1;
for (j = 0; j < mod->rate_matrix->size; j++) {
if (mod->tree_posteriors->base_probs[0][j][node][msa->ss->tuple_idx[i]] > maxprob) {
maxprob = mod->tree_posteriors->base_probs[0][j][node][msa->ss->tuple_idx[i]];
maxidx = j;
}
}
outseq[len++] = mod->rate_matrix->states[maxidx];
}
}
outseq[len] = '\0';
/* print in FASTA format */
sprintf(out_fname, "%s.%s.fa", out_root, n->name);
out_f = phast_fopen(out_fname, "w+");
print_seq_fasta(out_f, outseq, n->name, len);
phast_fclose(out_f);
sfree(outseq);
}
else { /* encoded sequence-by-sequence
output */
double error, tot_error = 0;
int ngaps = 0;
Vector *v;
unsigned *encoded;
/* first encode tuple by tuple */
v = vec_new(mod->rate_matrix->size);
encoded = smalloc(msa->ss->ntuples * sizeof(unsigned));
for (i = 0; i < msa->ss->ntuples; i++) {
if (mod->tree_posteriors->base_probs[0][0][node][i] == -1) {
encoded[i] = code->gap_code;
ngaps += msa->ss->counts[i];
}
else {
for (j = 0; j < mod->rate_matrix->size; j++)
vec_set(v, j, mod->tree_posteriors->base_probs[0][j][node][i]);
encoded[i] = pbs_get_index(code, v, &error);
tot_error += error * msa->ss->counts[i];
}
}
vec_free(v);
/* now write site by site */
sprintf(out_fname, "%s.%s.bin", out_root, n->name);
out_f = phast_fopen(out_fname, "w+");
for (i = 0; i < msa->length; i++) {
if (keep_gaps || encoded[msa->ss->tuple_idx[i]] != code->gap_code)
pbs_write_binary(code, encoded[msa->ss->tuple_idx[i]], out_f);
}
fprintf(stderr, "Average approximation error ('%s'): %f bits\n",
n->name, tot_error/(msa->length - ngaps));
sfree(encoded);
}
}
fprintf(stderr, "Done.\n");
return 0;
}
int main(int argc, char* argv[]) {
FILE* F;
GFF_Set *gff_real=NULL, *gff_pred=NULL;
char c;
List *real_fname_list = NULL, *pred_fname_list = NULL,
*feat_list = NULL, *seq_len_list = NULL, *l = NULL;
int nfile, i, j;
char *prefix = NULL;
int tot_tp = 0, tot_fp = 0, tot_nreal_pos = 0, tot_npred_pos = 0,
tot_seqlen = 0, tot_ncr = 0, tot_npca = 0, tot_nola = 0, tot_nme = 0,
tot_npcp = 0, tot_nolp = 0, tot_nwe = 0, tot_nexons_real = 0,
tot_nexons_pred = 0, dump_exons = 0, nnc = -1, tot_nnc = -1,
nc_threshold = 0;
while ((c = (char)getopt(argc, argv, "r:p:f:l:d:n:h")) != -1) {
switch(c) {
case 'r':
real_fname_list = get_arg_list(optarg);
break;
case 'p':
pred_fname_list = get_arg_list(optarg);
break;
case 'l':
l = get_arg_list(optarg);
/* convert to ints */
seq_len_list = lst_new_int(lst_size(l));
for (i = 0; i < lst_size(l); i++) {
int tmp;
if (str_as_int((String*)lst_get_ptr(l, i),
&tmp) != 0) {
die("ERROR: Bad integer in <seq_len_list>.\n");
}
lst_push_int(seq_len_list, tmp);
}
break;
case 'f':
feat_list = get_arg_list(optarg);
break;
case 'd':
dump_exons = 1;
prefix = optarg;
break;
case 'n':
nnc = tot_nnc = 0;
nc_threshold = get_arg_int(optarg);
break;
case 'h':
print_usage();
exit(0);
case '?':
die("Unrecognized option. Try \"eval_predictions -h\" for help.\n");
}
}
set_seed(-1);
if (feat_list == NULL) {
feat_list = lst_new_ptr(1);
lst_push_ptr(feat_list, str_new_charstr(GFF_CDS_TYPE));
}
if (real_fname_list == NULL || pred_fname_list == NULL ||
seq_len_list == NULL) {
die("ERROR: Must specify -r, -p, and -l. Try \"eval_predictions -h\" for help.\n");
}
if (lst_size(real_fname_list) != lst_size(pred_fname_list)) {
die("ERROR: Must specify lists of equal length for real and predicted filenames.\n\n.");
}
if (lst_size(seq_len_list) == 1 && lst_size(real_fname_list) > 1)
for (i = 1; i < lst_size(real_fname_list); i++)
lst_push_int(seq_len_list, lst_get_int(seq_len_list, 0));
else if (lst_size(seq_len_list) != lst_size(real_fname_list))
die("ERROR: List of sequence lengths does not match lists of real and predicted filenames.\n");
/* print header */
printf("%-25s %-25s %7s %7s %7s %7s %7s %7s %7s %7s %7s %7s %7s %7s %7s %7s", "Real_fname", "Pred_fname", "Sn", "Sp", "AC", "CC", "ESn", "ESp", "CRa", "PCa", "OLa", "ME", "CRp", "PCp", "OLp", "WE");
if (nnc != -1) printf(" %7s %7s %7s %7s", "NCa", "NCp", "CR+NCa", "CR+NCp");
printf("\n");
for (nfile = 0; nfile < lst_size(real_fname_list); nfile++) {
int tp, fp, nexons_real, nexons_pred, nwe, nme, ncr, npca,
npcp, nola, nolp, nreal_pos, npred_pos, len_real, len_pred, seqlen,
already_counted_real;
String *real_fname, *pred_fname;
GFF_Feature *feat_real, *feat_pred=NULL;
real_fname = (String*)lst_get_ptr(real_fname_list, nfile);
F = phast_fopen(real_fname->chars, "r");
if ((gff_real = gff_read_set(F)) == NULL) {
die("ERROR: Unable to read file \"%s\".\n",
real_fname->chars);
}
phast_fclose(F);
pred_fname = (String*)lst_get_ptr(pred_fname_list, nfile);
F = phast_fopen(pred_fname->chars, "r");
if ((gff_pred = gff_read_set(F)) == NULL) {
die("ERROR: Unable to read file \"%s\".\n",
pred_fname->chars);
}
phast_fclose(F);
seqlen = lst_get_int(seq_len_list, nfile);
/* sort ungrouped -- only cds exons will be considered, and each
one will be considered individually */
gff_ungroup(gff_real);
gff_ungroup(gff_pred);
gff_sort(gff_real);
gff_sort(gff_pred);
nexons_real = nexons_pred = nwe = nme = ncr = npca = npcp = nola =
nolp = tp = fp = nreal_pos = npred_pos = 0;
if (nnc != -1) nnc = 0;
i = j = 0;
already_counted_real = 0;
while (i < lst_size(gff_real->features)) {
feat_real = (GFF_Feature*)lst_get_ptr(gff_real->features, i);
if (!is_exon(feat_real, feat_list)) { i++; continue; }
len_real = feat_real->end - feat_real->start + 1;
if (!already_counted_real) {
nexons_real++;
nreal_pos += len_real;
}
/* look at all predicted exons up to and overlapping this real exon */
while (j < lst_size(gff_pred->features)) {
feat_pred = (GFF_Feature*)lst_get_ptr(gff_pred->features, j);
if (!is_exon(feat_pred, feat_list)) {
j++;
continue;
}
else if (feat_pred->start > feat_real->end) {
if (!already_counted_real) {
nme++;
if (dump_exons) dump(prefix, feat_real, NULL, ME, -1);
}
break;
}
/* otherwise we have a predicted exon to count (start of pred
<= end of real) */
nexons_pred++;
len_pred = feat_pred->end - feat_pred->start + 1;
npred_pos += len_pred;
j++; /* we'll be done with this prediction
one way or another; next time
through look at a new one */
if (feat_pred->end < feat_real->start) { /* WE */
nwe++;
fp += len_pred;
if (dump_exons) dump(prefix, NULL, feat_pred, WE, 0);
}
else if (feat_pred->start == feat_real->start && /* CR */
feat_pred->end == feat_real->end) {
ncr++;
tp += len_pred;
if (dump_exons) dump(prefix, feat_real, feat_pred, CR, 1);
break;
}
else if (feat_pred->start == feat_real->start || /* PC */
feat_pred->end == feat_real->end) {
pred_type type;
npca++;
npcp++;
if (nnc != -1 &&
max(abs(feat_pred->start - feat_real->start),
abs(feat_pred->end - feat_real->end)) <= nc_threshold) {
nnc++;
type = NC;
}
else type = PC;
if (len_pred < len_real)
tp += len_pred;
else {
tp += len_real;
fp += (len_pred - len_real);
}
if (dump_exons) dump(prefix, feat_real, feat_pred, type,
min(1, (double)len_real/len_pred));
break;
}
else { /* OL */
int overlap_size;
pred_type type;
nola++;
nolp++;
if (nnc != -1 &&
max(abs(feat_pred->start - feat_real->start),
abs(feat_pred->end - feat_real->end)) <= nc_threshold) {
nnc++;
type = NC;
}
else type = PC;
overlap_size = min(feat_pred->end, feat_real->end) -
max(feat_pred->start, feat_real->start) + 1;
tp += overlap_size;
fp += len_pred - overlap_size;
if (dump_exons) dump(prefix, feat_real, feat_pred, type,
(double)overlap_size/len_pred);
break;
}
/* NOTE: I'm ignoring the possibility that a predicted exon
could be a PC and/or OL with respect to multiple real
exons. The effect on the exon-level stats will be fairly
minor (at worst a predicted exon is scored as an OL when it
should be scored as an PC, and a real exon is erroneously
counted as a ME), but the effect on the nucleotide-level Sn
and Sp could conceivably be significant. */
}
/* if we have counted at least one prediction (and thus failed
to reach the end of the list), but the last prediction did
not extend as far as the end of the real exon, then delay
moving on to the next real exon */
