int ctx_clean(int argc, char **argv)
{
size_t nthreads = 0, use_ncols = 0;
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_ctx_path = NULL;
bool tip_cleaning = false, supernode_cleaning = false;
size_t min_keep_tip = 0;
Covg threshold = 0, fallback_thresh = 0;
const char *len_before_path = NULL, *len_after_path = NULL;
const char *covg_before_path = NULL, *covg_after_path = NULL;
// Arg parsing
char cmd[100];
char shortopts[300];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
// silence error messages from getopt_long
// opterr = 0;
while((c = getopt_long(argc, argv, shortopts, longopts, NULL)) != -1) {
cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd));
switch(c) {
case 0: /* flag set */ break;
case 'h': cmd_print_usage(NULL); break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'o':
if(out_ctx_path != NULL) cmd_print_usage(NULL);
out_ctx_path = optarg;
break;
case 'm': cmd_mem_args_set_memory(&memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break;
case 'N': use_ncols = cmd_uint32_nonzero(cmd, optarg); break;
case 't': cmd_check(!nthreads, cmd); nthreads = cmd_uint32_nonzero(cmd, optarg); break;
case 'T':
cmd_check(!tip_cleaning, cmd);
min_keep_tip = cmd_uint32_nonzero(cmd, optarg);
tip_cleaning = true;
break;
case 'S':
cmd_check(!supernode_cleaning, cmd);
if(optarg != NULL) threshold = cmd_uint32_nonzero(cmd, optarg);
supernode_cleaning = true;
break;
case 'B': cmd_check(!fallback_thresh, cmd); fallback_thresh = cmd_uint32_nonzero(cmd, optarg); break;
case 'l': cmd_check(!len_before_path, cmd); len_before_path = optarg; break;
case 'L': cmd_check(!len_after_path, cmd); len_after_path = optarg; break;
case 'c': cmd_check(!covg_before_path, cmd); covg_before_path = optarg; break;
case 'C': cmd_check(!covg_after_path, cmd); covg_after_path = optarg; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" clean -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
if(nthreads == 0) nthreads = DEFAULT_NTHREADS;
if(optind >= argc) cmd_print_usage("Please give input graph files");
// Default behaviour
if(!tip_cleaning && !supernode_cleaning) {
if(out_ctx_path != NULL)
supernode_cleaning = tip_cleaning = true; // do both
else
warn("No cleaning being done: you did not specify --out <out.ctx>");
}
bool doing_cleaning = (supernode_cleaning || tip_cleaning);
if(doing_cleaning && out_ctx_path == NULL) {
cmd_print_usage("Please specify --out <out.ctx> for cleaned graph");
}
if(!doing_cleaning && (covg_after_path || len_after_path)) {
cmd_print_usage("You gave --len-after <out> / --covg-after <out> without "
"any cleaning (set -s, --supernodes or -t, --tips)");
}
if(doing_cleaning && strcmp(out_ctx_path,"-") != 0 &&
!futil_get_force() && futil_file_exists(out_ctx_path))
{
cmd_print_usage("Output file already exists: %s", out_ctx_path);
}
if(fallback_thresh && !supernode_cleaning)
cmd_print_usage("-B, --fallback <T> without --supernodes");
// Use remaining args as graph files
char **gfile_paths = argv + optind;
size_t i, j, num_gfiles = (size_t)(argc - optind);
// Open graph files
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
size_t ncols, ctx_max_kmers = 0, ctx_sum_kmers = 0;
ncols = graph_files_open(gfile_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
size_t kmer_size = gfiles[0].hdr.kmer_size;
// default to one colour for now
if(use_ncols == 0) use_ncols = 1;
// Flatten if we don't have to remember colours / output a graph
if(!doing_cleaning)
{
ncols = use_ncols = 1;
for(i = 0; i < num_gfiles; i++)
file_filter_flatten(&gfiles[i].fltr, 0);
}
if(ncols < use_ncols) {
warn("I only need %zu colour%s ('--ncols %zu' ignored)",
ncols, util_plural_str(ncols), use_ncols);
use_ncols = ncols;
}
char max_kmers_str[100];
ulong_to_str(ctx_max_kmers, max_kmers_str);
status("%zu input graph%s, max kmers: %s, using %zu colours",
num_gfiles, util_plural_str(num_gfiles), max_kmers_str, use_ncols);
// If no arguments given we default to removing tips < 2*kmer_size
if(tip_cleaning && min_keep_tip == 0)
min_keep_tip = 2 * kmer_size;
// Warn if any graph files already cleaned
size_t fromcol, intocol;
ErrorCleaning *cleaning;
for(i = 0; i < num_gfiles; i++) {
for(j = 0; j < file_filter_num(&gfiles[i].fltr); j++) {
fromcol = file_filter_fromcol(&gfiles[i].fltr, j);
cleaning = &gfiles[i].hdr.ginfo[fromcol].cleaning;
if(cleaning->cleaned_snodes && supernode_cleaning) {
warn("%s:%zu already has supernode cleaning with threshold: <%zu",
file_filter_path(&gfiles[i].fltr), fromcol,
(size_t)cleaning->clean_snodes_thresh);
}
if(cleaning->cleaned_tips && tip_cleaning) {
warn("%s:%zu already has had tip cleaned",
file_filter_path(&gfiles[i].fltr), fromcol);
}
}
}
// Print steps
size_t step = 0;
status("Actions:\n");
if(covg_before_path != NULL)
status("%zu. Saving kmer coverage distribution to: %s", step++, covg_before_path);
if(len_before_path != NULL)
status("%zu. Saving supernode length distribution to: %s", step++, len_before_path);
if(tip_cleaning)
status("%zu. Cleaning tips shorter than %zu nodes", step++, min_keep_tip);
if(supernode_cleaning && threshold > 0)
status("%zu. Cleaning supernodes with coverage < %u", step++, threshold);
if(supernode_cleaning && threshold <= 0)
status("%zu. Cleaning supernodes with auto-detected threshold", step++);
if(covg_after_path != NULL)
status("%zu. Saving kmer coverage distribution to: %s", step++, covg_after_path);
if(len_after_path != NULL)
status("%zu. Saving supernode length distribution to: %s", step++, len_after_path);
//
// Decide memory usage
//
bool all_colours_loaded = (ncols <= use_ncols);
bool use_mem_limit = (memargs.mem_to_use_set && num_gfiles > 1) || !ctx_max_kmers;
size_t kmers_in_hash, bits_per_kmer, graph_mem;
size_t per_kmer_per_col_bits = (sizeof(BinaryKmer)+sizeof(Covg)+sizeof(Edges)) * 8;
size_t pop_edges_per_kmer_bits = (all_colours_loaded ? 0 : sizeof(Edges) * 8);
bits_per_kmer = per_kmer_per_col_bits * use_ncols + pop_edges_per_kmer_bits;
kmers_in_hash = cmd_get_kmers_in_hash(memargs.mem_to_use,
memargs.mem_to_use_set,
memargs.num_kmers,
memargs.num_kmers_set,
bits_per_kmer,
ctx_max_kmers, ctx_sum_kmers,
use_mem_limit, &graph_mem);
// Maximise the number of colours we load to fill the mem
size_t max_usencols = (memargs.mem_to_use*8 - pop_edges_per_kmer_bits * kmers_in_hash) /
(per_kmer_per_col_bits * kmers_in_hash);
use_ncols = MIN2(max_usencols, ncols);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
//
// Check output files are writable
//
futil_create_output(out_ctx_path);
// Does nothing if arg is NULL
futil_create_output(covg_before_path);
futil_create_output(covg_after_path);
futil_create_output(len_before_path);
futil_create_output(len_after_path);
// Create db_graph
// Load as many colours as possible
// Use an extra set of edge to take intersections
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, use_ncols, use_ncols,
kmers_in_hash, DBG_ALLOC_COVGS);
// Edges is a special case
size_t num_edges = db_graph.ht.capacity * (use_ncols + !all_colours_loaded);
db_graph.col_edges = ctx_calloc(num_edges, sizeof(Edges));
