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;
}
int main(int argc, char *argv[]) {
struct phastCons_struct *p = phastCons_struct_new(0);
struct option long_opts[] = {
{"states", 1, 0, 'S'},
{"hmm", 1, 0, 'H'},
{"viterbi", 1, 0, 'V'},
{"most-conserved", 1, 0, 'V'}, /* same as --viterbi */
{"no-post-probs", 0, 0, 'n'},
{"msa-format", 1, 0, 'i'},
{"FC", 0, 0, 'X'},
{"lambda", 1, 0, 'l'},
{"target-coverage", 1, 0, 'C'},
{"transitions", 1, 0, 't'},
{"expected-length", 1, 0, 'E'},
{"expected-lengths", 1, 0, 'E'}, /* for backward compatibility */
{"estimate-trees", 1, 0, 'T'},
{"estimate-rho", 1, 0, 'O'},
{"rho", 1, 0, 'R'},
{"gc", 1, 0, 'G'},
{"ignore-missing", 0, 0, 'z'},
{"nrates", 1, 0, 'k'},
{"log", 1, 0, 'g'},
{"refidx", 1, 0, 'r'},
{"suppress-missing", 0, 0, 'x'}, /* for backward compatibility */
{"reflect-strand", 1, 0, 'U'},
{"catmap", 1, 0, 'c'},
{"extrapolate", 1, 0, 'e'},
{"indels", 0, 0, 'I'},
{"max-micro-indel", 1, 0, 'Y'},
{"indel-params", 1, 0, 'D'},
{"min-informative-types", 1, 0, 'M'}, /* for backward compatibility */
{"require-informative", 1, 0, 'M'},
{"not-informative", 1, 0, 'F'},
{"lnl", 1, 0, 'L'},
{"seqname", 1, 0, 'N'},
{"idpref", 1, 0, 'P'},
{"score", 0, 0, 's'},
{"coding-potential", 0, 0, 'p'},
{"indels-only", 0, 0, 'J'},
{"alias", 1, 0, 'A'},
{"quiet", 0, 0, 'q'},
{"help", 0, 0, 'h'},
{0, 0, 0, 0}
};
/* other vars */
FILE *infile;
char *msa_fname;
char c;
int opt_idx, i, coding_potential=FALSE;
List *tmpl = NULL;
String *tmpstr;
char *mods_fname = NULL;
List *mod_fname_list;
msa_format_type msa_format = UNKNOWN_FORMAT;
while ((c = getopt_long(argc, argv,
"S:H:V:ni:k:l:C:G:zt:E:R:T:O:r:xL:sN:P:g:U:c:e:IY:D:JM:F:pA:Xqh",
long_opts, &opt_idx)) != -1) {
switch (c) {
case 'S':
p->states = get_arg_list(optarg);
break;
case 'H':
p->hmm = hmm_new_from_file(phast_fopen(optarg, "r"));
p->two_state = FALSE;
break;
case 'V':
p->viterbi_f = phast_fopen(optarg, "w+");
tmpstr = str_new_charstr(optarg);
if (str_ends_with_charstr(tmpstr, ".gff"))
p->gff = TRUE;
str_free(tmpstr);
break;
case 'n':
p->post_probs = FALSE;
break;
case 'i':
msa_format = msa_str_to_format(optarg);
if (msa_format == UNKNOWN_FORMAT)
die("ERROR: bad argument to --msa-format\n");
break;
case 'X':
p->FC = TRUE;
p->two_state = FALSE;
break;
case 'l':
if (optarg[0] != '~')
p->estim_lambda = FALSE;
else optarg = &optarg[1];
p->lambda = get_arg_dbl_bounds(optarg, 0, 1);
break;
case 'C':
p->gamma = get_arg_dbl_bounds(optarg, 0, 1);
break;
case 'G':
p->gc = get_arg_dbl_bounds(optarg, 0, 1);
break;
case 't':
p->set_transitions = TRUE;
if (optarg[0] != '~')
p->estim_transitions = FALSE;
else optarg = &optarg[1];
tmpl = get_arg_list_dbl(optarg);