if (j < lst_size(gff_pred->features) && feat_pred->end < feat_real->end)
already_counted_real = 1;
else {
/* if we reached the end of the list of predictions, then it
must not have contained any exons, and the real exon in
question is a ME (if it hasn't already been counted) */
if (j == lst_size(gff_pred->features) && !already_counted_real)
nme++;
i++;
already_counted_real = 0;
}
}
/* any remaining predictions must be wrong */
for (; j < lst_size(gff_pred->features); j++) {
if (is_exon((GFF_Feature*)lst_get_ptr(gff_pred->features, j),
feat_list)) {
nexons_pred++;
nwe++;
}
}
compute_and_print_stats(stdout, real_fname, pred_fname,
tp, fp, nreal_pos, npred_pos, seqlen, ncr,
npca, nola, nme, npcp, nolp, nwe,
nexons_real, nexons_pred, nnc);
tot_tp += tp;
tot_fp += fp;
tot_nreal_pos += nreal_pos;
tot_npred_pos += npred_pos;
tot_seqlen += seqlen;
tot_ncr += ncr;
tot_npca += npca;
tot_nola += nola;
tot_nme += nme;
tot_npcp += npcp;
tot_nolp += nolp;
tot_nwe += nwe;
tot_nexons_real += nexons_real;
tot_nexons_pred += nexons_pred;
if (nnc != -1) tot_nnc += nnc;
if (dump_exons && SUMF != NULL)
fprintf(SUMF, "# Total number of bases in real exons: %d\n", nreal_pos);
gff_free_set(gff_real);
gff_free_set(gff_pred);
}
if (lst_size(real_fname_list) > 1)
compute_and_print_stats(stdout, str_new_charstr("TOTAL"), str_new_charstr(""),
tot_tp, tot_fp, tot_nreal_pos, tot_npred_pos,
tot_seqlen, tot_ncr, tot_npca, tot_nola, tot_nme,
tot_npcp, tot_nolp, tot_nwe, tot_nexons_real,
tot_nexons_pred, tot_nnc);
return 0;
}
void ms_print_to_file(const char *filename, MS *ms) {
FILE *outfile = phast_fopen(filename, "w");
ms_print_fasta(outfile, ms);
phast_fclose(outfile);
}
void mafBlock_close_outfile(FILE *outfile) {
fprintf(outfile, "#eof\n");
if (outfile != stdout) phast_fclose(outfile);
}
int main(int argc, char *argv[]) {
char *msa_fname = NULL, *alph = "ACGT";
msa_format_type input_format = UNKNOWN_FORMAT;
char c;
int opt_idx, seed=-1;
String *optstr;
List *tmplist = NULL;
struct phyloFit_struct *pf;
FILE *infile;
struct option long_opts[] = {
{"msa", 1, 0, 'm'},
{"tree", 1, 0, 't'},
{"subst-mod", 1, 0, 's'},
{"msa-format", 1, 0, 'i'},
{"nrates", 1, 0, 'k'},
{"alpha", 1, 0, 'a'},
{"features", 1, 0, 'g'},
{"catmap", 1, 0, 'c'},
{"log", 1, 0, 'l'},
{"out-root", 1, 0, 'o'},
{"EM", 0, 0, 'E'},
{"error", 1, 0, 'e'},
{"precision", 1, 0, 'p'},
{"do-cats", 1, 0, 'C'},
{"non-overlapping", 0, 0, 'V'},
{"markov", 0, 0, 'N'},
{"reverse-groups", 1, 0, 'R'},
{"init-model", 1, 0, 'M'},
{"init-random", 0, 0, 'r'},
{"init-parsimony", 0, 0, 'y'},
{"print-parsimony", 1, 0, 'Y'},
{"lnl", 0, 0, 'L'},
{"scale-only", 0, 0, 'B'},
{"scale-subtree", 1, 0, 'S'},
{"estimate-freqs", 0, 0, 'F'},
{"sym-freqs", 0, 0, 'W'},
{"no-freqs", 0, 0, 'f'},
{"no-rates", 0, 0, 'n'},
{"no-opt", 1, 0, 'O'},
{"min-informative", 1, 0, 'I'},
{"gaps-as-bases", 0, 0, 'G'},
{"quiet", 0, 0, 'q'},
{"help", 0, 0, 'h'},
{"windows", 1, 0, 'w'},
{"windows-explicit", 1, 0, 'v'},
{"ancestor", 1, 0, 'A'},
{"post-probs", 0, 0, 'P'},
{"expected-subs", 0, 0, 'X'},
{"expected-total-subs", 0, 0, 'Z'},
{"expected-subs-col", 0, 0, 'J'},
{"column-probs", 0, 0, 'U'},
{"rate-constants", 1, 0, 'K'},
{"ignore-branches", 1, 0, 'b'},
{"clock", 0, 0, 'z'},
{"alt-model", 1, 0, 'd'},
{"label-branches", 1, 0, 0},
{"label-subtree", 1, 0, 0},
{"selection", 1, 0, 0},
{"bound", 1, 0, 'u'},
{"seed", 1, 0, 'D'},
{0, 0, 0, 0}
};
// NOTE: remaining shortcuts left: HjQx
pf = phyloFit_struct_new(0);
while ((c = (char)getopt_long(argc, argv, "m:t:s:g:c:C:i:o:k:a:l:w:v:M:p:A:I:K:S:b:d:O:u:Y:e:D:GVENRqLPXZUBFfnrzhWyJ", long_opts, &opt_idx)) != -1) {
switch(c) {
case 'm':
msa_fname = optarg;
break;
case 't':
if (optarg[0] == '(') /* in this case, assume topology given
at command line */
pf->tree = tr_new_from_string(optarg);
else
pf->tree = tr_new_from_file(phast_fopen(optarg, "r"));
break;
case 's':
pf->subst_mod = tm_get_subst_mod_type(optarg);
if (pf->subst_mod == UNDEF_MOD)
die("ERROR: illegal substitution model. Type \"phyloFit -h\" for usage.\n");
break;
case 'g':
pf->gff = gff_read_set(phast_fopen(optarg, "r"));
break;
case 'c':
pf->cm = cm_new_string_or_file(optarg);
break;
case 'C':
pf->cats_to_do_str = get_arg_list(optarg);
break;
case 'V':
pf->nonoverlapping = TRUE;
break;
case 'o':
pf->output_fname_root = optarg;
break;
case 'k':
pf->nratecats = get_arg_int_bounds(optarg, 0, INFTY);
break;
case 'a':
pf->alpha = get_arg_dbl(optarg);
break;
case 'R':
pf->reverse_group_tag = optarg;
break;
case 'i':
input_format = msa_str_to_format(optarg);
if (input_format == UNKNOWN_FORMAT)
die("ERROR: unrecognized alignment format. Type 'phyloFit -h' for usage.\n");
break;
case 'l':
if (!strcmp(optarg, "-"))
pf->logf = stderr;
else pf->logf = phast_fopen(optarg, "w+");
break;
case 'N':
pf->use_conditionals = 1;
break;
case 'w':
tmplist = get_arg_list(optarg);
if (lst_size(tmplist) != 2 ||
str_as_int(lst_get_ptr(tmplist, 0), &(pf->window_size)) != 0 ||
str_as_int(lst_get_ptr(tmplist, 1), &(pf->window_shift)) != 0)
die("ERROR: illegal arguments to --windows.\n");
lst_free_strings(tmplist);
lst_free(tmplist);
break;
case 'v':
tmplist = get_arg_list(optarg);
if (lst_size(tmplist) % 2 != 0)
die("ERROR: argument to --windows-explicit must be a list of even length.\n");
pf->window_coords = str_list_as_int(tmplist);
lst_free(tmplist);
break;
case 'E':
pf->use_em = TRUE;
break;
case 'e':
pf->error_fname=optarg;
break;
case 'p':
if (!strcmp(optarg, "LOW")) pf->precision = OPT_LOW_PREC;
else if (!strcmp(optarg, "MED")) pf->precision = OPT_MED_PREC;
else if (!strcmp(optarg, "HIGH")) pf->precision = OPT_HIGH_PREC;
else if (!strcmp(optarg, "VERY_HIGH")) pf->precision = OPT_VERY_HIGH_PREC;
else die("ERROR: --precision must be LOW, MED, or HIGH.\n\n");
break;
case 'M':
pf->input_mod = tm_new_from_file(phast_fopen(optarg, "r"), 1);
break;
case 'r':
pf->random_init = TRUE;
break;
case 'y':
pf->init_parsimony = TRUE;
break;
case 'Y':
pf->init_parsimony = TRUE;
pf->parsimony_cost_fname = optarg;
pf->parsimony_only=TRUE;
break;
case 'L':
pf->likelihood_only = TRUE;
break;
case 'q':
pf->quiet = TRUE;
break;
case 'P':
pf->do_bases = TRUE;
break;
case 'X':
pf->do_expected_nsubst = TRUE;
break;
case 'Z':
pf->do_expected_nsubst_tot = TRUE;
break;
case 'J':
pf->do_expected_nsubst_col = TRUE;
break;
case 'U':
pf->likelihood_only = TRUE; /* force -L */
pf->nsites_threshold = 0; /* also force this; typical use is
with small number of tuples, no
tuple_idx */
pf->do_column_probs = TRUE;
break;
case 'A':
pf->root_seqname = optarg;
break;
case 'I':
pf->nsites_threshold = get_arg_int(optarg);
break;
case 'G':
pf->gaps_as_bases = TRUE;
alph = "ACGT-";
break;
case 'B':
pf->estimate_scale_only = TRUE;
break;
case 'S':
pf->subtree_name = optarg;
break;
case 'F':
pf->estimate_backgd = TRUE;
break;
case 'W':
pf->estimate_backgd = TRUE;
pf->symfreq = TRUE;
break;
case 'f':
pf->no_freqs = TRUE;
break;
case 'n':
pf->no_rates = TRUE;
break;
case 'K':
tmplist = get_arg_list(optarg);
pf->rate_consts = str_list_as_dbl(tmplist);
pf->nratecats = lst_size(pf->rate_consts);
pf->use_em = 1;
lst_free_strings(tmplist); lst_free(tmplist);
break;
case 'b':
pf->ignore_branches = get_arg_list(optarg);
break;
case 'z':
pf->assume_clock = TRUE;
break;
case 'O':
if (pf->nooptstr == NULL)
pf->nooptstr = str_new_charstr(optarg);
else die("ERROR: no-opt argument can only be used once! parameters can be comma-separated list.");
break;
case 'd':
if (pf->alt_mod_str == NULL) {
pf->alt_mod_str = lst_new_ptr(1);
}
optstr = str_new_charstr(optarg);
lst_push_ptr(pf->alt_mod_str, optstr);
break;
case 0:
if (strcmp(long_opts[opt_idx].name, "label-branches") == 0 ||
strcmp(long_opts[opt_idx].name, "label-subtree") == 0) {
optstr = str_new_charstr(optarg);
if (pf->label_str == NULL) {
pf->label_str = lst_new_ptr(3);
pf->label_type = lst_new_int(3);
}
lst_push_ptr(pf->label_str, optstr);
lst_push_int(pf->label_type,
strcmp(long_opts[opt_idx].name, "label-branches") == 0 ?
