SEXP rph_gff_nonOverlapping_genes(SEXP gffP) {
GFF_Set *gff = (GFF_Set*)EXTPTR_PTR(gffP);
if (lst_size(gff->features) == 0) return gffP;
gff_register_protect(gff);
gff_group(gff, "transcript_id");
gff_remove_overlaps(gff, NULL);
return gffP;
}
SEXP rph_gff_fix_start_stop(SEXP gffP) {
GFF_Set *gff;
gff = (GFF_Set*)EXTPTR_PTR(gffP);
gff_register_protect(gff);
gff_group(gff, "transcript_id");
gff_fix_start_stop(gff);
return gffP;
}
SEXP rph_gff_add_signals(SEXP gffP) {
GFF_Set *gff;
gff = (GFF_Set*)EXTPTR_PTR(gffP);
gff_register_protect(gff);
gff_group(gff, "transcript_id");
gff_create_signals(gff);
return gffP;
}
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[]) {
char c;
int opt_idx;
GFF_Set *gff;
List *include = NULL;
char *groupby = "transcript_id", *exongroup_tag = NULL;
int unique = FALSE, sort = FALSE, simplebed = FALSE, fix_start_stop = FALSE,
add_utrs = FALSE, add_introns = FALSE, add_signals = FALSE;
enum {GFF, BED, GENEPRED, WIG} output_format = GFF;
FILE *discards_f = NULL, *groups_f = NULL;
struct option long_opts[] = {
{"output", 1, 0, 'o'},
{"include-only", 1, 0, 'i'},
{"include-groups", 1, 0, 'l'},
{"groupby", 1, 0, 'g'},
{"exongroup", 1, 0, 'e'},
{"add-utrs", 0, 0, 'U'},
{"add-introns", 0, 0, 'I'},
{"add-signals", 0, 0, 'S'},
{"fix-start-stop", 0, 0, 'f'},
{"unique", 0, 0, 'u'},
{"sort", 0, 0, 's'},
{"simplebed", 0, 0, 'b'},
{"discards", 1, 0, 'd'},
{"help", 0, 0, 'h'},
{0, 0, 0, 0}
};
while ((c = (char)getopt_long(argc, argv, "o:i:l:g:e:d:UISfusbh", long_opts, &opt_idx)) != -1) {
switch (c) {
case 'o':
if (!strcmp("bed", optarg)) output_format = BED;
else if (!strcmp("genepred", optarg)) output_format = GENEPRED;
else if (!strcmp("wig", optarg)) output_format = WIG;
else if (strcmp("gff", optarg)) die("ERROR: bad output format.\n");
break;
case 'i':
include = get_arg_list(optarg);
break;
case 'l':
groups_f = phast_fopen(optarg, "r");
break;
case 'g':
groupby = optarg;
break;
case 'e':
exongroup_tag = optarg;
break;
case 'U':
add_utrs = TRUE;
break;
case 'I':
add_introns = TRUE;
break;
case 'S':
add_signals = TRUE;
break;
case 'f':
fix_start_stop = TRUE;
break;
case 'u':
unique = TRUE;
break;
case 'b':
simplebed = TRUE;
output_format = BED;
break;
case 'd':
discards_f = phast_fopen(optarg, "w+");
break;
case 's':
sort = TRUE;
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);
gff = gff_read_set(phast_fopen(argv[optind], "r"));
if (lst_size(gff->features) == 0) exit(0); /* helps avoid unexpected
behavior below */
/* filter by type */
if (include != NULL) gff_filter_by_type(gff, include, FALSE, discards_f);
/* group */
gff_group(gff, groupby);
/* utrs, introns, & signals */
if (add_utrs) gff_create_utrs(gff);
if (add_introns) gff_create_introns(gff);
if (add_signals) gff_create_signals(gff);
/* subgroup */
if (exongroup_tag != NULL) gff_exon_group(gff, exongroup_tag);
/* filter by group */
if (groups_f != NULL) {
String *s = str_new(STR_LONG_LEN);
List *groups = lst_new_ptr(10000);
str_slurp(s, groups_f);
str_split(s, NULL, groups);
gff_filter_by_group(gff, groups);
lst_free_strings(groups); lst_free(groups);
str_free(s);
}
/* sort */
if (sort) gff_sort(gff);
/* make unique */
if (unique) gff_remove_overlaps(gff, discards_f);
if (fix_start_stop) gff_fix_start_stop(gff);
if (output_format == BED)
gff_print_bed(stdout, gff, !simplebed);
else if (output_format == GENEPRED)
gff_print_genepred(stdout, gff);
else if (output_format == WIG)
wig_print(stdout, gff);
else
gff_print_set(stdout, gff);
gff_free_set(gff);
return 0;
}
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 *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;
}
