void _construct_graph_with_paths(dBGraph *graph,
size_t kmer_size, size_t ncols,
char **seqs, size_t nseqs,
CorrectAlnParam path_params)
{
size_t i;
db_graph_alloc(graph, kmer_size, ncols, ncols, 1024);
// Graph data
graph->bktlocks = ctx_calloc(roundup_bits2bytes(graph->ht.num_of_buckets), 1);
graph->col_edges = ctx_calloc(graph->ht.capacity * ncols, sizeof(Edges));
graph->col_covgs = ctx_calloc(graph->ht.capacity * ncols, sizeof(Covg));
graph->node_in_cols = ctx_calloc(roundup_bits2bytes(graph->ht.capacity) * ncols, 1);
// Path data
path_store_alloc(&graph->pstore, 1024, true, graph->ht.capacity, ncols);
graph->pstore.kmer_locks = ctx_calloc(roundup_bits2bytes(graph->ht.capacity), 1);
// Build graph
for(i = 0; i < nseqs; i++)
build_graph_from_str_mt(graph, 0, seqs[i], strlen(seqs[i]));
graph->num_of_cols_used = MAX2(graph->num_of_cols_used, 1);
GenPathWorker *gen_path_wrkr = gen_paths_workers_alloc(1, graph, NULL);
for(i = 0; i < nseqs; i++)
gen_paths_from_str_mt(gen_path_wrkr, seqs[i], path_params);
gen_paths_workers_dealloc(gen_path_wrkr, 1);
}
/**
* Get coverage threshold for removing unitigs
*
* @param visited should be at least db_graph.ht.capcity bits long and initialised
* to zero. On return, it will be 1 at each original kmer index
* @param covgs_csv_path
* @param lens_csv_path paths to files to write CSV histogram of unitigs
coverages and lengths BEFORE ANY CLEANING.
* If NULL these are ignored.
* @return threshold to clean or -1 on error
*/
int cleaning_get_threshold(size_t num_threads,
const char *covgs_csv_path,
const char *lens_csv_path,
uint8_t *visited,
const dBGraph *db_graph)
{
// Estimate optimum cleaning threshold
status("[cleaning] Calculating unitig stats with %zu threads...", num_threads);
status("[cleaning] Using kmer gamma method");
// Get kmer coverages and unitig lengths
UnitigCleaner cl;
unitig_cleaner_alloc(&cl, num_threads, 0, 0, NULL, db_graph);
supernodes_iterate(num_threads, visited, db_graph, unitig_get_covg, &cl);
// Get kmer coverage only (faster)
// KmerCleanerIterator kcls[nthreads];
// for(i = 0; i < nthreads; i++)
// kcls[i] = (KmerCleanerIterator){.threadid = i, .nthreads = nthreads, .cl = &cl};
// util_run_threads(kcls, nthreads, sizeof(kcls[0]), nthreads, kmer_get_covg);
// Wipe visited kmer memory
memset(visited, 0, roundup_bits2bytes(db_graph->ht.capacity));
if(covgs_csv_path != NULL) {
cleaning_write_covg_histogram(covgs_csv_path,
cl.kmer_covgs_init,
cl.unitig_covgs_init,
cl.covg_arrsize);
}
if(lens_csv_path != NULL) {
cleaning_write_len_histogram(lens_csv_path,
cl.len_hist_init,
cl.len_arrsize,
db_graph->kmer_size);
}
// set threshold using histogram and genome size
double alpha = 0, beta = 0, false_pos = 0, false_neg = 0;
int threshold_est = cleaning_pick_kmer_threshold(cl.kmer_covgs_init,
cl.covg_arrsize,
&alpha, &beta,
&false_pos, &false_neg);
if(threshold_est < 0)
warn("Cannot pick a cleaning threshold");
else {
status("[cleaning] alpha=%f, beta=%f FP=%f FN=%f",
alpha, beta, false_pos, false_neg);
status("[cleaning] Recommended unitig cleaning threshold: < %i",
threshold_est);
}
unitig_cleaner_dealloc(&cl);
return threshold_est;
}
// Does not allocate ugraph
void unitig_printer_init(UnitigPrinter *printer, const dBGraph *db_graph,
size_t nthreads, UnitigSyntax syntax, FILE *fout)
{
memset(printer, 0, sizeof(UnitigPrinter));
printer->db_graph = db_graph;
printer->syntax = syntax;
printer->fout = fout;
printer->nthreads = nthreads;
printer->num_unitigs = 0;
printer->visited = ctx_calloc(roundup_bits2bytes(db_graph->ht.capacity), 1);
if(pthread_mutex_init(&printer->outlock, NULL) != 0) die("Mutex init failed");
}
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;
}
int ctx_correct(int argc, char **argv)
{