// Load graph into a single colour
LoadingStats stats = LOAD_STATS_INIT_MACRO;
GraphLoadingPrefs gprefs = {.db_graph = &db_graph,
.boolean_covgs = false,
.must_exist_in_graph = false,
.must_exist_in_edges = NULL,
.empty_colours = false};
// Construct cleaned graph header
GraphFileHeader outhdr;
memset(&outhdr, 0, sizeof(GraphFileHeader));
outhdr.version = CTX_GRAPH_FILEFORMAT;
outhdr.kmer_size = db_graph.kmer_size;
outhdr.num_of_cols = ncols;
outhdr.num_of_bitfields = (db_graph.kmer_size*2+63)/64;
graph_header_alloc(&outhdr, ncols);
// Merge info into header
size_t gcol = 0;
for(i = 0; i < num_gfiles; i++) {
for(j = 0; j < file_filter_num(&gfiles[i].fltr); j++, gcol++) {
fromcol = file_filter_fromcol(&gfiles[i].fltr, j);
intocol = file_filter_intocol(&gfiles[i].fltr, j);
graph_info_merge(&outhdr.ginfo[intocol], &gfiles[i].hdr.ginfo[fromcol]);
}
}
if(ncols > use_ncols) {
graph_files_load_flat(gfiles, num_gfiles, gprefs, &stats);
} else {
for(i = 0; i < num_gfiles; i++)
graph_load(&gfiles[i], gprefs, &stats);
}
char num_kmers_str[100];
ulong_to_str(db_graph.ht.num_kmers, num_kmers_str);
status("Total kmers loaded: %s\n", num_kmers_str);
size_t initial_nkmers = db_graph.ht.num_kmers;
hash_table_print_stats(&db_graph.ht);
uint8_t *visited = ctx_calloc(roundup_bits2bytes(db_graph.ht.capacity), 1);
uint8_t *keep = ctx_calloc(roundup_bits2bytes(db_graph.ht.capacity), 1);
if((supernode_cleaning && threshold <= 0) || covg_before_path || len_before_path)
{
// Get coverage distribution and estimate cleaning threshold
int est_threshold = cleaning_get_threshold(nthreads,
covg_before_path,
len_before_path,
visited, &db_graph);
if(est_threshold < 0) status("Cannot find recommended cleaning threshold");
else status("Recommended cleaning threshold is: %i", est_threshold);
// Use estimated threshold if threshold not set
if(threshold <= 0) {
if(fallback_thresh > 0 && est_threshold < (int)fallback_thresh) {
status("Using fallback threshold: %i", fallback_thresh);
threshold = fallback_thresh;
}
else if(est_threshold >= 0) threshold = est_threshold;
}
}
// Die if we failed to find suitable cleaning threshold
if(supernode_cleaning && threshold <= 0)
die("Need cleaning threshold (--supernodes=<D> or --fallback <D>)");
if(doing_cleaning) {
// Clean graph of tips (if min_keep_tip > 0) and supernodes (if threshold > 0)
clean_graph(nthreads, threshold, min_keep_tip,
covg_after_path, len_after_path,
visited, keep, &db_graph);
}
ctx_free(visited);
ctx_free(keep);
if(doing_cleaning)
{
// Output graph file
Edges *intersect_edges = NULL;
bool kmers_loaded = true;
size_t col, thresh;
// Set output header ginfo cleaned
for(col = 0; col < ncols; col++)
{
cleaning = &outhdr.ginfo[col].cleaning;
cleaning->cleaned_snodes |= supernode_cleaning;
cleaning->cleaned_tips |= tip_cleaning;
// if(tip_cleaning) {
// strbuf_append_str(&outhdr.ginfo[col].sample_name, ".tipclean");
// }
if(supernode_cleaning) {
thresh = cleaning->clean_snodes_thresh;
thresh = cleaning->cleaned_snodes ? MAX2(thresh, (uint32_t)threshold)
: (uint32_t)threshold;
cleaning->clean_snodes_thresh = thresh;
// char name_append[200];
// sprintf(name_append, ".supclean%zu", thresh);
// strbuf_append_str(&outhdr.ginfo[col].sample_name, name_append);
}
}
if(!all_colours_loaded)
{
// We haven't loaded all the colours
// intersect_edges are edges to mask with
// resets graph edges
intersect_edges = db_graph.col_edges;
db_graph.col_edges += db_graph.ht.capacity;
}
// Print stats on removed kmers
size_t removed_nkmers = initial_nkmers - db_graph.ht.num_kmers;
double removed_pct = (100.0 * removed_nkmers) / initial_nkmers;
char removed_str[100], init_str[100];
ulong_to_str(removed_nkmers, removed_str);
ulong_to_str(initial_nkmers, init_str);
status("Removed %s of %s (%.2f%%) kmers", removed_str, init_str, removed_pct);
graph_files_merge(out_ctx_path, gfiles, num_gfiles,
kmers_loaded, all_colours_loaded,
intersect_edges, &outhdr, &db_graph);
// Swap back
if(!all_colours_loaded)
db_graph.col_edges = intersect_edges;
}
ctx_check(db_graph.ht.num_kmers == hash_table_count_kmers(&db_graph.ht));
graph_header_dealloc(&outhdr);
for(i = 0; i < num_gfiles; i++) graph_file_close(&gfiles[i]);
ctx_free(gfiles);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
/*!
@see futil_fopen_create()
*/
gzFile futil_gzopen_create(const char *path, const char *mode)
{
ctx_assert(path != NULL);
futil_create_output(path);
return futil_gzopen(path, mode);
}
int ctx_calls2vcf(int argc, char **argv)
{
const char *in_path = NULL, *out_path = NULL, *out_type = NULL;
// Filtering parameters
int32_t min_mapq = -1, max_align_len = -1, max_allele_len = -1;
// Alignment parameters
int nwmatch = 1, nwmismatch = -2, nwgapopen = -4, nwgapextend = -1;
// ref paths
char const*const* ref_paths = NULL;
size_t nref_paths = 0;
// flank file
const char *sam_path = NULL;
//
// Things we figure out by looking at the input
//
bool isbubble = false;
// samples in VCF, (0 for bubble, does not include ref in breakpoint calls)
size_t i, kmer_size, num_samples;
//
// Reference genome
//
// Hash map of chromosome name -> sequence
ChromHash *genome;
ReadBuffer chroms;
// Arg parsing
char cmd[100];
char shortopts[300];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
// silence error messages from getopt_long
// opterr = 0;
while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) {
cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd));
switch(c) {
case 0: /* flag set */ break;
case 'h': cmd_print_usage(NULL); break;
case 'o': cmd_check(!out_path, cmd); out_path = optarg; break;
case 'O': cmd_check(!out_type, cmd); out_type = optarg; break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'F': cmd_check(!sam_path,cmd); sam_path = optarg; break;
case 'Q': cmd_check(min_mapq < 0,cmd); min_mapq = cmd_uint32(cmd, optarg); break;
case 'A': cmd_check(max_align_len < 0,cmd); max_align_len = cmd_uint32(cmd, optarg); break;
case 'L': cmd_check(max_allele_len < 0,cmd); max_allele_len = cmd_uint32(cmd, optarg); break;
case 'm': nwmatch = cmd_int32(cmd, optarg); break;
case 'M': nwmismatch = cmd_int32(cmd, optarg); break;
case 'g': nwgapopen = cmd_int32(cmd, optarg); break;
case 'G': nwgapextend = cmd_int32(cmd, optarg); break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
die("`"CMD" "SUBCMD" -h` for help. Bad option: %s", argv[optind-1]);
default: ctx_assert2(0, "shouldn't reach here: %c", c);
}
}
// Defaults for unset values
if(out_path == NULL) out_path = "-";
if(max_align_len < 0) max_align_len = DEFAULT_MAX_ALIGN;
if(max_allele_len < 0) max_allele_len = DEFAULT_MAX_ALLELE;
if(optind+2 > argc)
cmd_print_usage("Require <in.txt.gz> and at least one reference");
in_path = argv[optind++];
ref_paths = (char const*const*)argv + optind;
nref_paths = argc - optind;
// These functions call die() on error
gzFile gzin = futil_gzopen(in_path, "r");
// Read call file header
cJSON *json = json_hdr_load(gzin, in_path);
// Check we can handle the kmer size
kmer_size = json_hdr_get_kmer_size(json, in_path);
db_graph_check_kmer_size(kmer_size, in_path);