if (lst_size(tmpl) != 2)
die("ERROR: bad argument to --transitions.\n");
p->mu = lst_get_dbl(tmpl, 0);
p->nu = lst_get_dbl(tmpl, 1);
if (p->mu <= 0 || p->mu >= 1 || p->nu <= 0 || p->nu >= 1)
die("ERROR: bad argument to --transitions.\n");
lst_free(tmpl);
break;
case 'E':
if (optarg[0] != '~')
p->estim_transitions = FALSE;
else optarg = &optarg[1];
p->omega = get_arg_dbl_bounds(optarg, 1, INFTY);
p->mu = 1/p->omega;
break;
case 'T':
p->estim_trees = TRUE;
p->estim_trees_fname_root = optarg;
break;
case 'O':
p->estim_rho = TRUE;
p->estim_trees_fname_root = optarg;
break;
case 'z':
p->ignore_missing = TRUE;
break;
case 'k':
tmpl = get_arg_list_int(optarg);
if (lst_size(tmpl) > 2)
die("ERROR: too many arguments with --nrates.\n");
p->nrates = lst_get_int(tmpl, 0);
if (p->nrates <= 0)
die("ERROR: bad argument to --nrates (%d).\n", p->nrates);
if (lst_size(tmpl) == 2) {
p->nrates2 = lst_get_int(tmpl, 1);
if (p->nrates2 <= 0)
die("ERROR: bad argument to --nrates (%d).\n", p->nrates2);
}
lst_free(tmpl);
break;
case 'R':
p->rho = get_arg_dbl_bounds(optarg, 0, 1);
break;
case 'g':
if (!strcmp(optarg, "-"))
p->log_f = stderr;
else p->log_f = phast_fopen(optarg, "w+");
break;
case 'r':
p->refidx = get_arg_int_bounds(optarg, 0, INFTY);
break;
case 'x':
/* do nothing; left in for backward compatibility */
break;
case 'U':
p->pivot_states = get_arg_list(optarg); /* we want strings not ints
for phmm_new */
break;
case 'e':
p->extrapolate_tree_fname = optarg;
break;
case 'I':
p->indels = TRUE;
break;
case 'Y':
p->max_micro_indel = get_arg_int_bounds(optarg, 1, INFTY);
break;
case 'D':
if (optarg[0] != '~')
p->estim_indels = FALSE;
else optarg = &optarg[1];
tmpl = get_arg_list_dbl(optarg);
if (lst_size(tmpl) != 6) die("ERROR: bad argument to --indel-params.\n");
p->alpha_0 = lst_get_dbl(tmpl, 0);
p->beta_0 = lst_get_dbl(tmpl, 1);
p->tau_0 = lst_get_dbl(tmpl, 2);
p->alpha_1 = lst_get_dbl(tmpl, 3);
p->beta_1 = lst_get_dbl(tmpl, 4);
p->tau_1 = lst_get_dbl(tmpl, 5);
if (p->alpha_0 < 0 || p->beta_0 < 0 || p->tau_0 < 0 ||
p->alpha_1 < 0 || p->beta_1 < 0 || p->tau_1 < 0)
die("ERROR: bad argument to --indel-params.\n");
lst_free(tmpl);
break;
case 'J':
p->indels_only = TRUE;
p->two_state = FALSE;
p->indels = TRUE;
p->post_probs = FALSE;
break;
case 'M':
p->inform_reqd = get_arg_list(optarg);
break;
case 'F':
p->not_informative = get_arg_list(optarg);
break;
case 'c':
p->cm = cm_new_string_or_file(optarg);
break;
case 'L':
p->lnl_f = phast_fopen(optarg, "w+");
break;
case 'N':
p->seqname = optarg;
break;
case 'P':
p->idpref = optarg;
break;
case 's':
p->score = TRUE;
break;
case 'p':
coding_potential = TRUE;
break;
case 'A':