BRANCH_TYPE : SUBTREE_TYPE);
}
else if (strcmp(long_opts[opt_idx].name, "selection") == 0) {
pf->selection = get_arg_dbl(optarg);
pf->use_selection = TRUE;
}
else {
die("ERROR: unknown option. Type 'phyloFit -h' for usage.\n");
}
break;
case 'u':
if (pf->bound_arg == NULL)
pf->bound_arg = lst_new_ptr(1);
optstr = str_new_charstr(optarg);
lst_push_ptr(pf->bound_arg, optstr);
break;
case 'D':
seed = get_arg_int_bounds(optarg, 1, INFTY);
break;
case 'h':
printf("%s", HELP);
exit(0);
case '?':
die("ERROR: illegal argument. Type 'phyloFit -h' for usage.\n");
}
}
set_seed(seed);
if (msa_fname == NULL) {
if (optind >= argc)
die("ERROR: missing alignment filename. Type 'phyloFit -h' for usage.\n");
msa_fname = argv[optind];
pf->msa_fname = msa_fname;
}
infile = phast_fopen(msa_fname, "r");
if (input_format == UNKNOWN_FORMAT)
input_format = msa_format_for_content(infile, 1);
if (pf->nonoverlapping && (pf->use_conditionals || pf->gff != NULL ||
pf->cats_to_do_str || input_format == SS))
die("ERROR: cannot use --non-overlapping with --markov, --features,\n--msa-format SS, or --do-cats.\n");
/* read alignment */
if (!pf->quiet) fprintf(stderr, "Reading alignment from %s ...\n", msa_fname);
if (input_format == MAF) {
pf->msa = maf_read(infile, NULL,
tm_order(pf->subst_mod) + 1,
NULL, pf->gff, pf->cm,
pf->nonoverlapping ? tm_order(pf->subst_mod) + 1 : -1,
FALSE, pf->reverse_group_tag, NO_STRIP, FALSE);
if (pf->gaps_as_bases)
msa_reset_alphabet(pf->msa, alph);
}
else
pf->msa = msa_new_from_file_define_format(infile,
input_format, alph);
/* set up for categories */
/* first label sites, if necessary */
pf->label_categories = (input_format != MAF);
run_phyloFit(pf);
if (pf->logf != NULL && pf->logf != stderr && pf->logf != stdout)
phast_fclose(pf->logf);
if (!pf->quiet) fprintf(stderr, "Done.\n");
sfree(pf);
return 0;
}
SEXP rph_phyloFit(SEXP msaP,
SEXP treeStrP,
SEXP substModP,
SEXP scaleOnlyP,
SEXP scaleSubtreeP,
SEXP nratesP,
SEXP alphaP,
SEXP rateConstantsP,
SEXP initModP,
SEXP initBackgdFromDataP,
SEXP initRandomP,
SEXP initParsimonyP,
SEXP clockP,
SEXP emP,
SEXP maxEmItsP,
SEXP precisionP,
SEXP gffP,
SEXP ninfSitesP,
SEXP quietP,
SEXP noOptP,
SEXP boundP,
SEXP logFileP,
SEXP selectionP) {
struct phyloFit_struct *pf;
int numProtect=0, i;
double *doubleP;
char *die_message=NULL;
SEXP rv=R_NilValue;
List *new_rate_consts = NULL;
List *new_rate_weights = NULL;
GetRNGstate(); //seed R's random number generator
pf = phyloFit_struct_new(1); //sets appropriate defaults for RPHAST mode
pf->msa = (MSA*)EXTPTR_PTR(msaP);
if (treeStrP != R_NilValue)
pf->tree = rph_tree_new(treeStrP);
pf->use_em = LOGICAL_VALUE(emP);
if (rateConstantsP != R_NilValue) {
PROTECT(rateConstantsP = AS_NUMERIC(rateConstantsP));
numProtect++;
doubleP = NUMERIC_POINTER(rateConstantsP);
new_rate_consts = lst_new_dbl(LENGTH(rateConstantsP));
for (i=0; i < LENGTH(rateConstantsP); i++)
lst_push_dbl(new_rate_consts, doubleP[i]);
// pf->use_em = 1;
}
if (initModP != R_NilValue) {
pf->input_mod = (TreeModel*)EXTPTR_PTR(initModP);
pf->subst_mod = pf->input_mod->subst_mod;
tm_register_protect(pf->input_mod);
if (new_rate_consts == NULL && pf->input_mod->rK != NULL && pf->input_mod->nratecats > 1) {
new_rate_consts = lst_new_dbl(pf->input_mod->nratecats);
for (i=0; i < pf->input_mod->nratecats; i++)
lst_push_dbl(new_rate_consts, pf->input_mod->rK[i]);
// pf-> = 1;
}
if (pf->input_mod->empirical_rates && pf->input_mod->freqK != NULL && pf->input_mod->nratecats > 1) {
new_rate_weights = lst_new_dbl(pf->input_mod->nratecats);
for (i=0; i < pf->input_mod->nratecats; i++)
lst_push_dbl(new_rate_weights, pf->input_mod->freqK[i]);
}
tm_reinit(pf->input_mod,
rph_get_subst_mod(substModP),
nratesP == R_NilValue ? pf->input_mod->nratecats : INTEGER_VALUE(nratesP),
NUMERIC_VALUE(alphaP),
new_rate_consts,
new_rate_weights);
} else {
if (nratesP != R_NilValue)
pf->nratecats = INTEGER_VALUE(nratesP);
if (alphaP != R_NilValue)
pf->alpha = NUMERIC_VALUE(alphaP);
if (rateConstantsP != R_NilValue) {
pf->rate_consts = new_rate_consts;
if (nratesP == R_NilValue)
pf->nratecats = lst_size(new_rate_consts);
else if (lst_size(new_rate_consts) != pf->nratecats)
die("length of new_rate_consts does not match nratecats\n");
}
}
pf->subst_mod = rph_get_subst_mod(substModP);
pf->estimate_scale_only = LOGICAL_VALUE(scaleOnlyP);
if (scaleSubtreeP != R_NilValue) {
pf->subtree_name = smalloc((1+strlen(CHARACTER_VALUE(scaleSubtreeP)))*sizeof(char));
strcpy(pf->subtree_name, CHARACTER_VALUE(scaleSubtreeP));
}
pf->random_init = LOGICAL_VALUE(initRandomP);
pf->init_backgd_from_data = LOGICAL_VALUE(initBackgdFromDataP);
pf->init_parsimony = LOGICAL_VALUE(initParsimonyP);
pf->assume_clock = LOGICAL_VALUE(clockP);
if (maxEmItsP != R_NilValue)
pf->max_em_its = INTEGER_VALUE(maxEmItsP);
pf->precision = get_precision(CHARACTER_VALUE(precisionP));
if (pf->precision == OPT_UNKNOWN_PREC) {
die_message = "invalid precision";
goto rph_phyloFit_end;
}
if (gffP != R_NilValue) {
pf->gff = (GFF_Set*)EXTPTR_PTR(gffP);
gff_register_protect(pf->gff);
}
if (ninfSitesP != R_NilValue)
pf->nsites_threshold = INTEGER_VALUE(ninfSitesP);
pf->quiet = LOGICAL_VALUE(quietP);
if (noOptP != R_NilValue) {
int len=LENGTH(noOptP), pos=0;
char *temp;
for (i=0; i < LENGTH(noOptP); i++)
len += strlen(CHARACTER_VALUE(STRING_ELT(noOptP, i)));
temp = smalloc(len*sizeof(char));
for (i=0; i < LENGTH(noOptP); i++) {
if (i != 0) temp[pos++] = ',';
sprintf(&temp[pos], "%s", CHARACTER_VALUE(STRING_ELT(noOptP, i)));
pos += strlen(CHARACTER_VALUE(STRING_ELT(noOptP, i)));
}
if (pos != len-1) die("ERROR parsing noOpt len=%i pos=%i\n", len, pos);
temp[pos] = '\0';
pf->nooptstr = str_new_charstr(temp);
}
if (boundP != R_NilValue) {
pf->bound_arg = lst_new_ptr(LENGTH(boundP));
for (i=0; i < LENGTH(boundP); i++) {
String *temp = str_new_charstr(CHARACTER_VALUE(STRING_ELT(boundP, i)));
lst_push_ptr(pf->bound_arg, temp);
}
}
if (logFileP != R_NilValue) {
if (IS_CHARACTER(logFileP))
pf->logf = phast_fopen(CHARACTER_VALUE(logFileP), "w+");
else if (IS_LOGICAL(logFileP) &&
LOGICAL_VALUE(logFileP)) {
pf->logf = stdout;
}
}
if (selectionP != R_NilValue) {
pf->use_selection = TRUE;
pf->selection = NUMERIC_VALUE(selectionP);
}
msa_register_protect(pf->msa);
run_phyloFit(pf);
rv = PROTECT(rph_listOfLists_to_SEXP(pf->results));
numProtect++;
rph_phyloFit_end:
if (pf->logf != NULL && pf->logf != stdout && pf->logf != stderr)
phast_fclose(pf->logf);
PutRNGstate();
if (die_message != NULL) die(die_message);
if (numProtect > 0)
UNPROTECT(numProtect);
return rv;
}
int main(int argc, char* argv[]) {
FILE* F;
MSA *msa;
int *msa_gap_patterns = NULL;
HMM *hmm = NULL;
TreeNode *tree = NULL;
int i, input_format = SS, msa_idx, quiet_mode = FALSE,
ncats, nmsas, ncats_unspooled, indel_nseqs = -1;
String *msa_fname, *gff_fname;
List *gff_fname_list = NULL, *msa_fname_list = NULL,
*msa_length_list = NULL, *model_indels_str = NULL;
Matrix *traincounts = NULL;
Vector *begcounts = NULL, *statecounts = NULL;
CategoryMap *cm = NULL;
char c;
GapPatternMap *gpm = NULL;
GFF_Set *gff;
char *reverse_groups_tag = NULL;
while ((c = getopt(argc, argv, "i:g:c:m:M:R:I:n:t:P:G:qh")) != -1) {
switch(c) {
case 'i':
input_format = msa_str_to_format(optarg);
if (input_format == -1)
die("ERROR: bad alignment format.\n");
break;
case 'g':
gff_fname_list = get_arg_list(optarg);
break;
case 'c':
cm = cm_new_string_or_file(optarg);
break;
case 'm':
msa_fname_list = get_arg_list(optarg);