// Get format (bubble or breakpoint file)
cJSON *json_fmt = json_hdr_get(json, "file_format", cJSON_String, in_path);
if(strcmp(json_fmt->valuestring,"CtxBreakpoints") == 0) isbubble = false;
else if(strcmp(json_fmt->valuestring,"CtxBubbles") == 0) isbubble = true;
else die("Unknown format: '%s'", json_fmt->valuestring);
status("Reading %s in %s format", futil_inpath_str(in_path),
isbubble ? "bubble" : "breakpoint");
if(isbubble) {
// bubble specific
if(sam_path == NULL)
cmd_print_usage("Require -F <flanks.sam> with bubble file");
if(min_mapq < 0) min_mapq = DEFAULT_MIN_MAPQ;
}
else {
// breakpoint specific
if(min_mapq >= 0)
cmd_print_usage("-Q,--min-mapq <Q> only valid with bubble calls");
}
// Open flank file if it exists
htsFile *samfh = NULL;
bam_hdr_t *bam_hdr = NULL;
bam1_t *mflank = NULL;
if(sam_path)
{
if((samfh = hts_open(sam_path, "r")) == NULL)
die("Cannot open SAM/BAM %s", sam_path);
// Load BAM header
bam_hdr = sam_hdr_read(samfh);
if(bam_hdr == NULL) die("Cannot load BAM header: %s", sam_path);
mflank = bam_init1();
}
// Output VCF has 0 samples if bubbles file, otherwise has N where N is
// number of samples/colours in the breakpoint graph
size_t num_graph_samples = json_hdr_get_ncols(json, in_path);
size_t num_graph_nonref = json_hdr_get_nonref_ncols(json, in_path);
num_samples = 0;
if(!isbubble) {
// If last colour has "is_ref", drop number of samples by one
num_samples = num_graph_nonref < num_graph_samples ? num_graph_samples-1
: num_graph_samples;
}
//
// Open output file
//
if(!out_path) out_path = "-";
int mode = vcf_misc_get_outtype(out_type, out_path);
futil_create_output(out_path);
htsFile *vcffh = hts_open(out_path, modes_htslib[mode]);
status("[calls2vcf] Reading %s call file with %zu samples",
isbubble ? "Bubble" : "Breakpoint", num_graph_samples);
status("[calls2vcf] %zu sample output to: %s format: %s",
num_samples, futil_outpath_str(out_path), hsmodes_htslib[mode]);
if(isbubble) status("[calls2vcf] min. MAPQ: %i", min_mapq);
status("[calls2vcf] max alignment length: %i", max_align_len);
status("[calls2vcf] max VCF allele length: %i", max_allele_len);
status("[calls2vcf] alignment match:%i mismatch:%i gap open:%i extend:%i",
nwmatch, nwmismatch, nwgapopen, nwgapextend);
// Load reference genome
read_buf_alloc(&chroms, 1024);
genome = chrom_hash_init();
chrom_hash_load(ref_paths, nref_paths, &chroms, genome);
// convert to upper case
char *s;
for(i = 0; i < chroms.len; i++)
for(s = chroms.b[i].seq.b; *s; s++) *s = toupper(*s);
if(!isbubble) brkpnt_check_refs_match(json, genome, in_path);
bcf_hdr_t *vcfhdr = make_vcf_hdr(json, in_path, !isbubble, kmer_size,
ref_paths, nref_paths,
chroms.b, chroms.len);
if(bcf_hdr_write(vcffh, vcfhdr) != 0) die("Cannot write VCF header");
AlignedCall *call = acall_init();
CallDecomp *aligner = call_decomp_init(vcffh, vcfhdr);
scoring_t *scoring = call_decomp_get_scoring(aligner);
scoring_init(scoring, nwmatch, nwmismatch, nwgapopen, nwgapextend,
false, false, 0, 0, 0, 0);
CallFileEntry centry;
call_file_entry_alloc(¢ry);
char kmer_str[50];
sprintf(kmer_str, ";K%zu", kmer_size);
if(isbubble)
{
// Bubble calls
DecompBubble *bubbles = decomp_bubble_init();
// Set scoring for aligning 3' flank
scoring = decomp_bubble_get_scoring(bubbles);
scoring_init(scoring, nwmatch, nwmismatch, nwgapopen, nwgapextend,
true, true, 0, 0, 0, 0);
while(call_file_read(gzin, in_path, ¢ry)) {
do {
if(sam_read1(samfh, bam_hdr, mflank) < 0)
die("We've run out of SAM entries!");
} while(mflank->core.flag & (BAM_FSECONDARY | BAM_FSUPPLEMENTARY));
// Align call
strbuf_reset(&call->info);
decomp_bubble_call(bubbles, genome, kmer_size, min_mapq,
¢ry, mflank, bam_hdr, call);
strbuf_append_str(&call->info, kmer_str);
acall_decompose(aligner, call, max_align_len, max_allele_len);
}
// print bubble stats
DecompBubbleStats *bub_stats = ctx_calloc(1, sizeof(*bub_stats));
decomp_bubble_cpy_stats(bub_stats, bubbles);
print_bubble_stats(bub_stats);
ctx_free(bub_stats);
decomp_bubble_destroy(bubbles);
}
else
{
// Breakpoint calls
DecompBreakpoint *breakpoints = decomp_brkpt_init();
while(call_file_read(gzin, in_path, ¢ry)) {
strbuf_reset(&call->info);
decomp_brkpt_call(breakpoints, genome, num_samples, ¢ry, call);
strbuf_append_str(&call->info, kmer_str);
acall_decompose(aligner, call, max_align_len, max_allele_len);
}
// print bubble stats
DecompBreakpointStats *brk_stats = ctx_calloc(1, sizeof(*brk_stats));
decomp_brkpt_cpy_stats(brk_stats, breakpoints);
print_breakpoint_stats(brk_stats);
ctx_free(brk_stats);
decomp_brkpt_destroy(breakpoints);
}
// Print stats
DecomposeStats *astats = ctx_calloc(1, sizeof(*astats));
call_decomp_cpy_stats(astats, aligner);
print_acall_stats(astats);
ctx_free(astats);
call_file_entry_dealloc(¢ry);
call_decomp_destroy(aligner);
acall_destroy(call);
// Finished - clean up
cJSON_Delete(json);
gzclose(gzin);
bcf_hdr_destroy(vcfhdr);
hts_close(vcffh);
for(i = 0; i < chroms.len; i++) seq_read_dealloc(&chroms.b[i]);
read_buf_dealloc(&chroms);
chrom_hash_destroy(genome);
if(sam_path) {
hts_close(samfh);
bam_hdr_destroy(bam_hdr);
bam_destroy1(mflank);
}
return EXIT_SUCCESS;
}
int ctx_pop_bubbles(int argc, char **argv)
{
size_t nthreads = 0;
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_path = NULL;
int32_t max_covg = -1; // max mean coverage to remove <=0 => ignore
int32_t max_klen = -1; // max length (kmers) to remove <=0 => ignore
int32_t max_kdiff = -1; // max diff between bubble branch lengths <0 => ignore
// Arg parsing
char cmd[100];
char shortopts[300];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
// silence error messages from getopt_long
// opterr = 0;
while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) {
cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd));
switch(c) {
case 0: /* flag set */ break;
case 'h': cmd_print_usage(NULL); break;
case 'o': cmd_check(!out_path, cmd); out_path = optarg; break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 't': cmd_check(!nthreads, cmd); nthreads = cmd_uint32_nonzero(cmd, optarg); break;
case 'm': cmd_mem_args_set_memory(&memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break;
case 'C': cmd_check(max_covg<0, cmd); max_covg = cmd_uint32(cmd, optarg); break;
case 'L': cmd_check(max_klen<0, cmd); max_klen = cmd_uint32(cmd, optarg); break;
case 'D': cmd_check(max_kdiff<0, cmd); max_kdiff = cmd_uint32(cmd, optarg); break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" pop -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
// Defaults for unset values
if(out_path == NULL) out_path = "-";
if(nthreads == 0) nthreads = DEFAULT_NTHREADS;
if(optind >= argc) cmd_print_usage("Require input graph files (.ctx)");
//
// Open graph files
//
const size_t num_gfiles = argc - optind;
char **graph_paths = argv + optind;
ctx_assert(num_gfiles > 0);
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
size_t i, ncols, ctx_max_kmers = 0, ctx_sum_kmers = 0;
ncols = graph_files_open(graph_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
bool reread_graph_to_filter = (num_gfiles == 1 &&