size_t i, j;
struct ReadThreadCmdArgs args = READ_THREAD_CMD_ARGS_INIT;
read_thread_args_alloc(&args);
read_thread_args_parse(&args, argc, argv, longopts, true);
GraphFileReader *gfile = &args.gfile;
PathFileBuffer *pfiles = &args.pfiles;
CorrectAlnInputBuffer *inputs = &args.inputs;
size_t ctx_total_cols = gfile->hdr.num_of_cols;
size_t ctx_num_kmers = gfile->num_of_kmers;
if(args.colour > ctx_total_cols)
cmd_print_usage("-c %zu is too big [> %zu]", args.colour, ctx_total_cols);
size_t ctp_usedcols = 0;
for(i = 0; i < pfiles->len; i++) {
if(!file_filter_iscolloaded(&pfiles->data[i].fltr, args.colour)) {
cmd_print_usage("Path file doesn't load into colour %zu: %s",
args.colour, pfiles->data[i].fltr.orig_path.buff);
}
ctp_usedcols = MAX2(ctp_usedcols, path_file_usedcols(&pfiles->data[i]));
}
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem, path_mem, total_mem;
// 1 bit needed per kmer if we need to keep track of noreseed
bits_per_kmer = sizeof(Edges)*8 + ctx_num_kmers + sizeof(uint64_t)*8;
kmers_in_hash = cmd_get_kmers_in_hash2(args.memargs.mem_to_use,
args.memargs.mem_to_use_set,
args.memargs.num_kmers,
args.memargs.num_kmers_set,
bits_per_kmer,
ctx_num_kmers, ctx_num_kmers,
false, &graph_mem);
// Paths memory
path_mem = path_files_mem_required(pfiles->data, pfiles->len, false, false,
ctp_usedcols, 0);
cmd_print_mem(path_mem, "paths");
// Total memory
total_mem = graph_mem + path_mem;
cmd_check_mem_limit(args.memargs.mem_to_use, total_mem);
//
// Check we can read all output files
//
// Open output files
SeqOutput *outputs = ctx_calloc(inputs->len, sizeof(SeqOutput));
bool output_files_exist = false;
for(i = 0; i < inputs->len; i++)
{
CorrectAlnInput *input = &inputs->data[i];
input->crt_params.ctxcol = input->crt_params.ctpcol = args.colour;
SeqOutput *output = &outputs[i];
seq_output_alloc(output);
seq_output_set_paths(output, input->out_base,
async_task_pe_output(&input->files));
input->output = output;
// output check prints warnings and returns true if errors
output_files_exist |= seq_output_files_exist_check(output);
}
// Abandon if some of the output files already exist
if(output_files_exist) die("Output files already exist");
// Attempt to open all files
for(i = 0; i < inputs->len && seq_output_open(&outputs[i]); i++) {}
// Check if something went wrong - if so remove all output files
if(i < inputs->len) {
for(j = 0; j < i; j++) seq_output_delete(&outputs[i]);
die("Couldn't open output files");
}
//
// Allocate memory
//
dBGraph db_graph;
db_graph_alloc(&db_graph, gfile->hdr.kmer_size, ctx_total_cols, 1, kmers_in_hash);
size_t bytes_per_col = roundup_bits2bytes(db_graph.ht.capacity);
db_graph.col_edges = ctx_calloc(db_graph.ht.capacity, sizeof(Edges));
db_graph.node_in_cols = ctx_calloc(bytes_per_col * ctx_total_cols, 1);
// Paths
path_store_alloc(&db_graph.pstore, path_mem, false,
db_graph.ht.capacity, ctp_usedcols);
//
// Load Graph and Path files
//
LoadingStats gstats = 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 = true};
// Load graph, print stats, close file
graph_load(gfile, gprefs, &gstats);
hash_table_print_stats_brief(&db_graph.ht);
graph_file_close(gfile);
// Load path files (does nothing if num_fpiles == 0)
paths_format_merge(pfiles->data, pfiles->len, false, false,
args.num_of_threads, &db_graph);
//
// Run alignment
//
correct_reads(args.num_of_threads, MAX_IO_THREADS,
inputs->data, inputs->len,
&db_graph);
// Close and free output files
for(i = 0; i < inputs->len; i++) seq_output_dealloc(&outputs[i]);
ctx_free(outputs);
read_thread_args_dealloc(&args);
db_graph_dealloc(&db_graph);
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;
}
/**
* Remove unitigs with coverage < `covg_threshold` and tips shorter than
* `min_keep_tip`.