break;
case 'M':
msa_length_list = str_list_as_int(get_arg_list(optarg));
break;
case 'R':
reverse_groups_tag = optarg;
break;
case 'I':
model_indels_str = get_arg_list(optarg);
break;
case 'n':
indel_nseqs = get_arg_int(optarg);
break;
case 't':
if (optarg[0] == '(') /* in this case, assume topology given
at command line */
tree = tr_new_from_string(optarg);
else
tree = tr_new_from_file(phast_fopen(optarg, "r"));
break;
case 'q':
quiet_mode = TRUE;
break;
case 'h':
print_usage();
exit(0);
case '?':
die("ERROR: unrecognized option.\n\nType 'hmm_train -h' for usage.\n");
}
}
if (msa_fname_list == NULL)
die("ERROR: -m required. Type 'hmm_train -h' for usage.\n");
if (gff_fname_list == NULL)
die("ERROR: -g required in training mode. Type 'hmm_train -h' for usage.\n");
if (msa_length_list != NULL && msa_fname_list != NULL)
die("ERROR: -m and -M are mutually exclusive. Type 'hmm_train -h' for usage.\n");
if (model_indels_str != NULL && tree == NULL)
die("ERROR: -I requires -t. Type 'hmm_train -h' for usage.\n");
if (cm == NULL)
die("ERROR: category map required.\n");
set_seed(-1);
ncats = cm->ncats + 1;
ncats_unspooled = cm->unspooler != NULL ? cm->unspooler->nstates_unspooled :
ncats;
nmsas = (msa_length_list != NULL ? lst_size(msa_length_list) :
lst_size(msa_fname_list));
if (model_indels_str != NULL) {
if (tree == NULL)
die("ERROR: tree is NULL\n"); /*FIXME: indel_ncats broken */
gpm = gp_create_gapcats(cm, model_indels_str, tree, FALSE);
ncats = cm->ncats + 1; /* numbers will change */
ncats_unspooled = cm->unspooler == NULL ? ncats :
cm->unspooler->nstates_unspooled;
}
/* allocate memory for storage of "training paths" */
traincounts = mat_new(ncats_unspooled, ncats_unspooled);
statecounts = vec_new(ncats_unspooled);
begcounts = vec_new(ncats_unspooled);
mat_zero(traincounts);
vec_zero(statecounts);
vec_zero(begcounts);
/* create skeleton of new HMM. */
hmm = hmm_new_nstates(ncats_unspooled, 0, 0);
/* Main loop: consider each MSA in turn */
for (msa_idx = 0; msa_idx < nmsas; msa_idx++) {
if (msa_fname_list != NULL) {
msa_fname = (String*)lst_get_ptr(msa_fname_list, msa_idx);
F = phast_fopen(msa_fname->chars, "r");
if (!quiet_mode)
fprintf(stderr, "Reading alignment from %s ...\n",
F == stdin ? "stdin" : msa_fname->chars);
msa = msa_new_from_file(F, NULL);
phast_fclose(F);
}
else { /* only lengths of alignments specified */
msa = msa_new(NULL, NULL, 0, lst_get_int(msa_length_list, msa_idx), NULL);
/* just a shell in this case */
}
gff_fname = (String*)lst_get_ptr(gff_fname_list, msa_idx);
if (!quiet_mode)
fprintf(stderr, "Reading annotations from %s ...\n", gff_fname->chars);
gff = gff_read_set(phast_fopen(gff_fname->chars, "r"));
/* convert GFF to coordinate frame of alignment */
if (msa_length_list == NULL) {
if (!quiet_mode)
fprintf(stderr, "Mapping annotations to alignment ...\n");
msa_map_gff_coords(msa, gff, 1, 0, 0); /* assume seq 1 is ref */
}
if (model_indels_str != NULL) {
if (!quiet_mode)
fprintf(stderr, "Obtaining gap patterns ...\n");
msa_gap_patterns = smalloc(msa->length * sizeof(int));
gp_set_phylo_patterns(gpm, msa_gap_patterns, msa);
}
/* at this point, we don't actually need the alignment anymore;
if using ordered suff stats (likely with large data sets),
can free them now, to avoid running out of memory */
if (msa->ss != NULL) {
ss_free(msa->ss);
msa->ss = NULL;
}
if (reverse_groups_tag != NULL) {
if (!quiet_mode)
fprintf(stderr, "Reverse complementing features on negative strand (group by '%s') ...\n",
reverse_groups_tag);
/* we don't need to reverse complement the whole alignment --
just the gff and possibly the gap pattern array (pass a
NULL msa) */
gff_group(gff, reverse_groups_tag);
msa_reverse_compl_feats(NULL, gff, msa_gap_patterns);
}
if (!quiet_mode)
fprintf(stderr, "Labeling sites by category ...\n");
msa_label_categories(msa, gff, cm);
gff_free_set(gff);
if (model_indels_str != NULL) {
if (!quiet_mode)
fprintf(stderr, "Remapping categories according to gap patterns ...\n");
if (indel_nseqs > 0 && indel_nseqs != msa->nseqs) {
/* in this case, we'll simply reassign non-trivial gap
patterns randomly. This will achieve the desired
effect with minimal coding, as long as the number of
sites is not too small (the indel model is probably
useless anyway if the number is small) */
int pat, newpat;
int npatterns = 4 * indel_nseqs - 5;
int complex_allowed[cm->ncats+1];
List *no_complex_names, *no_complex_nums;
if (!quiet_mode)
fprintf(stderr, "(target number of sequences: %d)\n", indel_nseqs);
/* set up index indicating by cat no. whether complex gaps
are allowed */
for (i = 0; i < ncats; i++) complex_allowed[i] = 1;
no_complex_names = lst_new_ptr(10);
str_split(str_new_charstr(NO_COMPLEX), ",", no_complex_names);
no_complex_nums = cm_get_category_list(cm, no_complex_names, 1);
for (i = 0; i < lst_size(no_complex_nums); i++)
complex_allowed[lst_get_int(no_complex_nums, i)] = 0;
lst_free(no_complex_nums);
lst_free_strings(no_complex_names);
lst_free(no_complex_names);
/* now reassign all non-null numbers */
for (i = 0; i < msa->length; ) {
if ((pat = msa_gap_patterns[i]) != 0) {
if (complex_allowed[msa->categories[i]])
newpat = 1 + ((double)npatterns * unif_rand());
/* random number in interval [1, npatterns] */
else
newpat = 1 + ((double)(npatterns-1) * unif_rand());
/* random number in interval [1,npatterns-1]
(excludes complex gap pattern) */
for (; i < msa->length && msa_gap_patterns[i] == pat; i++)
msa_gap_patterns[i] = newpat; /* change for whole sequence */
}
else i++;
}
}
/* obtain gapped category number for each site */
for (i = 0; i < msa->length; i++)
if (gpm->cat_x_pattern_to_gapcat[msa->categories[i]] != NULL)
msa->categories[i] = gpm->cat_x_pattern_to_gapcat[msa->categories[i]][msa_gap_patterns[i]];
}
if (!quiet_mode)
fprintf(stderr, "Unspooling categories ...\n");
cm_spooled_to_unspooled(cm, msa->categories, msa->length);
if (!quiet_mode)
fprintf(stderr, "Collecting training data ...\n");
hmm_train_update_counts(traincounts, statecounts, begcounts,
msa->categories, msa->length,
ncats_unspooled);
if (msa_gap_patterns != NULL) sfree(msa_gap_patterns);
msa_free(msa);
}
/* now train HMM, using cumulative data */
hmm_train_from_counts(hmm, traincounts, NULL, statecounts, NULL,
begcounts, NULL);
/* if modeling indels, adjust begin transitions so probability is
distributed among different "gap pattern" states that all
correspond to the same ungapped state (category); this helps
avoid problems that occur when training on a few large sequences
(e.g., whole chromosomes) and then testing on many shorter ones */
if (model_indels_str != NULL) {
double tprob[gpm->ncats];
int nst[gpm->ncats]; /* total prob and number of states per
spooled, ungapped category */
for (i = 0; i < gpm->ncats; i++) tprob[i] = nst[i] = 0;
for (i = 0; i < hmm->nstates; i++) {
if (vec_get(hmm->begin_transitions, i) > 0)
/* have to go from unspooled space to spooled space, then to
ungapped space (HMM states correspond to unspooled,
gapped categories). Note that states with nonzero begin
probs shouldn't be conditioned on other states. */
tprob[gpm->gapcat_to_cat[cm_unspooled_to_spooled_cat(cm, i)]] +=
vec_get(hmm->begin_transitions, i);
nst[gpm->gapcat_to_cat[cm_unspooled_to_spooled_cat(cm, i)]]++;
}
for (i = 0; i < hmm->nstates; i++)
if (tprob[gpm->gapcat_to_cat[cm_unspooled_to_spooled_cat(cm, i)]] > 0)
vec_set(hmm->begin_transitions, i,
tprob[gpm->gapcat_to_cat[cm_unspooled_to_spooled_cat(cm, i)]] /