strcmp(file_filter_path(&gfiles[0].fltr),"-") != 0);
if(reread_graph_to_filter) {
file_filter_flatten(&gfiles[0].fltr, 0);
ncols = 1;
}
// Check graphs are compatible
graphs_gpaths_compatible(gfiles, num_gfiles, NULL, 0, -1);
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem;
bits_per_kmer = sizeof(BinaryKmer)*8 +
sizeof(Covg)*8*ncols +
sizeof(Edges)*8*ncols +
2; // 1 bit for visited, 1 for removed
kmers_in_hash = cmd_get_kmers_in_hash(memargs.mem_to_use,
memargs.mem_to_use_set,
memargs.num_kmers,
memargs.num_kmers_set,
bits_per_kmer,
ctx_max_kmers, ctx_sum_kmers,
false, &graph_mem);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
// Check out_path is writable
futil_create_output(out_path);
// Allocate memory
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, ncols, ncols,
kmers_in_hash, DBG_ALLOC_EDGES | DBG_ALLOC_COVGS);
size_t nkwords = roundup_bits2bytes(db_graph.ht.capacity);
uint8_t *visited = ctx_calloc(1, nkwords);
uint8_t *rmvbits = ctx_calloc(1, nkwords);
//
// Load graphs
//
GraphLoadingPrefs gprefs = graph_loading_prefs(&db_graph);
gprefs.empty_colours = true;
for(i = 0; i < num_gfiles; i++) {
graph_load(&gfiles[i], gprefs, NULL);
graph_file_close(&gfiles[i]);
gprefs.empty_colours = false;
}
ctx_free(gfiles);
hash_table_print_stats(&db_graph.ht);
PopBubblesPrefs prefs = {.max_rmv_covg = max_covg,
.max_rmv_klen = max_klen,
.max_rmv_kdiff = max_kdiff};
size_t npopped = 0;
char npopped_str[50];
status("Popping bubbles...");
npopped = pop_bubbles(&db_graph, nthreads, prefs, visited, rmvbits);
ulong_to_str(npopped, npopped_str);
status("Popped %s bubbles", npopped_str);
size_t nkmers0 = db_graph.ht.num_kmers;
status("Removing nodes...");
for(i = 0; i < nkwords; i++) rmvbits[i] = ~rmvbits[i];
prune_nodes_lacking_flag(nthreads, rmvbits, &db_graph);
size_t nkmers1 = db_graph.ht.num_kmers;
ctx_assert(nkmers1 <= nkmers0);
char nkmers0str[50], nkmers1str[50], ndiffstr[50];
ulong_to_str(nkmers0, nkmers0str);
ulong_to_str(nkmers1, nkmers1str);
ulong_to_str(nkmers0-nkmers1, ndiffstr);
status("Number of kmers %s -> %s (-%s)", nkmers0str, nkmers1str, ndiffstr);
if(reread_graph_to_filter)
{
status("Streaming filtered file to: %s\n", out_path);
GraphFileReader gfile;
memset(&gfile, 0, sizeof(GraphFileReader));
graph_file_open(&gfile, graph_paths[0]);
graph_writer_stream_mkhdr(out_path, &gfile, &db_graph,
db_graph.col_edges, NULL);
graph_file_close(&gfile);
}
else
{
status("Saving to: %s\n", out_path);
graph_writer_save_mkhdr(out_path, &db_graph, CTX_GRAPH_FILEFORMAT, NULL,
0, ncols);
}
ctx_free(visited);
ctx_free(rmvbits);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
int ctx_join(int argc, char **argv)
{
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_path = NULL;
size_t use_ncols = 0;
GraphFileReader tmp_gfile;
GraphFileBuffer isec_gfiles_buf;
gfile_buf_alloc(&isec_gfiles_buf, 8);
// Arg parsing
char cmd[100], shortopts[100];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) {
cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd));
switch(c) {
case 0: /* flag set */ break;
case 'h': cmd_print_usage(NULL); break;
case 'o': cmd_check(!out_path, cmd); out_path = optarg; break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'm': cmd_mem_args_set_memory(&memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break;
case 'N': cmd_check(!use_ncols, cmd); use_ncols = cmd_uint32_nonzero(cmd, optarg); break;
case 'i':
graph_file_reset(&tmp_gfile);
graph_file_open(&tmp_gfile, optarg);
if(file_filter_into_ncols(&tmp_gfile.fltr) > 1)
warn("Flattening intersection graph into colour 0: %s", optarg);
file_filter_flatten(&tmp_gfile.fltr, 0);
gfile_buf_push(&isec_gfiles_buf, &tmp_gfile, 1);
break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" join -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
GraphFileReader *igfiles = isec_gfiles_buf.b;
size_t num_igfiles = isec_gfiles_buf.len;
if(!out_path) cmd_print_usage("--out <out.ctx> required");
if(optind >= argc)
cmd_print_usage("Please specify at least one input graph file");
// optind .. argend-1 are graphs to load
size_t num_gfiles = (size_t)(argc - optind);
char **gfile_paths = argv + optind;
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
status("Probing %zu graph files and %zu intersect files", num_gfiles, num_igfiles);
// Check all binaries are valid binaries with matching kmer size
size_t i;
size_t ctx_max_cols = 0;
uint64_t min_intersect_num_kmers = 0, ctx_max_kmers = 0, ctx_sum_kmers = 0;
for(i = 0; i < num_gfiles; i++)
{
graph_file_open2(&gfiles[i], gfile_paths[i], "r", true, ctx_max_cols);
if(gfiles[0].hdr.kmer_size != gfiles[i].hdr.kmer_size) {
cmd_print_usage("Kmer sizes don't match [%u vs %u]",
gfiles[0].hdr.kmer_size, gfiles[i].hdr.kmer_size);
}
ctx_max_cols = MAX2(ctx_max_cols, file_filter_into_ncols(&gfiles[i].fltr));
ctx_max_kmers = MAX2(ctx_max_kmers, graph_file_nkmers(&gfiles[i]));
ctx_sum_kmers += graph_file_nkmers(&gfiles[i]);
}
// Probe intersection graph files
for(i = 0; i < num_igfiles; i++)
{
if(gfiles[0].hdr.kmer_size != igfiles[i].hdr.kmer_size) {
cmd_print_usage("Kmer sizes don't match [%u vs %u]",
gfiles[0].hdr.kmer_size, igfiles[i].hdr.kmer_size);
}
uint64_t nkmers = graph_file_nkmers(&igfiles[i]);
if(i == 0) min_intersect_num_kmers = nkmers;
else if(nkmers < min_intersect_num_kmers)
{
// Put smallest intersection binary first
SWAP(igfiles[i], igfiles[0]);
min_intersect_num_kmers = nkmers;
}
}
bool take_intersect = (num_igfiles > 0);
// If we are taking an intersection,
// all kmers intersection kmers will need to be loaded
if(take_intersect)
ctx_max_kmers = ctx_sum_kmers = min_intersect_num_kmers;
bool use_ncols_set = (use_ncols > 0);
bool output_to_stdout = (strcmp(out_path,"-") == 0);
// if(use_ncols == 0) use_ncols = 1;
if(use_ncols_set) {
if(use_ncols < ctx_max_cols && output_to_stdout)
die("I need %zu colours if outputting to STDOUT (--ncols)", ctx_max_cols);
if(use_ncols > ctx_max_cols) {
warn("I only need %zu colour%s ('--ncols %zu' ignored)",
ctx_max_cols, util_plural_str(ctx_max_cols), use_ncols);
use_ncols = ctx_max_cols;
}
}
else {
use_ncols = output_to_stdout ? ctx_max_cols : 1;
}
// Check out_path is writable
futil_create_output(out_path);
status("Output %zu cols; from %zu files; intersecting %zu graphs; ",
ctx_max_cols, num_gfiles, num_igfiles);
if(num_gfiles == 1 && num_igfiles == 0)
{
// Loading only one file with no intersection files
// Don't need to store a graph in memory, can filter as stream
// Don't actually store anything in the de Bruijn graph, but we need to
// pass it, so mock one up
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size,
file_filter_into_ncols(&gfiles[0].fltr), 0, 1024, 0);
graph_writer_stream_mkhdr(out_path, &gfiles[0], &db_graph, NULL, NULL);
graph_file_close(&gfiles[0]);
gfile_buf_dealloc(&isec_gfiles_buf);
ctx_free(gfiles);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem;
bits_per_kmer = sizeof(BinaryKmer)*8 +
(sizeof(Covg) + sizeof(Edges)) * 8 * use_ncols;
kmers_in_hash = cmd_get_kmers_in_hash(memargs.mem_to_use,