p->alias_hash = make_name_hash(optarg);
break;
case 'q':
p->results_f = NULL;
break;
case 'h':
printf("%s", HELP);
exit(0);
case '?':
die("Bad argument. Try '%s -h'.\n", argv[0]);
}
}
if ((!coding_potential && optind != argc - 2) ||
(coding_potential && optind != argc - 2 && optind != argc - 1))
die("ERROR: extra or missing arguments. Try '%s -h'.\n", argv[0]);
set_seed(-1);
if (p->extrapolate_tree_fname != NULL &&
!strcmp(p->extrapolate_tree_fname, "default")) {
p->extrapolate_tree_fname = smalloc((strlen(PHAST_HOME)+100)*sizeof(char));
#if defined(__MINGW32__)
sprintf(p->extrapolate_tree_fname,
"%s\\data\\exoniphy\\mammals\\cftr25_hybrid.nh", PHAST_HOME);
#else
sprintf(p->extrapolate_tree_fname,
"%s/data/exoniphy/mammals/cftr25_hybrid.nh", PHAST_HOME);
#endif
}
if (p->extrapolate_tree_fname != NULL)
p->extrapolate_tree = tr_new_from_file(phast_fopen(p->extrapolate_tree_fname, "r"));
mods_fname = (optind == argc - 2 ? argv[argc - 1] : NULL);
/* if there are two args, mods are the second one; otherwise will
use default mods for coding potential (see below) */
/* set defaults for coding-potential mode */
if (coding_potential) {
char tmp[5000];
p->two_state = FALSE;
if (p->cm == NULL)
p->cm = cm_new_string_or_file("NCATS=4; CNS 1; CDS 2-4");
if (p->hmm == NULL) {
#if defined(__MINGW32__)
sprintf(tmp, "%s\\data\\phastCons\\%s", PHAST_HOME,
p->indels ? "simple-coding-indels.hmm" : "simple-coding.hmm");
#else
sprintf(tmp, "%s/data/phastCons/%s", PHAST_HOME,
p->indels ? "simple-coding-indels.hmm" : "simple-coding.hmm");
#endif
if (p->results_f!=NULL)
fprintf(p->results_f, "Reading HMM from %s...\n", tmp);
p->hmm = hmm_new_from_file(phast_fopen(tmp, "r"));
}
if (mods_fname == NULL) {
#if defined(__MINGW32__)
sprintf(tmp, "%s\\data\\exoniphy\\mammals\\r3.ncns.mod, %s\\data\\exoniphy\\mammals\\r3.cns.mod, %s\\data\\exoniphy\\mammals\\r3.cds-1.mod, %s\\data\\exoniphy\\mammals\\r3.cds-2.mod, %s\\data\\exoniphy\\mammals\\r3.cds-3.mod", PHAST_HOME, PHAST_HOME, PHAST_HOME, PHAST_HOME, PHAST_HOME);
#else
sprintf(tmp, "\
%s/data/exoniphy/mammals/r3.ncns.mod,\
%s/data/exoniphy/mammals/r3.cns.mod,\
%s/data/exoniphy/mammals/r3.cds-1.mod,\
%s/data/exoniphy/mammals/r3.cds-2.mod,\
%s/data/exoniphy/mammals/r3.cds-3.mod",
PHAST_HOME, PHAST_HOME, PHAST_HOME, PHAST_HOME, PHAST_HOME);
#endif
mods_fname = tmp;
}
if (p->states == NULL)
p->states = get_arg_list("CDS");
if (p->pivot_states == NULL)
p->pivot_states = get_arg_list("background,CNS");
}
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 *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_cm_new_from_str(SEXP str) {
return rph_cm_new_extptr(cm_new_string_or_file(CHARACTER_VALUE(str)));
}