nst[gpm->gapcat_to_cat[cm_unspooled_to_spooled_cat(cm, i)]]);
/* (uniform prior) */
}
/* write trained HMM */
hmm_print(stdout, hmm);
if (!quiet_mode) fprintf(stderr, "Done.\n");
return 0;
}
List *pwm_read(const char *filename) {
List *result;
Matrix *pwm = NULL;
int i, currBase, nBases = 0;
FILE * F;
// char *motifName;
String *line = str_new(STR_MED_LEN);
List *l = lst_new_ptr(3);
List *probabilitiesStr = lst_new_ptr(4);
List *probabilitiesDbl;
Regex *pssm_re = NULL;
Regex *motif_name_re = NULL;
int alphabetLength;
result = lst_new_ptr(1);
//letter-probability matrix: alength= 4 w= 8 nsites= 2 E= 1.5e+004
pssm_re = str_re_new("^letter-probability matrix: alength= ([0-9]+) w= ([0-9]+)");
motif_name_re = str_re_new("^MOTIF[[:space:]]+(.+?)[[:space:]].*");
//open PWM file
F = phast_fopen(filename, "r");
currBase = 0;
nBases = -1;
//For each line in the MEME file
while ((str_readline(line, F)) != EOF) {
//If line matches Motif name
if (str_re_match(line, motif_name_re, l, 1) > 0) {
// motifName = copy_charstr(((String*)lst_get_ptr(l, 1))->chars);
//printf("motifName=%s\n", motifName);
}
//If line matches beginning of a probability matrix
else if (str_re_match(line, pssm_re, l, 2) > 0) {
//Extract the alphabet size & number of bases in matrix
if (str_as_int((String*)lst_get_ptr(l, 1), &alphabetLength) != 0)
die("ERROR: Unable to parse 'alength=' from MEME file, expected integer, read %s", ((String*)lst_get_ptr(l, 1))->chars);
if (str_as_int((String*)lst_get_ptr(l, 2), &nBases) != 0)
die("ERROR: Unable to parse 'w=' from MEME file, expected integer, read %s ", ((String*)lst_get_ptr(l, 2))->chars);
currBase = 0;
if (nBases <= 0) //We must have at least one base in the PWM
die("ERROR: No Position Weight Matrices were detected in the provided PWM file");
if (alphabetLength <= 0) //We must have a positive alphabet length
die("ERROR: Alphabet lengh specified in PWM file must be greater than zero");
pwm = mat_new(nBases, alphabetLength);
mat_set_all(pwm, -1);
continue;
//If this row contains matrix data
} else if (currBase < nBases) {
//Parse row of probabilities
str_double_trim(line);
str_split(line, NULL, probabilitiesStr);
probabilitiesDbl = str_list_as_dbl(probabilitiesStr);
for (i = 0; i < lst_size(probabilitiesDbl); i++)
mat_set(pwm, currBase, i, log(lst_get_dbl(probabilitiesDbl, i)));
currBase++;
} else if ((currBase == nBases) && (pwm != NULL)) {
//Push full matrix
lst_push_ptr(result, pwm);
pwm = NULL;
}
}
if (currBase == nBases && pwm != NULL)
lst_push_ptr(result, pwm);
else if (pwm != NULL)
die("Premature end of PWM file\n");
str_re_free(motif_name_re);
str_re_free(pssm_re);
phast_fclose(F);
return result;
}
MS *ms_read(const char *filename, const char *alphabet) {
List *names = lst_new_ptr(10);
List *seqs = lst_new_ptr(10);
static Regex *descrip_re = NULL;
int i, nseqs, j, do_toupper, line_no;
String *line = str_new(STR_MED_LEN);
List *l = lst_new_ptr(2);
String *n, *s, *new_str = NULL;
MS *ms;
FILE * F;
F = phast_fopen(filename, "r");
if (descrip_re == NULL)
descrip_re = str_re_new("[[:space:]]*>[[:space:]]*(.+)");
line_no=1;
while ((str_readline(line, F)) != EOF) {
if (str_re_match(line, descrip_re, l, 1) > 0) {
lst_push_ptr(names, lst_get_ptr(l, 1));
str_free((String*)lst_get_ptr(l, 0));
new_str = str_new(STR_MED_LEN);
lst_push_ptr(seqs, new_str);
continue;
}
str_double_trim(line);
if (line->length == 0) continue;
if (new_str == NULL)
die("ERROR in FASTA file: non-blank line preceding first description ('>') line.\n");
str_append(new_str, line);
checkInterruptN(line_no++, 1000);
}
if (lst_size(seqs) == 0)
die("ERROR: empty FASTA file.\n");
// now create MS
nseqs = lst_size(names);
if (nseqs != lst_size(seqs))
die("ERROR ms_read: nseqs (%i) != lst_size(seqs) (%i)\n",
nseqs, lst_size(seqs));
ms = ms_new(NULL, NULL, nseqs, NULL, 0, 1);
ms->names = (char**)smalloc(nseqs * sizeof(char*));
ms->seqs = (char**)smalloc(nseqs * sizeof(char*));
ms->idx_offsets = (int*)smalloc(nseqs * sizeof(int));
// upcase chars unless there are lowercase characters in the alphabet
do_toupper = !ms_alph_has_lowercase(ms);
for (i = 0; i < nseqs; i++) {
n = (String*)lst_get_ptr(names, i);
ms->names[i] = (char*)smalloc((n->length + 1) * sizeof(char));
strcpy(ms->names[i], n->chars);
str_free(n);
s = (String*)lst_get_ptr(seqs, i);
ms->seqs[i] = (char*)smalloc((s->length + 1) * sizeof(char));
ms->idx_offsets[i] = 0;
// scan chars and adjust if necessary
for (j = 0; j < s->length; j++) {
ms->seqs[i][j] = do_toupper ? (char)toupper(s->chars[j]) : s->chars[j];
if (ms->seqs[i][j] == '.' && ms->inv_alphabet[(int)'.'] == -1)
ms->seqs[i][j] = ms->missing[0]; // interpret '.' as missing
// data; maybe no longer
// necessary
if (isalpha(ms->seqs[i][j]) && ms->inv_alphabet[(int)ms->seqs[i][j]] == -1 && get_iupac_map()[(int)ms->seqs[i][j]] == NULL)
ms->seqs[i][j] = 'N'; // assume 'N' if unrecognized letter
}
ms->seqs[i][s->length] = '\0';
str_free(s);
// str_free(n);
}
lst_free(names);
lst_free(seqs);
lst_free(l);
str_free(line);
phast_fclose(F);
return ms;
}
int main(int argc, char *argv[]) {
List *pruned_names = lst_new_ptr(5);
TreeModel *source_mod;
MSA *msa = NULL, *out_msa;
IndelHistory *ih;
char *read_hist_fname = NULL;
char c;
int opt_idx, old_nnodes, i;
msa_format_type msa_format = UNKNOWN_FORMAT;
int output_alignment = FALSE, ia_names = FALSE;
struct option long_opts[] = {
{"msa-format", 1, 0, 'i'},
{"output-alignment", 0, 0, 'A'},
{"read-history", 1, 0, 'H'},
{"ia-names", 0, 0, 'I'},
{"help", 0, 0, 'h'},
{0, 0, 0, 0}
};
while ((c = getopt_long(argc, argv, "i:H:AIh", long_opts, &opt_idx)) != -1) {
switch (c) {
case 'i':
msa_format = msa_str_to_format(optarg);
if (msa_format == -1)
die("ERROR: unrecognized alignment format.\n");
break;
case 'A':
output_alignment = TRUE;
break;
case 'H':
read_hist_fname = optarg;
break;
case 'I':
ia_names = TRUE;
break;
case 'h':
printf("%s", HELP);
exit(0);
case '?':
die("Bad argument. Try 'indelHistory -h'.\n");
}
}
set_seed(-1);
if (read_hist_fname != NULL) {
fprintf(stderr, "Reading indel history from %s...\n", read_hist_fname);
ih = ih_new_from_file(phast_fopen(read_hist_fname, "r"));
}
else {
FILE *mfile;
if (optind != argc - 2)
die("Two arguments required. Try 'indelHistory -h'.\n");
fprintf(stderr, "Reading alignment from %s...\n", argv[optind]);
mfile = phast_fopen(argv[optind], "r");
if (msa_format == UNKNOWN_FORMAT)
msa_format = msa_format_for_content(mfile, 1);
msa = msa_new_from_file_define_format(mfile, msa_format, "ACGTNB^.-");
phast_fclose(mfile);
if (msa->seqs == NULL && (msa->ss == NULL || msa->ss->tuple_idx == NULL))
die("ERROR: ordered representation of alignment required.\n");
fprintf(stderr, "Reading tree from %s...\n", argv[optind+1]);
source_mod = tm_new_from_file(phast_fopen(argv[optind+1], "r"), 1);
/* prune tree, if necessary */
old_nnodes = source_mod->tree->nnodes;
tm_prune(source_mod, msa, pruned_names);
if (lst_size(pruned_names) == (old_nnodes + 1) / 2)
die("ERROR: no match for leaves of tree in alignment (leaf names must match alignment names).\n");
if (lst_size(pruned_names) > 0) {
fprintf(stderr, "WARNING: pruned away leaves of tree with no match in alignment (");
for (i = 0; i < lst_size(pruned_names); i++)
fprintf(stderr, "%s%s", ((String*)lst_get_ptr(pruned_names, i))->chars,
i < lst_size(pruned_names) - 1 ? ", " : ").\n");
}
lst_free(pruned_names);
tr_name_ancestors(source_mod->tree);
if (msa->nseqs > (source_mod->tree->nnodes + 1) / 2) { /* assume ancestral