memargs.mem_to_use_set,
memargs.num_kmers,
memargs.num_kmers_set,
bits_per_kmer,
ctx_max_kmers, ctx_sum_kmers,
true, &graph_mem);
if(!use_ncols_set)
{
// Maximise use_ncols
size_t max_usencols = (memargs.mem_to_use*8) / bits_per_kmer;
use_ncols = MIN2(max_usencols, ctx_max_cols);
bits_per_kmer = sizeof(BinaryKmer)*8 +
(sizeof(Covg) + sizeof(Edges)) * 8 * use_ncols;
// Re-check memory used
kmers_in_hash = cmd_get_kmers_in_hash(memargs.mem_to_use,
memargs.mem_to_use_set,
memargs.num_kmers,
memargs.num_kmers_set,
bits_per_kmer,
ctx_max_kmers, ctx_sum_kmers,
true, &graph_mem);
}
status("Using %zu colour%s in memory", use_ncols, util_plural_str(use_ncols));
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
// Create db_graph
dBGraph db_graph;
Edges *intersect_edges = NULL;
size_t edge_cols = (use_ncols + take_intersect);
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, use_ncols, use_ncols,
kmers_in_hash, DBG_ALLOC_COVGS);
// We allocate edges ourself since it's a special case
db_graph.col_edges = ctx_calloc(db_graph.ht.capacity*edge_cols, sizeof(Edges));
// Load intersection binaries
char *intsct_gname_ptr = NULL;
StrBuf intersect_gname;
strbuf_alloc(&intersect_gname, 1024);
if(take_intersect)
{
GraphLoadingPrefs gprefs = graph_loading_prefs(&db_graph);
gprefs.boolean_covgs = true; // covg++ only
for(i = 0; i < num_igfiles; i++)
{
graph_load(&igfiles[i], gprefs, NULL);
// Update intersect header
// note: intersection graphs all load exactly one colour into colour 0
graph_info_make_intersect(&igfiles[i].hdr.ginfo[0], &intersect_gname);
gprefs.must_exist_in_graph = true;
gprefs.must_exist_in_edges = db_graph.col_edges;
}
if(num_igfiles > 1)
{
// Remove nodes where covg != num_igfiles
HASH_ITERATE_SAFE(&db_graph.ht, remove_non_intersect_nodes,
db_graph.col_covgs, (Covg)num_igfiles, &db_graph.ht);
}
status("Loaded intersection set\n");
intsct_gname_ptr = intersect_gname.b;
for(i = 0; i < num_igfiles; i++) graph_file_close(&igfiles[i]);
// Reset graph info
for(i = 0; i < db_graph.num_of_cols; i++)
graph_info_init(&db_graph.ginfo[i]);
// Zero covgs
memset(db_graph.col_covgs, 0, db_graph.ht.capacity * sizeof(Covg));
// Use union edges we loaded to intersect new edges
intersect_edges = db_graph.col_edges;
db_graph.col_edges += db_graph.ht.capacity;
}
bool kmers_loaded = take_intersect, colours_loaded = false;
graph_writer_merge_mkhdr(out_path, gfiles, num_gfiles,
kmers_loaded, colours_loaded, intersect_edges,
intsct_gname_ptr, &db_graph);
if(take_intersect)
db_graph.col_edges -= db_graph.ht.capacity;
for(i = 0; i < num_gfiles; i++) graph_file_close(&gfiles[i]);
strbuf_dealloc(&intersect_gname);
gfile_buf_dealloc(&isec_gfiles_buf);
ctx_free(gfiles);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
int ctx_clean(int argc, char **argv)
{
size_t nthreads = 0, use_ncols = 0;
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_ctx_path = NULL;
int min_keep_tip = -1, unitig_min = -1; // <0 => default, 0 => noclean
uint32_t fallback_thresh = 0;
const char *len_before_path = NULL, *len_after_path = NULL;
const char *covg_before_path = NULL, *covg_after_path = NULL;
// Arg parsing
char cmd[100];
char shortopts[300];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
// silence error messages from getopt_long
// opterr = 0;
while((c = getopt_long(argc, argv, shortopts, longopts, NULL)) != -1) {
cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd));
switch(c) {
case 0: /* flag set */ break;
case 'h': cmd_print_usage(NULL); break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'o':
if(out_ctx_path != NULL) cmd_print_usage(NULL);
out_ctx_path = optarg;
break;
case 'm': cmd_mem_args_set_memory(&memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break;
case 'N': use_ncols = cmd_uint32_nonzero(cmd, optarg); break;
case 't': cmd_check(!nthreads, cmd); nthreads = cmd_uint32_nonzero(cmd, optarg); break;
case 'T':
cmd_check(min_keep_tip<0, cmd);
min_keep_tip = (optarg != NULL ? (int)cmd_uint32(cmd, optarg) : -1);
break;
case 'S':
case 'U':
cmd_check(unitig_min<0, cmd);
unitig_min = (optarg != NULL ? cmd_uint32(cmd, optarg) : -1);
break;
case 'B': cmd_check(!fallback_thresh, cmd); fallback_thresh = cmd_uint32_nonzero(cmd, optarg); break;
case 'l': cmd_check(!len_before_path, cmd); len_before_path = optarg; break;
case 'L': cmd_check(!len_after_path, cmd); len_after_path = optarg; break;
case 'c': cmd_check(!covg_before_path, cmd); covg_before_path = optarg; break;
case 'C': cmd_check(!covg_after_path, cmd); covg_after_path = optarg; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" clean -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
if(nthreads == 0) nthreads = DEFAULT_NTHREADS;
if(optind >= argc) cmd_print_usage("Please give input graph files");
bool unitig_cleaning = (unitig_min != 0);
bool tip_cleaning = (min_keep_tip != 0);
bool doing_cleaning = (unitig_cleaning || tip_cleaning);
// If you ever want to estimate cleaning threshold without outputting
// a graph, change this to a warning
if(doing_cleaning && out_ctx_path == NULL) {
cmd_print_usage("Please specify --out <out.ctx> for cleaned graph");
// warn("No cleaning being done: you did not specify --out <out.ctx>");
}
if(!doing_cleaning && (covg_after_path || len_after_path)) {
warn("You gave --len-after <out> / --covg-after <out> without "
"any cleaning (set -U, --unitigs or -t, --tips)");
}
if(doing_cleaning && strcmp(out_ctx_path,"-") != 0 &&
!futil_get_force() && futil_file_exists(out_ctx_path))
{
cmd_print_usage("Output file already exists: %s", out_ctx_path);
}
if(fallback_thresh && !unitig_cleaning)
warn("-B, --fallback <T> without --unitigs");
// Use remaining args as graph files
char **gfile_paths = argv + optind;
size_t i, j, num_gfiles = (size_t)(argc - optind);
// Open graph files
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
size_t col, ncols, ctx_max_kmers = 0, ctx_sum_kmers = 0;
ncols = graph_files_open(gfile_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
size_t kmer_size = gfiles[0].hdr.kmer_size;
// default to one colour for now
if(use_ncols == 0) use_ncols = 1;
// Flatten if we don't have to remember colours / output a graph
if(out_ctx_path == NULL)
{
ncols = use_ncols = 1;
for(i = 0; i < num_gfiles; i++)
file_filter_flatten(&gfiles[i].fltr, 0);
}
if(ncols < use_ncols) {
warn("I only need %zu colour%s ('--ncols %zu' ignored)",
ncols, util_plural_str(ncols), use_ncols);
use_ncols = ncols;
}
char max_kmers_str[100];
ulong_to_str(ctx_max_kmers, max_kmers_str);
status("%zu input graph%s, max kmers: %s, using %zu colours",
num_gfiles, util_plural_str(num_gfiles), max_kmers_str, use_ncols);
// If no arguments given we default to removing tips < 2*kmer_size
if(min_keep_tip < 0)
min_keep_tip = 2 * kmer_size;
// Warn if any graph files already cleaned
size_t fromcol;
ErrorCleaning *cleaning;
for(i = 0; i < num_gfiles; i++) {
for(j = 0; j < file_filter_num(&gfiles[i].fltr); j++) {
fromcol = file_filter_fromcol(&gfiles[i].fltr, j);
cleaning = &gfiles[i].hdr.ginfo[fromcol].cleaning;