*
* @param num_threads Number of threads to use
* @param covg_threshold Remove unitigs with mean covg < `covg_threshold`.
* Ignored if 0.
* @param min_keep_tip Remove tips with length < `min_keep_tip`. Ignored if 0.
* @param covgs_csv_path Path to write CSV of kmer coverage histogram
* @param lens_csv_path Path to write CSV of unitig length histogram
*
* `visited`, `keep` should each be at least db_graph.ht.capcity bits long
* and initialised to zero. On return,
* `visited` will be 1 at each original kmer index
* `keep` will be 1 at each retained kmer index
**/
void clean_graph(size_t num_threads,
size_t covg_threshold, size_t min_keep_tip,
const char *covgs_csv_path, const char *lens_csv_path,
uint8_t *visited, uint8_t *keep, dBGraph *db_graph)
{
ctx_assert(db_graph->num_of_cols == 1);
ctx_assert(db_graph->num_edge_cols > 0);
size_t init_nkmers = db_graph->ht.num_kmers;
if(db_graph->ht.num_kmers == 0) return;
if(covg_threshold == 0 && min_keep_tip == 0) {
warn("[cleaning] No cleaning specified");
return;
}
if(covg_threshold > 0) {
status("[cleaning] Removing unitigs with coverage < %zu...", covg_threshold);
status("[cleaning] Using kmer gamma method");
}
if(min_keep_tip > 0)
status("[cleaning] Removing tips shorter than %zu...", min_keep_tip);
status("[cleaning] using %zu threads", num_threads);
// Mark nodes to keep
UnitigCleaner cl;
unitig_cleaner_alloc(&cl, num_threads, covg_threshold,
min_keep_tip, keep, db_graph);
supernodes_iterate(num_threads, visited, db_graph, unitig_mark, &cl);
// Print numbers of kmers that are being removed
char num_snodes_str[50], num_tips_str[50], num_tip_snodes_str[50];
char num_snode_kmers_str[50], num_tip_kmers_str[50], num_tip_snode_kmers_str[50];
ulong_to_str(cl.num_low_covg_snodes, num_snodes_str);
ulong_to_str(cl.num_tips, num_tips_str);
ulong_to_str(cl.num_tip_and_low_snodes, num_tip_snodes_str);
ulong_to_str(cl.num_low_covg_snode_kmers, num_snode_kmers_str);
ulong_to_str(cl.num_tip_kmers, num_tip_kmers_str);
ulong_to_str(cl.num_tip_and_low_snode_kmers, num_tip_snode_kmers_str);
status("[cleaning] Removing %s low coverage unitigs [%s kmer%s], "
"%s unitig tips [%s kmer%s] "
"and %s of both [%s kmer%s]",
num_snodes_str,
num_snode_kmers_str, util_plural_str(cl.num_low_covg_snode_kmers),
num_tips_str,
num_tip_kmers_str, util_plural_str(cl.num_tip_kmers),
num_tip_snodes_str,
num_tip_snode_kmers_str, util_plural_str(cl.num_tip_and_low_snode_kmers));
// Remove nodes not marked to keep
prune_nodes_lacking_flag(num_threads, keep, db_graph);
// Wipe memory
memset(visited, 0, roundup_bits2bytes(db_graph->ht.capacity));
memset(keep, 0, roundup_bits2bytes(db_graph->ht.capacity));
// Print status update
char remain_nkmers_str[100], removed_nkmers_str[100];
size_t remain_nkmers = db_graph->ht.num_kmers;
size_t removed_nkmers = init_nkmers - remain_nkmers;
ulong_to_str(remain_nkmers, remain_nkmers_str);
ulong_to_str(removed_nkmers, removed_nkmers_str);
status("[cleaning] Remaining kmers: %s removed: %s (%.1f%%)",
remain_nkmers_str, removed_nkmers_str,
(100.0*removed_nkmers)/init_nkmers);
if(covgs_csv_path != NULL) {
cleaning_write_covg_histogram(covgs_csv_path,
cl.kmer_covgs_clean,
cl.unitig_covg_clean,
cl.covg_arrsize);
}
if(lens_csv_path != NULL) {
cleaning_write_len_histogram(lens_csv_path,
cl.len_hist_clean,
cl.len_arrsize,
db_graph->kmer_size);
}
unitig_cleaner_dealloc(&cl);
}
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;
}
/**
* Get coverage threshold for removing supernodes
*
* @param visited should be at least db_graph.ht.capcity bits long and initialised
* to zero. On return, it will be 1 at each original kmer index
* @param covgs_csv_path
* @param lens_csv_path paths to files to write CSV histogram of supernodes
coverages and lengths BEFORE ANY CLEANING.