seqs specified
in this case */
if (ia_names) {
fprintf(stderr, "Converting sequence names...\n");
ih_convert_ia_names(msa, source_mod->tree);
}
fprintf(stderr, "Extracting indel history from alignment...\n");
ih = ih_extract_from_alignment(msa, source_mod->tree);
}
else { /* infer by parsimony */
if (msa->ss == NULL) {
fprintf(stderr, "Extracting sufficient statistics...\n");
ss_from_msas(msa, 1, TRUE, NULL, NULL, NULL, -1, 0);
}
fprintf(stderr, "Inferring indel history by parsimony...\n");
ih = ih_reconstruct(msa, source_mod->tree);
}
}
if (output_alignment) {
out_msa = ih_as_alignment(ih, msa);
msa_print(stdout, out_msa, FASTA, FALSE);
}
else
ih_print(ih, stdout,
read_hist_fname != NULL ? read_hist_fname : argv[optind],
"indelHistory");
fprintf(stderr, "Done.\n");
return 0;
}
int main(int argc, char* argv[]) {
FILE* F;
MSA *msa;
msa_format_type format = UNKNOWN_FORMAT;
int src_ref = -1, dest_ref = 0, offset = 0;
char *msa_fname = NULL, *feat_fname = NULL;
GFF_Set *gff;
char c;
while ((c = (char)getopt(argc, argv, "hm:f:s:d:i:p:n:")) != -1) {
switch(c) {
case 'm':
msa_fname = optarg;
break;
case 'f':
feat_fname = optarg;
break;
case 's':
src_ref = get_arg_int(optarg);
break;
case 'd':
dest_ref = get_arg_int(optarg);
break;
case 'i':
format = msa_str_to_format(optarg);
if (format == UNKNOWN_FORMAT) die("ERROR: bad alignment format.\n");
break;
case 'p':
offset = get_arg_int(optarg);
break;
case 'n':
offset = -1 * get_arg_int(optarg);
break;
case 'h':
print_usage();
exit(1);
case '?':
print_usage();
exit(1);
}
}
if (msa_fname == NULL || feat_fname == NULL) {
print_usage();
exit(1);
}
set_seed(-1);
F = phast_fopen(feat_fname, "r");
if ((gff = gff_read_set(F)) == NULL) {
die("ERROR: error reading %s.\n", feat_fname);
}
phast_fclose(F);
/* handle case of local alignment specially -- avoid representing
the alignment explicitly */
F = phast_fopen(msa_fname, "r");
if (format == UNKNOWN_FORMAT)
format = msa_format_for_content(F, 1);
if (format == LAV) {
LocalPwAlignment *lpwa = NULL;
/* int i; */
fprintf(stderr, "WARNING: in local alignment mode, coordinates may only be mapped from query (reference) sequence to target (aligned) sequence.\n");
lpwa = la_read_lav(F, 0);
la_gff_transform(lpwa, gff);
/* for (i = 0; i < lst_size(gff->features); i++) { */
/* GFF_Feature *feat = lst_get_ptr(gff->features, i); */
/* feat->start = la_get_target_coord(lpwa, feat->start); */
/* feat->end = la_get_target_coord(lpwa, feat->end); */
/* } */
}
else { /* normal alignment */
msa = msa_new_from_file_define_format(F, format, NULL);
phast_fclose(F);
msa_map_gff_coords(msa, gff, src_ref, dest_ref, offset);
msa_free(msa);
}
gff_print_set(stdout, gff);
gff_free_set(gff);
return 0;
}
int main(int argc, char* argv[]) {
char *maf_fname = NULL, *out_root_fname = "maf_parse", *masked_fn = NULL;
String *refseq = NULL, *currRefseq;
int opt_idx, startcol = 1, endcol = -1, include = 1, splitInterval = -1;
char c, outfilename[1000], splitFormat[100]="%s%.1i.maf", *group_tag = NULL;
List *order_list = NULL, *seqlist_str = NULL, *cats_to_do_str=NULL, *cats_to_do=NULL;
MafBlock *block;
FILE *mfile, *outfile=NULL, *masked_file=NULL;
int useRefseq=TRUE, currLen=-1, blockIdx=0, currSize, sortWarned=0;
int lastIdx = 0, currStart=0, by_category = FALSE, i, pretty_print = FALSE;
int lastStart = -1, gffSearchIdx=0;
GFF_Set *gff = NULL, *gffSub;
GFF_Feature *feat;
CategoryMap *cm = NULL;
int base_mask_cutoff = -1, stripILines=FALSE, stripELines=FALSE;//, numspec=0;
List *outfileList=NULL;
Hashtable *outfileHash=NULL;//, *specNameHash=NULL;
msa_format_type output_format = MAF;
MSA *msa = NULL;//, **catMsa;
char *mask_features_spec_arg=NULL;
List *mask_features_spec=NULL;
struct option long_opts[] = {
{"start", 1, 0, 's'},
{"end", 1, 0, 'e'},
{"seqs", 1, 0, 'l'},
{"exclude", 0, 0, 'x'},
{"order", 1, 0, 'O'},
{"split", 1, 0, 'S'},
{"out-root", 1, 0, 'r'},
{"out-root-digits", 1, 0, 'd'},
{"no-refseq", 0, 0, 'n'},
{"features", 1, 0, 'g'},
{"by-category", 0, 0, 'L'},
{"do-cats", 1, 0, 'C'},
{"catmap", 1, 0, 'c'},
{"by-group", 1, 0, 'P'},
{"mask-bases", 1, 0, 'b'},
{"masked-file", 1, 0, 'm'},
{"strip-i-lines", 0, 0, 'I'},
{"strip-e-lines", 0, 0, 'E'},
{"mask-features", 1, 0, 'M'},
{"help", 0, 0, 'h'},
{0, 0, 0, 0}
};
while ((c = getopt_long(argc, argv, "s:e:l:O:r:S:d:g:c:P:b:o:m:M:pLnxEIh", long_opts, &opt_idx)) != -1) {
switch(c) {
case 's':
startcol = get_arg_int(optarg);
break;
case 'e':
endcol = get_arg_int(optarg);
break;
case 'l':
seqlist_str = get_arg_list(optarg);
break;
case 'O':
order_list = get_arg_list(optarg);
break;
case 'x':
include = FALSE;
break;
case 'S':
splitInterval = atoi(optarg);
break;
case 'r':
out_root_fname = optarg;
break;
case 'd':
sprintf(splitFormat, "%%s%%.%si.%%s", optarg);
break;
case 'n':
useRefseq = FALSE;
break;
case 'g':
gff = gff_read_set(phast_fopen(optarg, "r"));
gff_sort(gff);
stripILines=TRUE;
stripELines=TRUE;
break;
case 'c':
cm = cm_new_string_or_file(optarg);
break;
case 'C':
cats_to_do_str = get_arg_list(optarg);
break;
case 'L':
by_category = TRUE;
break;
case 'P':
group_tag = optarg;
break;
case 'b':
base_mask_cutoff = atoi(optarg);
break;
case 'm':
masked_fn = optarg;
break;
case 'M':
mask_features_spec_arg = optarg;
break;
case 'E':
stripELines=TRUE;
break;
case 'I':
stripILines=TRUE;
break;
case 'o':
output_format = msa_str_to_format(optarg);
if (output_format == UNKNOWN_FORMAT)
die("ERROR: bad output format. Try \"maf_parse -h\" for help.\n");
if (output_format != MAF)
die("Sorry, only MAF format output has been implemented right now.\n");
break;
case 'p':
pretty_print = TRUE;
break;
case 'h':
print_usage();
exit(0);
case '?':
die("Bad argument. Try 'maf_parse -h' for help.\n");
}
}
if (optind >= argc)
die("Missing alignment filename. Try 'maf_parse -h' for help.\n");
else if (optind == argc - 1)
maf_fname = argv[optind];
else
die("ERROR: Too many arguments. Try 'maf_parse -h' for help.\n");
set_seed(-1);
if (startcol < 1 || (endcol != -1 && endcol < startcol))
die("ERROR: must have 1 <= start <= end <= [msa_length]\n");
if ((group_tag != NULL || by_category) && gff == NULL)
die("ERROR: --by-category and --by-group require --features. Try \"maf_parse -h\""
" for help.\n");
if (group_tag != NULL && by_category)
die("ERROR: --by-category and --by-group cannot be used together. Try \"maf_parse -h\""
" for help.\n");
if (splitInterval != -1 && gff != NULL)
die("ERROR: can't use --split and --features together. Try \"maf_parse -h\""
"for help\n");
if (group_tag != NULL || by_category) {
outfileList = lst_new_ptr(10);
outfileHash = hsh_new(100);
}
if (gff != NULL && cm == NULL)
cm = cm_new_from_features(gff);
if (cats_to_do_str != NULL) {
cats_to_do = cm_get_category_str_list(cm, cats_to_do_str, FALSE);
if (gff != NULL)
gff_filter_by_type(gff, cats_to_do, 0, NULL);
}
if (masked_fn != NULL) {
if (base_mask_cutoff == -1)
die("ERROR: need to use --mask-bases with --masked-file");
masked_file = phast_fopen(masked_fn, "w");
}
if (mask_features_spec_arg != NULL) {
if (gff==NULL)
die("ERROR: need --features with --mask-features");
mask_features_spec = lst_new_ptr(10);
str_split(str_new_charstr(mask_features_spec_arg), ",", mask_features_spec);
for (i=0; i < lst_size(mask_features_spec); i++) {
fprintf(stderr, "masking species %s within features\n",
((String*)lst_get_ptr(mask_features_spec, i))->chars);
}
}
/* Check to see if --do-cats names a feature which is length 1.