if(cleaning->cleaned_snodes && unitig_cleaning) {
warn("%s:%zu already has unitig cleaning with threshold: <%zu",
file_filter_path(&gfiles[i].fltr), fromcol,
(size_t)cleaning->clean_snodes_thresh);
}
if(cleaning->cleaned_tips && tip_cleaning) {
warn("%s:%zu already has had tip cleaned",
file_filter_path(&gfiles[i].fltr), fromcol);
}
}
}
// Print steps
size_t step = 0;
status("Actions:\n");
if(covg_before_path != NULL)
status("%zu. Saving kmer coverage distribution to: %s", step++, covg_before_path);
if(len_before_path != NULL)
status("%zu. Saving unitig length distribution to: %s", step++, len_before_path);
if(min_keep_tip > 0)
status("%zu. Cleaning tips shorter than %i nodes", step++, min_keep_tip);
if(unitig_min > 0)
status("%zu. Cleaning unitigs with coverage < %i", step++, unitig_min);
if(unitig_min < 0)
status("%zu. Cleaning unitigs with auto-detected threshold", step++);
if(covg_after_path != NULL)
status("%zu. Saving kmer coverage distribution to: %s", step++, covg_after_path);
if(len_after_path != NULL)
status("%zu. Saving unitig length distribution to: %s", step++, len_after_path);
//
// Decide memory usage
//
bool all_colours_loaded = (ncols <= use_ncols);
bool use_mem_limit = (memargs.mem_to_use_set && num_gfiles > 1) || !ctx_max_kmers;
size_t kmers_in_hash, bits_per_kmer, graph_mem;
size_t per_col_bits = (sizeof(Covg)+sizeof(Edges)) * 8;
size_t extra_edge_bits = (all_colours_loaded ? 0 : sizeof(Edges) * 8);
bits_per_kmer = sizeof(BinaryKmer)*8 +
per_col_bits * use_ncols +
extra_edge_bits;
kmers_in_hash = cmd_get_kmers_in_hash(memargs.mem_to_use,
memargs.mem_to_use_set,
memargs.num_kmers,
memargs.num_kmers_set,
bits_per_kmer,
ctx_max_kmers, ctx_sum_kmers,
use_mem_limit, &graph_mem);
// Maximise the number of colours we load to fill the mem
size_t max_usencols = (memargs.mem_to_use*8 -
sizeof(BinaryKmer)*8*kmers_in_hash +
extra_edge_bits*kmers_in_hash) /
(per_col_bits*kmers_in_hash);
use_ncols = MIN2(max_usencols, ncols);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
//
// Check output files are writable
//
futil_create_output(out_ctx_path);
// Does nothing if arg is NULL
futil_create_output(covg_before_path);
futil_create_output(covg_after_path);
futil_create_output(len_before_path);
futil_create_output(len_after_path);
// Create db_graph
// Load as many colours as possible
// Use an extra set of edge to take intersections
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, use_ncols, use_ncols,
kmers_in_hash, DBG_ALLOC_EDGES | DBG_ALLOC_COVGS);
// Extra edges required to hold union of kept edges
Edges *edges_union = NULL;
if(use_ncols < ncols)
edges_union = ctx_calloc(db_graph.ht.capacity, sizeof(Edges));
// Load graph into a single colour
GraphLoadingPrefs gprefs = graph_loading_prefs(&db_graph);
// Construct cleaned graph header
GraphFileHeader outhdr;
memset(&outhdr, 0, sizeof(GraphFileHeader));
for(i = 0; i < num_gfiles; i++)
graph_file_merge_header(&outhdr, &gfiles[i]);
if(ncols > use_ncols)
{
db_graph.num_of_cols = db_graph.num_edge_cols = 1;
SWAP(edges_union, db_graph.col_edges);
graphs_load_files_flat(gfiles, num_gfiles, gprefs, NULL);
SWAP(edges_union, db_graph.col_edges);
db_graph.num_of_cols = db_graph.num_edge_cols = use_ncols;
}
else {
for(i = 0; i < num_gfiles; i++)
graph_load(&gfiles[i], gprefs, NULL);
}
char num_kmers_str[100];
ulong_to_str(db_graph.ht.num_kmers, num_kmers_str);
status("Total kmers loaded: %s\n", num_kmers_str);
size_t initial_nkmers = db_graph.ht.num_kmers;
hash_table_print_stats(&db_graph.ht);
uint8_t *visited = ctx_calloc(roundup_bits2bytes(db_graph.ht.capacity), 1);
uint8_t *keep = ctx_calloc(roundup_bits2bytes(db_graph.ht.capacity), 1);
// Always estimate cleaning threshold
// if(unitig_min <= 0 || covg_before_path || len_before_path)
// {
// Get coverage distribution and estimate cleaning threshold
int est_min_covg = cleaning_get_threshold(nthreads,
covg_before_path,
len_before_path,
visited, &db_graph);
if(est_min_covg < 0) status("Cannot find recommended cleaning threshold");
else status("Recommended cleaning threshold is: %i", est_min_covg);
// Use estimated threshold if threshold not set
if(unitig_min < 0) {
if(fallback_thresh > 0 && est_min_covg < (int)fallback_thresh) {
status("Using fallback threshold: %i", fallback_thresh);
unitig_min = fallback_thresh;
}
else if(est_min_covg >= 0) unitig_min = est_min_covg;
}
// }
// Die if we failed to find suitable cleaning threshold
if(unitig_min < 0)
die("Need cleaning threshold (--unitigs=<D> or --fallback <D>)");
// Cleaning parameters should now be set (>0) or turned off (==0)
ctx_assert(unitig_min >= 0);
ctx_assert(min_keep_tip >= 0);
if(unitig_min || min_keep_tip)
{
// Clean graph of tips (if min_keep_tip > 0) and unitigs (if threshold > 0)
clean_graph(nthreads, unitig_min, min_keep_tip,
covg_after_path, len_after_path,
visited, keep, &db_graph);
}
ctx_free(visited);
ctx_free(keep);
if(out_ctx_path != NULL)
{
// Set output header ginfo cleaned
for(col = 0; col < ncols; col++)
{
cleaning = &outhdr.ginfo[col].cleaning;
cleaning->cleaned_snodes |= unitig_cleaning;
cleaning->cleaned_tips |= tip_cleaning;
// if(tip_cleaning) {
// strbuf_append_str(&outhdr.ginfo[col].sample_name, ".tipclean");
// }
if(unitig_cleaning) {
size_t thresh = cleaning->clean_snodes_thresh;
thresh = cleaning->cleaned_snodes ? MAX2(thresh, (uint32_t)unitig_min)
: (uint32_t)unitig_min;
cleaning->clean_snodes_thresh = thresh;
// char name_append[200];
// sprintf(name_append, ".supclean%zu", thresh);
// strbuf_append_str(&outhdr.ginfo[col].sample_name, name_append);
}
}
// Print stats on removed kmers
size_t removed_nkmers = initial_nkmers - db_graph.ht.num_kmers;
double removed_pct = (100.0 * removed_nkmers) / initial_nkmers;
char removed_str[100], init_str[100];
ulong_to_str(removed_nkmers, removed_str);
ulong_to_str(initial_nkmers, init_str);
status("Removed %s of %s (%.2f%%) kmers", removed_str, init_str, removed_pct);
// kmers_loaded=true
graph_writer_merge(out_ctx_path, gfiles, num_gfiles,
true, all_colours_loaded,
edges_union, &outhdr, &db_graph);
}
ctx_check(db_graph.ht.num_kmers == hash_table_count_kmers(&db_graph.ht));
// TODO: report kmer coverage for each sample
graph_header_dealloc(&outhdr);
for(i = 0; i < num_gfiles; i++) graph_file_close(&gfiles[i]);
ctx_free(gfiles);
ctx_free(edges_union);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
void read_thread_args_parse(struct ReadThreadCmdArgs *args,
int argc, char **argv,
const struct option *longopts, bool correct_cmd)
{
size_t i;
CorrectAlnInput task = CORRECT_ALN_INPUT_INIT;
uint8_t fq_offset = 0;
GPathReader tmp_gpfile;
CorrectAlnInputBuffer *inputs = &args->inputs;
args->memargs = (struct MemArgs)MEM_ARGS_INIT;
args->fmt = SEQ_FMT_FASTQ;
// Arg parsing
char cmd[100];
char shortopts[300];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int used = 1, c;
char *tmp_path;
// silence error messages from getopt_long
// opterr = 0;
while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) {
cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd));
switch(c) {
