static char const rcsid[] = "$Id: blastclust.c,v 6.49 2007/12/21 14:00:47 madden Exp $"; /* $RCSfile: blastclust.c,v $ $Revision: 6.49 $ $Date: 2007/12/21 14:00:47 $ * =========================================================================== * * PUBLIC DOMAIN NOTICE * National Center for Biotechnology Information * * This software/database is a "United States Government Work" under the * terms of the United States Copyright Act. It was written as part of * the author's official duties as a United States Government employee and * thus cannot be copyrighted. This software/database is freely available * to the public for use. The National Library of Medicine and the U.S. * Government have not placed any restriction on its use or reproduction. * * Although all reasonable efforts have been taken to ensure the accuracy * and reliability of the software and data, the NLM and the U.S. * Government do not and cannot warrant the performance or results that * may be obtained by using this software or data. The NLM and the U.S. * Government disclaim all warranties, express or implied, including * warranties of performance, merchantability or fitness for any particular * purpose. * * Please cite the author in any work or product based on this material. * * =========================================================================== * * Author: Ilya Dondoshansky * * File Description: * Program to perform a single-linkage clustering of a set of protein or DNA * sequences. See file README.bcl for description * * --------------------------------------------------------------------------- * $Log: blastclust.c,v $ * Revision 6.49 2007/12/21 14:00:47 madden * Exit if query cannot be processed, JIRA SB-32 * * Revision 6.48 2005/08/19 20:12:32 dondosha * Added extensive doxygen-style comments * * Revision 6.47 2005/07/28 16:16:46 coulouri * remove dead code * * Revision 6.46 2005/07/27 15:52:02 coulouri * remove unused queue_callback * * Revision 6.45 2004/08/02 20:08:49 dondosha * Write and read header integers separately instead of as a structure; changed first number from 1 byte to 4, to prevent padding with garbage * * Revision 6.44 2004/06/30 21:03:57 madden * Add include for blfmtutl.h * * Revision 6.43 2004/05/03 19:59:48 dondosha * Display version number when run with no arguments * * Revision 6.42 2004/02/24 15:18:15 dondosha * Fixed several memory leaks for nucleotide clustering * * Revision 6.41 2004/02/18 20:42:59 dondosha * Cleaned up how id strings are retrieved for output * * Revision 6.40 2004/02/18 15:18:46 dondosha * Minor fix when freeing title strings * * Revision 6.39 2003/07/22 17:59:48 dondosha * Added call to BlastErrorPrint to make warnings more informative * * Revision 6.38 2003/06/13 19:59:29 dondosha * Fixed 2 purify errors * * Revision 6.37 2003/05/30 17:31:09 coulouri * add rcsid * * Revision 6.36 2003/05/13 16:02:42 coulouri * make ErrPostEx(SEV_FATAL, ...) exit with nonzero status * * Revision 6.35 2003/05/06 20:21:13 dondosha * Typo correction * * Revision 6.34 2003/05/06 20:19:45 dondosha * Changed a confusing error message to a meaningful one * * Revision 6.33 2002/10/08 15:32:42 dondosha * Corrected file description comment in the NCBI header * * Revision 6.32 2002/09/03 16:01:11 dondosha * Introduced a restriction on the total length of concatenated sequences for nucleotide case * * Revision 6.31 2002/05/30 16:20:16 dondosha * 1. Correction in evaluation of hits for DNA case, * 2. Added word size parameter to make DNA clustering more flexible * * Revision 6.30 2002/03/14 21:07:11 dondosha * Slight improvement in finding the used id list from a gi list when reclustering * * Revision 6.29 2002/02/11 21:32:10 dondosha * Free Bioseqs after use to avoid accumulation * * Revision 6.28 2001/11/16 20:34:17 dondosha * Bug fix: BioseqUnlock was missing; also added ObjMgrFreeCache calls * * Revision 6.27 2001/08/15 17:49:08 dondosha * Correction in previous change * * Revision 6.26 2001/08/15 