If so, set output_format to SS ? or FASTA ? */
mfile = phast_fopen(maf_fname, "r");
block = mafBlock_read_next(mfile, NULL, NULL);
if (splitInterval == -1 && gff==NULL) {
//TODO: do we want to copy header from original MAF in this case?
mafBlock_open_outfile(NULL, argc, argv);
}
while (block != NULL) {
if (order_list != NULL)
mafBlock_reorder(block, order_list);
if (seqlist_str != NULL)
mafBlock_subSpec(block, seqlist_str, include);
if (mafBlock_numSpec(block)==0 || mafBlock_all_gaps(block))
goto get_next_block;
if (stripILines)
mafBlock_strip_iLines(block);
if (stripELines)
mafBlock_strip_eLines(block);
if (base_mask_cutoff != -1)
mafBlock_mask_bases(block, base_mask_cutoff, masked_file);
//TODO: still need to implement (either here or elsewhere)
// if (indel_mask_cutoff != -1)
// mafBlock_mask_indels(block, indel_mask_cutoff, mfile);
if (useRefseq) { //get refseq and check that it is consistent in MAF file
currRefseq = mafBlock_get_refSpec(block);
if (refseq == NULL)
refseq = str_new_charstr(currRefseq->chars);
else if (str_compare(refseq, currRefseq)!=0)
die("Error: refseq not consistent in MAF (got %s, %s)\n",
refseq->chars, currRefseq->chars);
}
if (startcol != 1 || endcol != -1)
if (0 == mafBlock_trim(block, startcol, endcol, refseq, useRefseq ? 0 : lastIdx))
goto get_next_block;
currSize = mafBlock_get_size(block, refseq);
if (useRefseq) {
currStart = mafBlock_get_start(block, refseq);
if (currStart < lastIdx && sortWarned == 0) {
fprintf(stderr, "Warning: input MAF not sorted with respect to refseq. Output files may not represent contiguous alignments. (%i, %i)\n", lastIdx, currStart);
sortWarned = 1;
}
}
else currStart = lastIdx;
if (currStart < lastStart) gffSearchIdx = 0;
lastStart = currStart;
lastIdx = currStart + currSize;
//split by length
if (splitInterval != -1) {
if (currLen == -1 || currLen+currSize > splitInterval) {
sprintf(outfilename, splitFormat, out_root_fname, ++blockIdx,
msa_suffix_for_format(output_format));
if (output_format == MAF) {
if (outfile != NULL) mafBlock_close_outfile(outfile);
outfile = mafBlock_open_outfile(outfilename, argc, argv);
}
else if (output_format != MAF && msa != NULL) {
// msa_print_to_filename(msa, outfilename, output_format, pretty_print);
msa_free(msa);
msa = NULL;
}
currLen = 0;
}
currLen += currSize;
}
else outfile = stdout;
if (gff != NULL && mask_features_spec != NULL) {
gffSub = gff_subset_range_overlap_sorted(gff, currStart+1, lastIdx,
&gffSearchIdx);
if (gffSub != NULL) {
mafBlock_mask_region(block, gffSub, mask_features_spec);
gff_free_set(gffSub);
}
mafBlock_print(outfile, block, pretty_print);
} else if (gff != NULL) {
gffSub = gff_subset_range_overlap_sorted(gff, currStart+1, lastIdx,
&gffSearchIdx);
if (gffSub != NULL) {
if (by_category) gff_group_by_feature(gffSub);
else if (group_tag != NULL) gff_group(gffSub, group_tag);
gff_sort(gffSub);
gff_flatten_within_groups(gffSub);
for (i=0; i<lst_size(gffSub->features); i++) {
feat = (GFF_Feature*)lst_get_ptr(gffSub->features, i);
MafBlock *subBlock = mafBlock_copy(block);
mafBlock_trim(subBlock, feat->start, feat->end, refseq, 0);
if (by_category)
outfile = get_outfile(outfileList, outfileHash, feat->feature, out_root_fname,
argc, argv);
else if (group_tag != NULL)
outfile = get_outfile(outfileList, outfileHash,
gff_group_name(gffSub, feat), out_root_fname,
argc, argv);
else outfile = stdout;
if (output_format == MAF)
mafBlock_print(outfile, subBlock, pretty_print);
// else msa_add_mafBlock(msa);
mafBlock_free(subBlock);
}
gff_free_set(gffSub);
}
}
else {
if (output_format == MAF)
mafBlock_print(outfile, block, pretty_print);
// else msa = msa_add_mafBlock(mafBlock, msa, );
}
get_next_block:
mafBlock_free(block);
block = mafBlock_read_next(mfile, NULL, NULL);
}
if (masked_file != NULL) fclose(masked_file);
if (output_format == MAF) {
if (by_category || group_tag != NULL)
close_outfiles(outfileList, outfileHash);
else if (outfile!=NULL) mafBlock_close_outfile(outfile);
} else {
msa_print(stdout, msa, output_format, pretty_print);
msa_free(msa);
}
if (gff != NULL) gff_free_set(gff);
phast_fclose(mfile);
return 0;
}
int main(int argc, char *argv[]) {
char c;
int i, j, t, opt_idx, ntrees, nleaves = -1;
TreeNode *n, *node_i, *node_j, *lca, *nametree = NULL;
TreeNode **tree;
List *leaves, ***distance, *tree_fnames, *tot_dist;
int mod = FALSE;
char **leaf_name;
String *trees_arg;
FILE *F;
struct option long_opts[] = {
{"mod", 0, 0, 'm'},
{"tree", 1, 0, 't'},
{"help", 0, 0, 'h'},
{0, 0, 0, 0}
};
while ((c = getopt_long(argc, argv, "mt:h", long_opts, &opt_idx)) != -1) {
switch (c) {
case 'm':
mod = TRUE;
break;
case 't':
if (optarg[0] == '(')
nametree = tr_new_from_string(optarg);
else
nametree = tr_new_from_file(phast_fopen(optarg, "r"));
break;
case 'h':
usage(argv[0]);
case '?':
die("Bad argument. Try '%s -h'.\n", argv[0]);
}
}
if (optind > argc - 1)
die("Input filename required. Try '%s -h'.\n", argv[0]);
set_seed(-1);
/* build a comma-delimited list and pass to get_arg_list; allows
possibility of reading from file via '*' operator */
trees_arg = str_new(1000);
for (i = optind; i < argc; i++) {
str_append_charstr(trees_arg, argv[i]);
if (i < argc - 1) str_append_char(trees_arg, ',');
}
tree_fnames = get_arg_list(trees_arg->chars);
ntrees = lst_size(tree_fnames);
tree = smalloc(ntrees * sizeof(void*));
/* read trees */
for (t = 0; t < ntrees; t++) {
String *fname = lst_get_ptr(tree_fnames, t);
if (mod) {
TreeModel *m = tm_new_from_file(F = phast_fopen(fname->chars, "r"), 1);
tree[t] = tr_create_copy(m->tree);
tm_free(m);
phast_fclose(F);
}
else
tree[t] = tr_new_from_file(phast_fopen(fname->chars, "r"));
}
/* initialization */
nleaves = (tree[0]->nnodes + 1)/2;
leaves = lst_new_ptr(nleaves);
distance = smalloc(nleaves * sizeof(void*));
leaf_name = smalloc(nleaves * sizeof(void*));
for (i = 0; i < nleaves; i++) {
distance[i] = smalloc(nleaves * sizeof(void*));
for (j = i+1; j < nleaves; j++)
distance[i][j] = lst_new_dbl(ntrees);
}
if (nametree == NULL) nametree = tree[0];
for (i = 0, j = 0; i < lst_size(nametree->nodes); i++) {
n = lst_get_ptr(nametree->nodes, i);
if (n->lchild == NULL && n->rchild == NULL)
leaf_name[j++] = n->name;
}
tot_dist = lst_new_dbl(ntrees);
/* now compute distances */
for (t = 0; t < ntrees; t++) {
/* obtain list of leaves */
lst_clear(leaves);
for (i = 0; i < lst_size(tree[t]->nodes); i++) {
n = lst_get_ptr(tree[t]->nodes, i);
if (n->lchild == NULL && n->rchild == NULL)
lst_push_ptr(leaves, n);
}
if (lst_size(leaves) != nleaves)
die("ERROR: trees have different numbers of leaves.\n");
/* look at all pairs */
for (i = 0; i < nleaves; i++) {
node_i = lst_get_ptr(leaves, i);
for (j = i+1; j < nleaves; j++) {
double dist = 0;
node_j = lst_get_ptr(leaves, j);
/* because ids are assigned in pre-order, the first ancestor of
node j that has an id less than i is the LCA of i and j; we
seek the sum of distances from both i and j to this node */
for (n = node_j; n->id >= node_i->id; n = n->parent)
dist += n->dparent;
lca = n;
for (n = node_i; n != lca; n = n->parent)
dist += n->dparent;
lst_push_dbl(distance[i][j], dist);
}
}
lst_push_dbl(tot_dist, tr_total_len(tree[t]));
}
/* print distances and (optionally) stats */
if (ntrees == 1) {
for (i = 0; i < nleaves; i++) {
for (j = i+1; j < nleaves; j++) {
printf ("%s\t%s\t%f\n", leaf_name[i], leaf_name[j],
lst_get_dbl(distance[i][j], 0));
}
}
printf ("%s\t%s\t%f\n", "(total)", "-", lst_get_dbl(tot_dist, 0));
}
else {
double mean, stdev;
double quantiles[] = {0, 0.025, 0.05, 0.5, 0.95, 0.975, 1};
double quantile_vals[7];
printf("%-15s %-15s %9s %9s %9s %9s %9s %9s %9s %9s %9s\n", "leaf1",
"leaf2", "mean", "stdev", "median", "min", "max", "95%_min",
"95%_max", "90%_min", "90%_max");
for (i = 0; i < nleaves; i++) {
for (j = i+1; j < nleaves; j++) {
mean = lst_dbl_mean(distance[i][j]);
stdev = lst_dbl_stdev(distance[i][j]);
lst_qsort_dbl(distance[i][j], ASCENDING);
lst_dbl_quantiles(distance[i][j], quantiles, 7, quantile_vals);
printf("%-15s %-15s %9.5f %9.5f %9.5f %9.5f %9.5f %9.5f %9.5f %9.5f %9.5f\n",
leaf_name[i], leaf_name[j], mean, stdev, quantile_vals[3], quantile_vals[0],
quantile_vals[6], quantile_vals[1], quantile_vals[5], quantile_vals[2],
quantile_vals[4]);
}
}
/* also do total branch len */
mean = lst_dbl_mean(tot_dist);
stdev = lst_dbl_stdev(tot_dist);