case 0: /* flag set */ break;
case 'h': cmd_print_usage(NULL); break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'o': cmd_check(!args->out_ctp_path,cmd); args->out_ctp_path = optarg; break;
case 'p':
memset(&tmp_gpfile, 0, sizeof(GPathReader));
gpath_reader_open(&tmp_gpfile, optarg);
gpfile_buf_push(&args->gpfiles, &tmp_gpfile, 1);
break;
case 't':
cmd_check(!args->nthreads, cmd);
args->nthreads = cmd_uint32_nonzero(cmd, optarg);
break;
case 'm': cmd_mem_args_set_memory(&args->memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&args->memargs, optarg); break;
case 'c': args->colour = cmd_uint32(cmd, optarg); break;
case 'F':
cmd_check(args->fmt == SEQ_FMT_FASTQ, cmd);
args->fmt = cmd_parse_format(cmd, optarg);
break;
case '1':
case '2':
case 'i':
used = 1;
correct_aln_input_buf_push(inputs, &task, 1);
asyncio_task_parse(&inputs->b[inputs->len-1].files, c, optarg,
fq_offset, correct_cmd ? &tmp_path : NULL);
if(correct_cmd) inputs->b[inputs->len-1].out_base = tmp_path;
break;
case 'M':
if(!strcmp(optarg,"FF")) task.matedir = READPAIR_FF;
else if(!strcmp(optarg,"FR")) task.matedir = READPAIR_FR;
else if(!strcmp(optarg,"RF")) task.matedir = READPAIR_RF;
else if(!strcmp(optarg,"RR")) task.matedir = READPAIR_RR;
else die("-M,--matepair <orient> must be one of: FF,FR,RF,RR");
used = 0; break;
case 'O': fq_offset = cmd_uint8(cmd, optarg); used = 0; break;
case 'Q': task.fq_cutoff = cmd_uint8(cmd, optarg); used = 0; break;
case 'H': task.hp_cutoff = cmd_uint8(cmd, optarg); used = 0; break;
case 'l': task.crt_params.frag_len_min = cmd_uint32(cmd, optarg); used = 0; break;
case 'L': task.crt_params.frag_len_max = cmd_uint32(cmd, optarg); used = 0; break;
case 'w': task.crt_params.one_way_gap_traverse = true; used = 0; break;
case 'W': task.crt_params.one_way_gap_traverse = false; used = 0; break;
case 'd': task.crt_params.gap_wiggle = cmd_udouble(cmd, optarg); used = 0; break;
case 'D': task.crt_params.gap_variance = cmd_udouble(cmd, optarg); used = 0; break;
case 'X': task.crt_params.max_context = cmd_uint32(cmd, optarg); used = 0; break;
case 'e': task.crt_params.use_end_check = true; used = 0; break;
case 'E': task.crt_params.use_end_check = false; used = 0; break;
case 'g': cmd_check(!args->dump_seq_sizes, cmd); args->dump_seq_sizes = optarg; break;
case 'G': cmd_check(!args->dump_frag_sizes, cmd); args->dump_frag_sizes = optarg; break;
case 'u': args->use_new_paths = true; break;
case 'x': gen_paths_print_contigs = true; break;
case 'y': gen_paths_print_paths = true; break;
case 'z': gen_paths_print_reads = true; break;
case 'Z':
cmd_check(!args->fq_zero, cmd);
if(strlen(optarg) != 1)
cmd_print_usage("--fq-zero <c> requires a single char");
args->fq_zero = optarg[0];
break;
case 'P': cmd_check(!args->append_orig_seq,cmd); args->append_orig_seq = true; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" thread/correct -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
if(args->nthreads == 0) args->nthreads = DEFAULT_NTHREADS;
// Check that optind+1 == argc
if(optind+1 > argc)
cmd_print_usage("Expected exactly one graph file");
else if(optind+1 < argc)
cmd_print_usage("Expected only one graph file. What is this: '%s'", argv[optind]);
char *graph_path = argv[optind];
status("Reading graph: %s", graph_path);
if(!used) cmd_print_usage("Ignored arguments after last --seq");
// ctx_thread requires output file
if(!correct_cmd && !args->out_ctp_path)
cmd_print_usage("--out <out.ctp> is required");
//
// Open graph graph file
//
GraphFileReader *gfile = &args->gfile;
graph_file_open(gfile, graph_path);
if(!correct_cmd && file_filter_into_ncols(&gfile->fltr) > 1)
die("Please specify a single colour e.g. %s:0", file_filter_path(&gfile->fltr));
//
// Open path files
//
size_t path_max_usedcols = 0;
for(i = 0; i < args->gpfiles.len; i++) {
// file_filter_update_intocol(&args->pfiles.b[i].fltr, 0);
if(!correct_cmd && file_filter_into_ncols(&args->gpfiles.b[i].fltr) > 1) {
die("Please specify a single colour e.g. %s:0",
file_filter_path(&args->gpfiles.b[i].fltr));
}
path_max_usedcols = MAX2(path_max_usedcols,
file_filter_into_ncols(&args->gpfiles.b[i].fltr));
}
args->path_max_usedcols = path_max_usedcols;
// Check for compatibility between graph files and path files
graphs_gpaths_compatible(gfile, 1, args->gpfiles.b, args->gpfiles.len, -1);
// if no paths loaded, set all max_context values to 1, since >1 kmer only
// useful if can pickup paths
if(args->gpfiles.len == 0) {
for(i = 0; i < inputs->len; i++)
inputs->b[i].crt_params.max_context = 1;
}
// Check frag_len_min < frag_len_max
for(i = 0; i < inputs->len; i++)
{
CorrectAlnInput *t = &inputs->b[i];
t->files.ptr = t;
if(t->crt_params.frag_len_min > t->crt_params.frag_len_max) {
die("--min-ins %u is greater than --max-ins %u",
t->crt_params.frag_len_min, t->crt_params.frag_len_max);
}
correct_aln_input_print(&inputs->b[i]);
args->max_gap_limit = MAX2(args->max_gap_limit, t->crt_params.frag_len_max);
}
futil_create_output(args->dump_seq_sizes);
futil_create_output(args->dump_frag_sizes);
}
int ctx_vcfcov(int argc, char **argv)
{
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_path = NULL, *out_type = NULL;
uint32_t max_allele_len = 0, max_gt_vars = 0;
char *ref_path = NULL;
bool low_mem = false;
// Arg parsing
char cmd[100];
char shortopts[300];
cmd_long_opts_to_short(longopts, shortopts, sizeof(shortopts));
int c;
size_t i;
// silence error messages from getopt_long
// opterr = 0;
while((c = getopt_long_only(argc, argv, shortopts, longopts, NULL)) != -1) {
cmd_get_longopt_str(longopts, c, cmd, sizeof(cmd));
switch(c) {
case 0: /* flag set */ break;
case 'h': cmd_print_usage(NULL); break;
case 'o': cmd_check(!out_path, cmd); out_path = optarg; break;
case 'O': cmd_check(!out_type, cmd); out_type = optarg; break;
case 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'm': cmd_mem_args_set_memory(&memargs, optarg); break;
case 'n': cmd_mem_args_set_nkmers(&memargs, optarg); break;
case 'r': cmd_check(!ref_path, cmd); ref_path = optarg; break;
case 'L': cmd_check(!max_allele_len,cmd); max_allele_len = cmd_uint32(cmd,optarg); break;
case 'N': cmd_check(!max_gt_vars,cmd); max_gt_vars = cmd_uint32(cmd,optarg); break;
case 'M': cmd_check(!low_mem, cmd); low_mem = true; break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
// cmd_print_usage(NULL);
die("`"CMD" "SUBCMD" -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
// Defaults for unset values
if(out_path == NULL) out_path = "-";
if(ref_path == NULL) cmd_print_usage("Require a reference (-r,--ref <ref.fa>)");
if(optind+2 > argc) cmd_print_usage("Require VCF and graph files");
if(!max_allele_len) max_allele_len = DEFAULT_MAX_ALLELE_LEN;
if(!max_gt_vars) max_gt_vars = DEFAULT_MAX_GT_VARS;
status("[vcfcov] max allele length: %u; max number of variants: %u",
max_allele_len, max_gt_vars);
// open ref
// index fasta with: samtools faidx ref.fa
faidx_t *fai = fai_load(ref_path);
if(fai == NULL) die("Cannot load ref index: %s / %s.fai", ref_path, ref_path);
// Open input VCF file
const char *vcf_path = argv[optind++];