17:27:17 dondosha * Consider and save only the top HSP for each pair of sequences being compared * * Revision 6.25 2001/07/26 18:23:48 dondosha * Added a -c option: configuration file with advanced BLAST options * * Revision 6.24 2001/02/13 16:26:03 dondosha * Bug fix: memory freed in the wrong place * * Revision 6.23 2001/01/31 19:48:14 dondosha * Change #define TMPDIR to environment variable * * Revision 6.22 2001/01/26 20:13:05 dondosha * If at least one sequence id non-numeric, make them all non-numeric * * Revision 6.21 2000/12/19 18:52:38 dondosha * Added option to disallow id parsing when formatting database * * Revision 6.20 2000/12/14 20:51:14 dondosha * Do several Mega BLAST searches if more than 16383 sequences in input * * Revision 6.19 2000/12/07 22:57:34 dondosha * Implemented nucleotide version using Mega BLAST * * Revision 6.18 2000/12/01 17:57:04 dondosha * Handle local and general ids differently to avoid non-informative output * * Revision 6.17 2000/09/01 18:30:55 dondosha * Create database memory map only once for all searches * * Revision 6.16 2000/08/08 17:58:55 dondosha * Change back to create database with indexes * * Revision 6.15 2000/08/07 19:40:57 dondosha * Removed sequence lengths from ClusterLogInfo structure - redundant information * * Revision 6.14 2000/08/07 15:13:36 dondosha * Changed comment for -S option * * Revision 6.13 2000/08/04 23:24:40 dondosha * Added functionality to choose neighbors based on percentage of identities * * Revision 6.12 2000/06/08 20:40:15 dondosha * In case of equal lengths of sequences within a cluster or equal number of cluster elements, use alphabetic or numeric order when sorting ids and clusters * * Revision 6.11 2000/06/08 13:34:57 dondosha * Increased buffer size for reading id file * * Revision 6.10 2000/06/07 21:57:04 dondosha * Added option to provide a file with id list for reclustering * * Revision 6.9 2000/06/07 14:23:20 dondosha * Changed ctime_r call to portable DayTimeStr * * Revision 6.8 2000/06/06 19:32:41 dondosha * Changed progress interval back to 1000 * * Revision 6.7 2000/06/06 19:23:42 dondosha * Added option to finish incomplete clustering from the point where previous search ended * * Revision 6.5 2000/05/22 18:19:59 dondosha * In case search set up fails, destruct all necessary stuff before going to next query * * Revision 6.4 2000/05/17 17:44:50 dondosha * Cleaned from unused variables * * Revision 6.3 2000/05/05 18:16:56 dondosha * Enhanced to process all types of SeqIds * * Revision 6.2 2000/05/03 16:50:40 dondosha * Removed system calls, added sorting of clusters by size, sequences within clusters by length * * Revision 6.1 2000/04/27 14:47:18 dondosha * Clustering of protein neighbours - initial revision * * Revision 1.1 2000/04/12 12:53:04 dondosha * Clustering of protein neighbors * * =========================================================================== */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define DEBUG 0 /** File to write cluster output. */ FILE *global_fp=NULL; /** Array of indices pointing to root sequences of each cluster. */ static Int4Ptr root; /** Mutex for modifying the cluster root array. */ static TNlmMutex root_mutex; /** HSP information, used internally. */ typedef struct blastclust_hsp_info { Int4 id; /**< Subject ordinal number. */ Int4 score; /**< HSP score */ Int4 index; } BlastClustHspInfo, PNTR BlastClustHspInfoPtr; /** User parameters for clustering. */ typedef struct cluster_parameters { Boolean bidirectional; /**< Is bydirectional coverage required? */ FloatHi length_threshold; /**< Length coverage threshold. */ FloatHi score_threshold; /**< Score coverage threshold. */ FILE *logfp; /**< File to write hit data to. */ } ClusterParameters, PNTR ClusterParametersPtr; /** Globally accessible parameters structure. */ static ClusterParametersPtr global_parameters; /** Basic structure written to the hit file. */ typedef struct cluster_log_info { Int4 id1; /**< Index of