* If NULL these are ignored.
* @return threshold to clean or -1 on error
*/
int cleaning_get_threshold(size_t num_threads,
const char *covgs_csv_path,
const char *lens_csv_path,
uint8_t *visited,
const dBGraph *db_graph)
{
// Estimate optimum cleaning threshold
status("[cleaning] Calculating supernode stats with %zu threads...", num_threads);
status("[cleaning] Using kmer gamma method");
// Get kmer coverages and supernode lengths
SupernodeCleaner cl;
supernode_cleaner_alloc(&cl, num_threads, 0, 0, NULL, db_graph);
supernodes_iterate(num_threads, visited, db_graph, supernode_get_covg, &cl);
// Get kmer coverage only (faster)
// KmerCleanerIterator kcls[nthreads];
// for(i = 0; i < nthreads; i++)
// kcls[i] = (KmerCleanerIterator){.threadid = i, .nthreads = nthreads, .cl = &cl};
// util_run_threads(kcls, nthreads, sizeof(kcls[0]), nthreads, kmer_get_covg);
// Wipe visited kmer memory
memset(visited, 0, roundup_bits2bytes(db_graph->ht.capacity));
if(covgs_csv_path != NULL) {
cleaning_write_covg_histogram(covgs_csv_path,
cl.covg_hist_init,
cl.mean_covg_hist_init,
cl.covg_arrsize);
}
if(lens_csv_path != NULL) {
cleaning_write_len_histogram(lens_csv_path,
cl.len_hist_init,
cl.len_arrsize,
db_graph->kmer_size);
}
// set threshold using histogram and genome size
int threshold_est = -1;
double fdr = 0.001, alpha = 0, beta = 0;
while(fdr < 1) {
threshold_est = cleaning_pick_kmer_threshold(cl.covg_hist_init,
cl.covg_arrsize,
fdr, &alpha, &beta);
if(threshold_est >= 0) break;
fdr *= 10;
}
if(threshold_est < 0)
warn("Cannot pick a cleaning threshold");
else
status("[cleaning] FDR set to %f [alpha=%f, beta=%f]", fdr, alpha, beta);
if(threshold_est >= 0) {
status("[cleaning] Recommended supernode cleaning threshold: < %i",
threshold_est);
}
supernode_cleaner_dealloc(&cl);
return threshold_est;
}
int ctx_contigs(int argc, char **argv)
{
size_t nthreads = 0;
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_path = NULL;
size_t i, contig_limit = 0, colour = 0;
bool cmd_reseed = false, cmd_no_reseed = false; // -r, -R
const char *conf_table_path = NULL; // save confidence table to here
bool use_missing_info_check = true, seed_with_unused_paths = false;
double min_step_confid = -1.0, min_cumul_confid = -1.0; // < 0 => no min
// Read length and expected depth for calculating confidences
size_t genome_size = 0;
seq_file_t *tmp_seed_file = NULL;
SeqFilePtrBuffer seed_buf;
seq_file_ptr_buf_alloc(&seed_buf, 16);
GPathReader tmp_gpfile;
GPathFileBuffer gpfiles;
gpfile_buf_alloc(&gpfiles, 8);
// Arg parsing
char cmd[100], 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 'f': cmd_check(!futil_get_force(), cmd); futil_set_force(true); break;
case 'o': cmd_check(!out_path,cmd); out_path = optarg; 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 'p':
memset(&tmp_gpfile, 0, sizeof(GPathReader));
gpath_reader_open(&tmp_gpfile, optarg);
gpfile_buf_push(&gpfiles, &tmp_gpfile, 1);
break;
case '1':
case 's': // --seed <in.fa>
if((tmp_seed_file = seq_open(optarg)) == NULL)
die("Cannot read --seed file: %s", optarg);
seq_file_ptr_buf_add(&seed_buf, tmp_seed_file);
break;
case 'r': cmd_check(!cmd_reseed,cmd); cmd_reseed = true; break;