htsFile *vcffh = hts_open(vcf_path, "r");
if(vcffh == NULL) die("Cannot open VCF file: %s", vcf_path);
bcf_hdr_t *vcfhdr = bcf_hdr_read(vcffh);
if(vcfhdr == NULL) die("Cannot read VCF header: %s", vcf_path);
// Test we can close and reopen files
if(low_mem) {
if((vcffh = hts_open(vcf_path, "r")) == NULL)
die("Cannot re-open VCF file: %s", vcf_path);
if((vcfhdr = bcf_hdr_read(vcffh)) == NULL)
die("Cannot re-read VCF header: %s", vcf_path);
}
//
// Open graph files
//
const size_t num_gfiles = argc - optind;
char **graph_paths = argv + optind;
ctx_assert(num_gfiles > 0);
GraphFileReader *gfiles = ctx_calloc(num_gfiles, sizeof(GraphFileReader));
size_t ncols, ctx_max_kmers = 0, ctx_sum_kmers = 0;
ncols = graph_files_open(graph_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
// Check graph + paths are compatible
graphs_gpaths_compatible(gfiles, num_gfiles, NULL, 0, -1);
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem;
bits_per_kmer = sizeof(BinaryKmer)*8 + sizeof(Covg)*8 * ncols;
kmers_in_hash = cmd_get_kmers_in_hash(memargs.mem_to_use,
memargs.mem_to_use_set,
memargs.num_kmers,
memargs.num_kmers_set,
bits_per_kmer,
low_mem ? -1 : (int64_t)ctx_max_kmers,
ctx_sum_kmers,
true, &graph_mem);
cmd_check_mem_limit(memargs.mem_to_use, graph_mem);
//
// Open output file
//
// v=>vcf, z=>compressed vcf, b=>bcf, bu=>uncompressed bcf
int mode = vcf_misc_get_outtype(out_type, out_path);
futil_create_output(out_path);
htsFile *outfh = hts_open(out_path, modes_htslib[mode]);
status("[vcfcov] Output format: %s", hsmodes_htslib[mode]);
// Allocate memory
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, ncols, 1, kmers_in_hash,
DBG_ALLOC_COVGS);
//
// Set up tag names
//
// *R => ref, *A => alt
sprintf(kcov_ref_tag, "K%zuR", db_graph.kmer_size); // mean coverage
sprintf(kcov_alt_tag, "K%zuA", db_graph.kmer_size);
// #SAMPLE=<ID=...,K29KCOV=...,K29NK=...,K29RLK>
// - K29_kcov is empirical kmer coverage
// - K29_nkmers is the number of kmers in the sample
// - mean_read_length is the mean read length in bases
char sample_kcov_tag[20], sample_nk_tag[20], sample_rlk_tag[20];
sprintf(sample_kcov_tag, "K%zu_kcov", db_graph.kmer_size); // mean coverage
sprintf(sample_nk_tag, "K%zu_nkmers", db_graph.kmer_size);
sprintf(sample_rlk_tag, "mean_read_length");
//
// Load kmers if we are using --low-mem
//
VcfCovStats st;
memset(&st, 0, sizeof(st));
VcfCovPrefs prefs = {.kcov_ref_tag = kcov_ref_tag,
.kcov_alt_tag = kcov_alt_tag,
.max_allele_len = max_allele_len,
.max_gt_vars = max_gt_vars,
.load_kmers_only = false};
if(low_mem)
{
status("[vcfcov] Loading kmers from VCF+ref");
prefs.load_kmers_only = true;
vcfcov_file(vcffh, vcfhdr, NULL, NULL, vcf_path, fai,
NULL, &prefs, &st, &db_graph);
// Close files
hts_close(vcffh);
bcf_hdr_destroy(vcfhdr);
// Re-open files
if((vcffh = hts_open(vcf_path, "r")) == NULL)
die("Cannot re-open VCF file: %s", vcf_path);
if((vcfhdr = bcf_hdr_read(vcffh)) == NULL)
die("Cannot re-read VCF header: %s", vcf_path);
prefs.load_kmers_only = false;
}
//
// Load graphs
//
GraphLoadingStats gstats;
memset(&gstats, 0, sizeof(gstats));
GraphLoadingPrefs gprefs = graph_loading_prefs(&db_graph);
gprefs.must_exist_in_graph = low_mem;
for(i = 0; i < num_gfiles; i++) {
graph_load(&gfiles[i], gprefs, &gstats);
graph_file_close(&gfiles[i]);
}
ctx_free(gfiles);
hash_table_print_stats(&db_graph.ht);
//
// Set up VCF header / graph matchup
//
size_t *samplehdrids = ctx_malloc(db_graph.num_of_cols * sizeof(size_t));
// Add samples to vcf header
bcf_hdr_t *outhdr = bcf_hdr_dup(vcfhdr);
bcf_hrec_t *hrec;
int sid;
char hdrstr[200];
for(i = 0; i < db_graph.num_of_cols; i++) {
char *sname = db_graph.ginfo[i].sample_name.b;
if((sid = bcf_hdr_id2int(outhdr, BCF_DT_SAMPLE, sname)) < 0) {
bcf_hdr_add_sample(outhdr, sname);
sid = bcf_hdr_id2int(outhdr, BCF_DT_SAMPLE, sname);
}
samplehdrids[i] = sid;
// Add SAMPLE field
hrec = bcf_hdr_get_hrec(outhdr, BCF_HL_STR, "ID", sname, "SAMPLE");
if(hrec == NULL) {
sprintf(hdrstr, "##SAMPLE=<ID=%s,%s=%"PRIu64",%s=%"PRIu64",%s=%zu>", sname,
sample_kcov_tag,
gstats.nkmers[i] ? gstats.sumcov[i] / gstats.nkmers[i] : 0,
sample_nk_tag, gstats.nkmers[i],
sample_rlk_tag, (size_t)db_graph.ginfo[i].mean_read_length);
bcf_hdr_append(outhdr, hdrstr);
}
else {
// mean kcovg
sprintf(hdrstr, "%"PRIu64, gstats.sumcov[i] / gstats.nkmers[i]);
vcf_misc_add_update_hrec(hrec, sample_kcov_tag, hdrstr);
// num kmers
sprintf(hdrstr, "%"PRIu64, gstats.nkmers[i]);
vcf_misc_add_update_hrec(hrec, sample_nk_tag, hdrstr);
// mean read length in kmers
sprintf(hdrstr, "%zu", (size_t)db_graph.ginfo[i].mean_read_length);
vcf_misc_add_update_hrec(hrec, sample_rlk_tag, hdrstr);
}
status("[vcfcov] Colour %zu: %s [VCF column %zu]", i, sname, samplehdrids[i]);
}
// Add genotype format fields
// One field per alternative allele
sprintf(hdrstr, "##FORMAT=<ID=%s,Number=A,Type=Integer,"
"Description=\"Coverage on ref (k=%zu): sum(kmer_covs) / exp_num_kmers\">\n",
kcov_ref_tag, db_graph.kmer_size);
bcf_hdr_append(outhdr, hdrstr);
sprintf(hdrstr, "##FORMAT=<ID=%s,Number=A,Type=Integer,"
"Description=\"Coverage on alt (k=%zu): sum(kmer_covs) / exp_num_kmers\">\n",
kcov_alt_tag, db_graph.kmer_size);
bcf_hdr_append(outhdr, hdrstr);
bcf_hdr_set_version(outhdr, "VCFv4.2");
// Add command string to header
vcf_misc_hdr_add_cmd(outhdr, cmd_get_cmdline(), cmd_get_cwd());
if(bcf_hdr_write(outfh, outhdr) != 0)
die("Cannot write header to: %s", futil_outpath_str(out_path));
status("[vcfcov] Reading %s and adding coverage", vcf_path);
// Reset stats and get coverage
memset(&st, 0, sizeof(st));
vcfcov_file(vcffh, vcfhdr, outfh, outhdr, vcf_path, fai,
samplehdrids, &prefs, &st, &db_graph);
// Print statistics
char ns0[50], ns1[50];
status("[vcfcov] Read %s VCF lines", ulong_to_str(st.nvcf_lines, ns0));
status("[vcfcov] Read %s ALTs", ulong_to_str(st.nalts_read, ns0));
status("[vcfcov] Used %s kmers", ulong_to_str(st.ngt_kmers, ns0));
status("[vcfcov] ALTs used: %s / %s (%.2f%%)",
ulong_to_str(st.nalts_loaded, ns0), ulong_to_str(st.nalts_read, ns1),
st.nalts_read ? (100.0*st.nalts_loaded) / st.nalts_read : 0.0);
status("[vcfcov] ALTs too long (>%ubp): %s / %s (%.2f%%)", max_allele_len,
ulong_to_str(st.nalts_too_long, ns0), ulong_to_str(st.nalts_read, ns1),
st.nalts_read ? (100.0*st.nalts_too_long) / st.nalts_read : 0.0);
status("[vcfcov] ALTs too dense (>%u within %zubp): %s / %s (%.2f%%)",
max_gt_vars, db_graph.kmer_size,
ulong_to_str(st.nalts_no_covg, ns0), ulong_to_str(st.nalts_read, ns1),
st.nalts_read ? (100.0*st.nalts_no_covg) / st.nalts_read : 0.0);
status("[vcfcov] ALTs printed with coverage: %s / %s (%.2f%%)",
ulong_to_str(st.nalts_with_covg, ns0), ulong_to_str(st.nalts_read, ns1),
st.nalts_read ? (100.0*st.nalts_with_covg) / st.nalts_read : 0.0);
status("[vcfcov] Saved to: %s\n", out_path);
ctx_free(samplehdrids);
graph_loading_stats_destroy(&gstats);
bcf_hdr_destroy(vcfhdr);
bcf_hdr_destroy(outhdr);
hts_close(vcffh);
hts_close(outfh);
fai_destroy(fai);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