lst_qsort_dbl(tot_dist, ASCENDING);
lst_dbl_quantiles(tot_dist, quantiles, 7, quantile_vals);
printf("%-15s %-15s %9.5f %9.5f %9.5f %9.5f %9.5f %9.5f %9.5f %9.5f %9.5f\n",
"(total)", "-", mean, stdev, quantile_vals[3], quantile_vals[0],
quantile_vals[6], quantile_vals[1], quantile_vals[5], quantile_vals[2],
quantile_vals[4]);
}
return 0;
}
int main(int argc, char *argv[]) {
TreeNode *tree = NULL;
TreeModel *backgd_mod = NULL;
int i, j,
size = DEFAULT_SIZE, meme_mode = 0, profile_mode = 0,
nrestarts = 10, npseudocounts = 5, nsamples = -1,
nmostprevalent = -1, tuple_size = -1, nbest = -1, sample_parms = 0,
nmotifs = DEFAULT_NUMBER, nseqs = -1, do_html = 0, do_bed = 0,
suppress_stdout = 0;
List *msa_name_list = NULL, *pos_examples = NULL, *init_list = NULL, *tmpl;
List *msas, *motifs;
SeqSet *seqset = NULL;
PooledMSA *pmsa = NULL;
msa_format_type msa_format = UNKNOWN_FORMAT;
Vector *backgd_mnmod = NULL;
Hashtable *hash=NULL;
String *output_prefix = str_new_charstr("phastm.");
double *has_motif = NULL;
double prior = PRIOR;
char c;
GFF_Set *bedfeats = NULL;
while ((c = getopt(argc, argv, "t:i:b:sk:md:pn:I:R:P:w:c:SB:o:HDxh")) != -1) {
switch (c) {
case 't':
tree = tr_new_from_file(phast_fopen(optarg, "r"));
break;
case 'i':
msa_format = msa_str_to_format(optarg);
if (msa_format == UNKNOWN_FORMAT)
die("ERROR: bad input format.\n");
break;
case 'b':
backgd_mod = tm_new_from_file(phast_fopen(optarg, "r"), 1);
break;
case 's':
break;
case 'k':
size = get_arg_int(optarg);
break;
case 'm':
meme_mode = 1;
break;
case 'd':
pos_examples = get_arg_list(optarg);
break;
case 'p':
profile_mode = 1;
break;
case 'n':
nrestarts = get_arg_int(optarg);
break;
case 'I':
init_list = get_arg_list(optarg);
break;
case 'P':
tmpl = str_list_as_int(get_arg_list(optarg));
if (lst_size(tmpl) != 2) die("ERROR: bad argument to -P.\n");
nmostprevalent = lst_get_int(tmpl, 0);
tuple_size = lst_get_int(tmpl, 1);
if (!(nmostprevalent > 0 && tuple_size > 0))
die("ERROR: bad argument nmostprevalent=%i tuple_size=%i\n",
nmostprevalent, tuple_size);
lst_free(tmpl);
break;
case 'R':
tmpl = str_list_as_int(get_arg_list(optarg));
if (lst_size(tmpl) != 2) die("ERROR: bad argument to -R.\n");
nsamples = lst_get_int(tmpl, 0);
tuple_size = lst_get_int(tmpl, 1);
if (!(nsamples > 0 && tuple_size > 0))
die("ERROR nsamples=%i tuple_sizse=%i\n", nsamples, tuple_size);
lst_free(tmpl);
break;
case 'c':
npseudocounts = get_arg_int(optarg);
break;
case 'w':
nbest = get_arg_int(optarg);
break;
case 'S':
sample_parms = 1;
break;
case 'B':
nmotifs = get_arg_int(optarg);
break;
case 'o':
str_free(output_prefix);
output_prefix = str_new_charstr(optarg);
str_append_char(output_prefix, '.');
break;
case 'H':
do_html = 1;
break;
case 'D':
do_bed = 1;
break;
case 'x':
suppress_stdout = 1;
break;
case 'h':
usage(argv[0]);
case '?':
die("Bad argument. Try '%s -h'.\n", argv[0]);
}
}
if (optind != argc - 1)
die("ERROR: List of alignment files required. Try '%s -h'.\n", argv[0]);
if ((nsamples > 0 && nmostprevalent > 0) ||
(nsamples > 0 && init_list != NULL) ||
(nmostprevalent > 0 && init_list != NULL))
die("ERROR: -I, -P, and -R are mutually exclusive.");
set_seed(-1);
msa_name_list = get_arg_list(argv[optind]);
if (backgd_mod != NULL && tree == NULL) tree = backgd_mod->tree;
if (tree == NULL && !meme_mode && !profile_mode)
die("ERROR: Must specify -t, -m, or -p.\n");
if ((init_list != NULL || nsamples > 0 || nmostprevalent > 0) &&
!sample_parms)
nrestarts = 1;
if (pos_examples != NULL) {
hash = hsh_new(lst_size(pos_examples));
for (i = 0; i < lst_size(pos_examples); i++)
hsh_put_int(hash, ((String*)lst_get_ptr(pos_examples, i))->chars, 1);
has_motif = smalloc(lst_size(msa_name_list) * sizeof(double));
}
/* open all MSAs */
msas = lst_new_ptr(lst_size(msa_name_list));
fprintf(stderr, "Reading alignment(s) ...\n");
for (i = 0, j = 0; i < lst_size(msa_name_list); i++) {
String *name = lst_get_ptr(msa_name_list, i);
FILE *mfile = phast_fopen(name->chars, "r");
msa_format_type temp_format;
MSA *msa;
if (msa_format == UNKNOWN_FORMAT)
temp_format = msa_format_for_content(mfile, 1);
else temp_format = msa_format;
msa = msa_new_from_file_define_format(mfile, temp_format, NULL);
phast_fclose(mfile);
if (nseqs == -1) nseqs = msa->nseqs;
if (!meme_mode &&
(msa->length - msa_num_gapped_cols(msa, STRIP_ANY_GAPS, -1, -1) < 300 ||
msa->nseqs != nseqs)) {
fprintf(stderr, "WARNING: ignoring alignment '%s' -- too few informative sites.\n", name->chars);
msa_free(msa);
continue;
}
if (msa_alph_has_lowercase(msa)) msa_toupper(msa);
msa_remove_N_from_alph(msa); /* Ns can be a problem */
lst_push_ptr(msas, msa);
if (has_motif != NULL) {
int k, hm = (hsh_get_int(hash, name->chars) == 1);
if (meme_mode) { /* here need to record at individ seq level */
has_motif = srealloc(has_motif,
(j + msa->nseqs + 1) * sizeof(double)); /* FIXME */
for (k = 0; k < msa->nseqs; k++) has_motif[j++] = hm;
}
else has_motif[j++] = hm;
}
}
if (!meme_mode) {
fprintf(stderr, "Extracting and pooling sufficient statistics ...\n");
pmsa = ss_pooled_from_msas(msas, 1, size, NULL, 0);
msa_remove_N_from_alph(pmsa->pooled_msa);
}
/* obtain individual sequences, if necessary */
if (nmostprevalent > 0 || nsamples > 0 || meme_mode) {
if (meme_mode) fprintf(stderr, "Converting to individual sequences ...\n");
else fprintf(stderr, "Obtaining reference sequences for pre-processing ...\n");
seqset = mtf_get_seqset(msas, meme_mode ? -1 : 1, 10 * size);
/* for now, assume 1st seq is reference */
msa_remove_N_from_alph(seqset->set);
}
if (nmostprevalent > 0) {
fprintf(stderr, "Obtaining %d most prevalent %d-tuples ...\n",
nmostprevalent, tuple_size);
init_list = lst_new_ptr(nmostprevalent);
mtf_get_common_ntuples(seqset, init_list, tuple_size, nmostprevalent);
}
else if (nsamples > 0) {
fprintf(stderr, "Sampling %d %d-tuples ...\n", nsamples, tuple_size);
init_list = lst_new_ptr(nsamples);
mtf_sample_ntuples(seqset, init_list, tuple_size, nsamples);
}
/* in meme_mode, backgd model can be specified as eq freqs in a .mod file */
if (meme_mode && backgd_mod != NULL && has_motif == NULL)
backgd_mnmod = backgd_mod->backgd_freqs;
/* estimate background model, if necessary */
else if (backgd_mod == NULL && (!meme_mode || has_motif == NULL)) {
fprintf(stderr, "Fitting background model%s ...\n",
has_motif == NULL ? "" : " (for use in initialization)");
/* if discriminative, be clear
backgd isn't really part of the
estimation procedure */
if (meme_mode) {
backgd_mnmod = vec_new(strlen(seqset->set->alphabet));
mtf_estim_backgd_mn(seqset, backgd_mnmod);
}
else {
backgd_mod = tm_new(tr_create_copy(tree), NULL, NULL, F81,
pmsa->pooled_msa->alphabet, 1, 0, NULL, -1);
tm_fit(backgd_mod, pmsa->pooled_msa,
tm_params_init(backgd_mod, .1, 5, 0),
-1, OPT_MED_PREC, NULL, 0, NULL);
}
}
/* select subset of init strings, if necessary */
if (nbest > 0 && init_list != NULL) {
fprintf(stderr, "Winnowing candidate start strings ...\n");
tmpl = lst_new_ptr(nbest);
mtf_winnow_starts(meme_mode ? (void*)seqset : (void*)pmsa,
init_list, nbest, tmpl, !meme_mode, size, tree,
meme_mode ? (void*)backgd_mnmod : (void*)backgd_mod,
has_motif);
lst_free(init_list);
init_list = tmpl;
}
/* Now find motifs */
motifs = mtf_find(meme_mode ? (void*)seqset : (void*)pmsa,
!meme_mode, size, nmotifs, tree,
meme_mode ? (void*)backgd_mnmod : (void*)backgd_mod,
has_motif, prior, nrestarts, init_list, sample_parms,
npseudocounts);
fprintf(stderr, "\n\n");
if (do_bed)
bedfeats = gff_new_set_init("phast_motif", "0.1b");
/* generate output */
for (i = 0; i < lst_size(motifs); i++) {
Motif *m = lst_get_ptr(motifs, i);
if (!suppress_stdout) {
if (lst_size(motifs) > 1)
printf("\n**********\nMOTIF #%d\n**********\n\n", i+1);
mtf_print(stdout, m);
}
if (do_html) {
String *fname = str_dup(output_prefix);
str_append_int(fname, i+1);
str_append_charstr(fname, ".html");
mtf_print_html(phast_fopen(fname->chars, "w+"), m);
str_free(fname);
}
if (do_bed)
mtf_add_features(m, bedfeats);
}
if (do_html) {
String *fname = str_dup(output_prefix);
str_append_charstr(fname, "index.html");
mtf_print_summary_html(phast_fopen(fname->chars, "w+"),
motifs, output_prefix);
str_free(fname);
}
if (do_bed) {
String *fname = str_dup(output_prefix);
str_append_charstr(fname, "bed");
gff_print_bed(phast_fopen(fname->chars, "w+"),
bedfeats, FALSE);
str_free(fname);
}
return 0;
}