case 'R': cmd_check(!cmd_no_reseed,cmd); cmd_no_reseed = true; break;
case 'N':
cmd_check(!contig_limit,cmd);
contig_limit = cmd_uint32_nonzero(cmd, optarg);
break;
case 'c': cmd_check(!colour,cmd); colour = cmd_uint32(cmd, optarg); break;
case 'G': cmd_check(!genome_size,cmd); genome_size = cmd_bases(cmd, optarg); break;
case 'S': cmd_check(!conf_table_path,cmd); conf_table_path = optarg; break;
case 'M': cmd_check(use_missing_info_check,cmd); use_missing_info_check = false; break;
case 'P': cmd_check(!seed_with_unused_paths,cmd); seed_with_unused_paths = true; break;
case 'C':
cmd_check(min_cumul_confid < 0,cmd);
min_cumul_confid = cmd_udouble(cmd,optarg);
if(min_cumul_confid > 1) die("%s must be 0 <= x <= 1", cmd);
break;
case 'T':
cmd_check(min_step_confid < 0,cmd);
min_step_confid = cmd_udouble(cmd,optarg);
if(min_step_confid > 1) die("%s must be 0 <= x <= 1", cmd);
break;
case ':': /* BADARG */
case '?': /* BADCH getopt_long has already printed error */
die("`"CMD" contigs -h` for help. Bad option: %s", argv[optind-1]);
default: abort();
}
}
if(cmd_no_reseed && cmd_reseed)
cmd_print_usage("Cannot specify both -r and -R");
if(contig_limit && seed_with_unused_paths)
cmd_print_usage("Cannot combine --ncontigs with --use-seed-paths");
bool sample_with_replacement = cmd_reseed;
// Defaults
if(nthreads == 0) nthreads = DEFAULT_NTHREADS;
if(!seed_buf.len && !contig_limit && sample_with_replacement) {
cmd_print_usage("Please specify one or more of: "
"--no-reseed | --ncontigs | --seed <in.fa>");
}
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 ncols, ctx_max_kmers = 0, ctx_sum_kmers = 0;
graph_files_open(graph_paths, gfiles, num_gfiles,
&ctx_max_kmers, &ctx_sum_kmers);
// char *ctx_path = argv[optind];
//
// Open Graph file
//
// GraphFileReader gfile;
// memset(&gfile, 0, sizeof(GraphFileReader));
// graph_file_open(&gfile, ctx_path);
// Update colours in graph file - sample in 0, all others in 1
// never need more than two colours
ncols = gpath_load_sample_pop(gfiles, num_gfiles,
gpfiles.b, gpfiles.len, colour);
// Check for compatibility between graph files and path files
// pop_colour is colour 1
graphs_gpaths_compatible(gfiles, num_gfiles, gpfiles.b, gpfiles.len, 1);
if(!genome_size)
{
char nk_str[50];
if(ctx_max_kmers <= 0) die("Please pass --genome <G> if streaming");
genome_size = ctx_max_kmers;
ulong_to_str(genome_size, nk_str);
status("Taking number of kmers as genome size: %s", nk_str);
}
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem, path_mem, total_mem;
// 1 bit needed per kmer if we need to keep track of kmer usage
bits_per_kmer = sizeof(BinaryKmer)*8 + sizeof(Edges)*8 + sizeof(GPath*)*8 +
ncols + !sample_with_replacement;
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);
// Paths memory
size_t rem_mem = memargs.mem_to_use - MIN2(memargs.mem_to_use, graph_mem);
path_mem = gpath_reader_mem_req(gpfiles.b, gpfiles.len, ncols, rem_mem, false);
// Shift path store memory from graphs->paths
graph_mem -= sizeof(GPath*)*kmers_in_hash;
path_mem += sizeof(GPath*)*kmers_in_hash;
cmd_print_mem(path_mem, "paths");
// Total memory
total_mem = graph_mem + path_mem;