the first sequence for this HSP. */ Int4 id2; /**< Index of the second sequence for this HSP. */ Int4 hsp_length1; /**< Length of this HSP in first sequence. */ Int4 hsp_length2; /**< Length of this HSP in second sequence. */ FloatHi bit_score;/**< This HSP's bit score. */ FloatHi perc_identity; /**< This HSP's percent identity. */ } ClusterLogInfo, PNTR ClusterLogInfoPtr; /** Input sequence information. */ typedef struct blast_cluster_element { Int4 gi; /**< Gi or other numeric identifier. */ CharPtr id;/**< String identifier. */ Int4 len; /**< Sequence length. */ } BlastClusterElement, PNTR BlastClusterElementPtr; /** Information about all input sequences. */ typedef struct blast_cluster { Int4 size; /** Number of sequences. */ BlastClusterElementPtr PNTR elements; /** Array of pointers to input sequence information structures. */ } BlastCluster, PNTR BlastClusterPtr; /** Comparison function for sorting input sequences in the clusters. * Sequences are sorted as follows, with decreasing priorities * 1. Decreasing order of length * 2. Increasing order of numeric identifiers * 3. Lexicographic order of string identifiers. */ static int LIBCALLBACK compare_cluster_elements(VoidPtr v1, VoidPtr v2) { BlastClusterElementPtr e1, e2; e1 = *(BlastClusterElementPtr PNTR) v1; e2 = *(BlastClusterElementPtr PNTR) v2; if (e1->len > e2->len) return -1; else if (e1->len < e2->len) return 1; else if (e1->gi > 0 && e2->gi>0) { if (e1->gi < e2->gi) return -1; else if (e1->gi > e2->gi) return 1; } else return StrCmp(e1->id, e2->id); return 0; } /** Comparison function for sorting clusters by size. Sequences within clusters * are sorted in a specific order @sa compare_cluster_elements. */ static int LIBCALLBACK compare_clusters(VoidPtr v1, VoidPtr v2) { BlastClusterPtr c1, c2; c1 = *(BlastClusterPtr PNTR) v1; c2 = *(BlastClusterPtr PNTR) v2; if (c1->size > c2->size) return -1; else if (c1->size < c2->size) return 1; else return compare_cluster_elements(c1->elements, c2->elements); } /** Comparison function for sorting HSP information structures. Sorting is * performed according to the following rules with decreasing priority: * 1. In increasing order of sequence index; * 2. In decreasing order of scores. */ static int LIBCALLBACK hsp_info_id_score_compare(VoidPtr v1, VoidPtr v2) { BlastClustHspInfoPtr c1, c2; c1 = *(BlastClustHspInfoPtr PNTR) v1; c2 = *(BlastClustHspInfoPtr PNTR) v2; if (c1->id == c2->id && c1->score == c2->score) return 0; else if ((c1->id < c2->id) || (c1->id == c2->id && c1->score > c2->score)) return -1; else return 1; } /** Starting size for a cluster, which can be dynamically reallocated when * necessary. */ #define ORIGINAL_CLUSTER_SIZE 10 /** Creates the clusters, based on the root array, and prints them to the * cluster output file. * @param num_queries Number of sequences being clustered. [in] * @param seq_len Array of sequence lengths [in] * @param id_list Array of sequence string identifiers [in] * @param gi_list array of sequence numeric identifiers [in] * @param used_id_index Array of flags indicating whether id with a given index * has already been processed. [in] */ static int BlastClusterNeighbours(Int4 num_queries, Int4Ptr seq_len, CharPtr PNTR id_list, Int4Ptr gi_list, Boolean PNTR used_id_index) { BlastClusterPtr PNTR cluster; Int4 num_clusters, available_size, index, i; Boolean numeric_id_type; if (gi_list) numeric_id_type = TRUE; else if (id_list) numeric_id_type = FALSE; else return 0; for (index=1; indexsize = 1; cluster[num_clusters]->elements = (BlastClusterElementPtr PNTR) Malloc(ORIGINAL_CLUSTER_SIZE*sizeof(BlastClusterElementPtr)); cluster[num_clusters]->elements[0] = (BlastClusterElementPtr) MemNew(sizeof(BlastClusterElement)); if (numeric_id_type) cluster[num_clusters]->elements[0]->gi = gi_list[index]; else cluster[num_clusters]->elements[0]->id = id_list[index]; cluster[num_clusters]->elements[0]->len = seq_len[index]; available_size = ORIGINAL_CLUSTER_SIZE; /* Find