cmd_check_mem_limit(memargs.mem_to_use, total_mem);
// Load contig hist distribution from ctp files
ZeroSizeBuffer contig_hist;
memset(&contig_hist, 0, sizeof(contig_hist));
for(i = 0; i < gpfiles.len; i++) {
gpath_reader_load_contig_hist(gpfiles.b[i].json,
gpfiles.b[i].fltr.path.b,
file_filter_fromcol(&gpfiles.b[i].fltr, 0),
&contig_hist);
}
// Calculate confidences, only for one colour
ContigConfidenceTable conf_table;
conf_table_alloc(&conf_table, 1);
conf_table_update_hist(&conf_table, 0, genome_size,
contig_hist.b, contig_hist.len);
if(conf_table_path != NULL) {
conf_table_save(&conf_table, conf_table_path);
}
zsize_buf_dealloc(&contig_hist);
//
// Output file if printing
//
FILE *fout = out_path ? futil_fopen_create(out_path, "w") : NULL;
// Allocate
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, ncols, 1, kmers_in_hash,
DBG_ALLOC_EDGES | DBG_ALLOC_NODE_IN_COL);
// Paths
gpath_reader_alloc_gpstore(gpfiles.b, gpfiles.len, path_mem,
false, &db_graph);
uint8_t *visited = NULL;
if(!sample_with_replacement)
visited = ctx_calloc(roundup_bits2bytes(db_graph.ht.capacity), 1);
// Load graph
LoadingStats stats = LOAD_STATS_INIT_MACRO;
GraphLoadingPrefs gprefs = {.db_graph = &db_graph,
.boolean_covgs = false,
.must_exist_in_graph = false,
.empty_colours = true};
for(i = 0; i < num_gfiles; i++) {
graph_load(&gfiles[i], gprefs, &stats);
graph_file_close(&gfiles[i]);
gprefs.empty_colours = false;
}
ctx_free(gfiles);
hash_table_print_stats(&db_graph.ht);
// Load path files
for(i = 0; i < gpfiles.len; i++) {
gpath_reader_load(&gpfiles.b[i], GPATH_DIE_MISSING_KMERS, &db_graph);
gpath_reader_close(&gpfiles.b[i]);
}
gpfile_buf_dealloc(&gpfiles);
AssembleContigStats assem_stats;
assemble_contigs_stats_init(&assem_stats);
assemble_contigs(nthreads, seed_buf.b, seed_buf.len,
contig_limit, visited,
use_missing_info_check, seed_with_unused_paths,
min_step_confid, min_cumul_confid,
fout, out_path, &assem_stats, &conf_table,
&db_graph, 0); // Sample always loaded into colour zero
if(fout && fout != stdout) fclose(fout);
assemble_contigs_stats_print(&assem_stats);
assemble_contigs_stats_destroy(&assem_stats);
conf_table_dealloc(&conf_table);
for(i = 0; i < seed_buf.len; i++)
seq_close(seed_buf.b[i]);
seq_file_ptr_buf_dealloc(&seed_buf);
ctx_free(visited);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
int ctx_bubbles(int argc, char **argv)
{
size_t nthreads = 0;
struct MemArgs memargs = MEM_ARGS_INIT;
const char *out_path = NULL;
size_t max_allele_len = 0, max_flank_len = 0;
bool remove_serial_bubbles = true;
// List of haploid colours
size_t *hapcols = NULL;
int nhapcols = 0;
char *hapcols_arg = NULL;
GPathReader tmp_gpfile;
GPathFileBuffer gpfiles;
gpfile_buf_alloc(&gpfiles, 8);
// 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 'p':
memset(&tmp_gpfile, 0, sizeof(GPathReader));
gpath_reader_open(&tmp_gpfile, optarg);
gpfile_buf_push(&gpfiles, &tmp_gpfile, 1);
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 'H': cmd_check(!hapcols_arg, cmd); hapcols_arg = optarg; break;
case 'A': cmd_check(!max_allele_len, cmd); max_allele_len = cmd_uint32_nonzero(cmd, optarg); break;
case 'F': cmd_check(!max_flank_len, cmd); max_flank_len = cmd_uint32_nonzero(cmd, optarg); break;
case 'S': cmd_check(remove_serial_bubbles,cmd); remove_serial_bubbles = false; 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(nthreads == 0) nthreads = DEFAULT_NTHREADS;
if(max_allele_len == 0) max_allele_len = DEFAULT_MAX_ALLELE;
if(max_flank_len == 0) max_flank_len = DEFAULT_MAX_FLANK;
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);
// Check graph + paths are compatible
graphs_gpaths_compatible(gfiles, num_gfiles, gpfiles.b, gpfiles.len, -1);
//
// Check haploid colours are valid
//
if(hapcols_arg != NULL) {
if((nhapcols = range_get_num(hapcols_arg, ncols)) < 0)
die("Invalid haploid colour list: %s", hapcols_arg);
hapcols = ctx_calloc(nhapcols, sizeof(hapcols[0]));
if(range_parse_array(hapcols_arg, hapcols, ncols) < 0)
die("Invalid haploid colour list: %s", hapcols_arg);
}
//
// Decide on memory
//
size_t bits_per_kmer, kmers_in_hash, graph_mem, path_mem, thread_mem;
char thread_mem_str[100];
// edges(1bytes) + kmer_paths(8bytes) + in_colour(1bit/col) +
// visitedfw/rv(2bits/thread)
bits_per_kmer = sizeof(BinaryKmer)*8 + sizeof(Edges)*8 +
(gpfiles.len > 0 ? sizeof(GPath*)*8 : 0) +
ncols + 2*nthreads;
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);
// Thread memory
thread_mem = roundup_bits2bytes(kmers_in_hash) * 2;
bytes_to_str(thread_mem * nthreads, 1, thread_mem_str);
status("[memory] (of which threads: %zu x %zu = %s)\n",
nthreads, thread_mem, thread_mem_str);
// Paths memory
size_t rem_mem = memargs.mem_to_use - MIN2(memargs.mem_to_use, graph_mem+thread_mem);
path_mem = gpath_reader_mem_req(gpfiles.b, gpfiles.len, ncols, rem_mem, false,
kmers_in_hash, false);
// Shift path store memory from graphs->paths
graph_mem -= sizeof(GPath*)*kmers_in_hash;
path_mem += sizeof(GPath*)*kmers_in_hash;
cmd_print_mem(path_mem, "paths");
size_t total_mem = graph_mem + thread_mem + path_mem;
cmd_check_mem_limit(memargs.mem_to_use, total_mem);
//
// Open output file
//
gzFile gzout = futil_gzopen_create(out_path, "w");
// Allocate memory
dBGraph db_graph;
db_graph_alloc(&db_graph, gfiles[0].hdr.kmer_size, ncols, 1, kmers_in_hash,
DBG_ALLOC_EDGES | DBG_ALLOC_NODE_IN_COL);
// Paths
gpath_reader_alloc_gpstore(gpfiles.b, gpfiles.len, path_mem, false, &db_graph);
//
// 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);
// Load link files
for(i = 0; i < gpfiles.len; i++)
gpath_reader_load(&gpfiles.b[i], GPATH_DIE_MISSING_KMERS, &db_graph);
// Create array of cJSON** from input files
cJSON **hdrs = ctx_malloc(gpfiles.len * sizeof(cJSON*));
for(i = 0; i < gpfiles.len; i++) hdrs[i] = gpfiles.b[i].json;
// Now call variants
BubbleCallingPrefs call_prefs = {.max_allele_len = max_allele_len,
.max_flank_len = max_flank_len,
.haploid_cols = hapcols,
.nhaploid_cols = nhapcols,
.remove_serial_bubbles = remove_serial_bubbles};
invoke_bubble_caller(nthreads, &call_prefs,
gzout, out_path,
hdrs, gpfiles.len,
&db_graph);
status(" saved to: %s\n", out_path);
gzclose(gzout);
ctx_free(hdrs);
// Close input link files
for(i = 0; i < gpfiles.len; i++)
gpath_reader_close(&gpfiles.b[i]);
gpfile_buf_dealloc(&gpfiles);
ctx_free(hapcols);
db_graph_dealloc(&db_graph);
return EXIT_SUCCESS;
}