other sequences belonging to this cluster and add them. */ for (i = index+1; i < num_queries; i++) { if (root[i] == index) { Int4 size = cluster[num_clusters]->size; /* Reallocate the elements array if necessary. */ if (size >= available_size) { available_size *= 2; cluster[num_clusters]->elements = (BlastClusterElementPtr PNTR) Realloc(cluster[num_clusters]->elements, available_size*sizeof(BlastClusterElementPtr)); } cluster[num_clusters]->elements[size] = (BlastClusterElementPtr) MemNew(sizeof(BlastClusterElement)); if (numeric_id_type) cluster[num_clusters]->elements[size]->gi = gi_list[i]; else cluster[num_clusters]->elements[size]->id = id_list[i]; cluster[num_clusters]->elements[size]->len = seq_len[i]; ++cluster[num_clusters]->size; } } num_clusters++; } } /* Sort each cluster in decreasing order of sequence lengths */ for (index=0; indexelements, cluster[index]->size, sizeof(BlastClusterElementPtr), compare_cluster_elements); /* Sort clusters in decreasing order of sizes */ HeapSort(cluster, num_clusters, sizeof(BlastClusterPtr), compare_clusters); /* Print out all clusters */ for (index=0; indexsize; i++) { if (numeric_id_type) fprintf(global_fp, "%ld ", cluster[index]->elements[i]->gi); else fprintf(global_fp, "%s ", cluster[index]->elements[i]->id); MemFree(cluster[index]->elements[i]); } fprintf(global_fp, "\n"); MemFree(cluster[index]->elements); MemFree(cluster[index]); } MemFree(cluster); return 1; } #define INFO_LIST_SIZE 1000 #define FILE_BUFFER_SIZE 4096 /* Reclusters saved hits and returns the ordinal id of the last query found in * the hits file, plus 1. * @param infofp File to read hit information from [in] * @param outfp File to write clusters to [in] * @param gilp Pointer to an array of gis (numeric ids) found in the hits * file [out] * @param idlp Pointer to an array of string sequence ids found in the hits * file [out] * @param seqlp Pointer to an array of lengths of sequences found in the hits * file [out] * @param idfp File to read a list of ids that need to be reclustered * (optional) [in] * @return Index of the last query found in the hits file ( == ordinal id + 1) */ static Int4 ReclusterFromFile(FILE *infofp, FILE *outfp, Int4Ptr PNTR gilp, CharPtr PNTR PNTR idlp, Int4Ptr PNTR seqlp, FILE *idfp) { ClusterLogInfoPtr info; Int4 num_queries, num_hits, i, root1, root2, total_id_len; Int4Ptr gi_list = NULL, seq_len = NULL; CharPtr PNTR id_list = NULL; CharPtr ptr, id_string = NULL; FloatHi length_coverage, score_coverage; Uint4 header_size, numeric_id_type; Int4 last_seq = -1; CharPtr id = NULL; Boolean PNTR used_id_index = NULL; /* Read header data from the hits file. */ FileRead(&numeric_id_type, sizeof(Int4), 1, infofp); FileRead(&header_size, sizeof(Int4), 1, infofp); if (numeric_id_type) { num_queries = header_size; gi_list = (Int4Ptr) MemNew(num_queries*sizeof(Int4)); FileRead(gi_list, sizeof(Int4), num_queries, infofp); } else { total_id_len = header_size; num_queries = 0; id_string = (CharPtr) MemNew(total_id_len+1); FileRead(id_string, sizeof(Char), total_id_len, infofp); ptr = id_string; /* Count the ID's and change delimiter from space to null character */ for (i=0; i 0) { for (i=0; ibidirectional) length_coverage = MIN(((FloatHi)info[i].hsp_length1) / seq_len[info[i].id1], ((FloatHi)info[i].hsp_length2) / seq_len[info[i].id2]); else length_coverage = MAX(((FloatHi)info[i].hsp_length1) / seq_len[info[i].id1], ((FloatHi)info[i].hsp_length2) / seq_len[info[i].id2]); if (global_parameters->score_threshold < 3.0) score_coverage = info[i].bit_score / (MAX(info[i].hsp_length1, info[i].hsp_length2)); else score_coverage = info[i].perc_identity; /* If coverage satisfies the input requirements, update the cluster root information. */ if (length_coverage >= global_parameters->length_threshold && score_coverage >= global_parameters->score_threshold) { root1 = info[i].id1; while (root[root1] != root1) root1 = root[root1]; root2 = info[i].id2; while (root[root2] != root2) root2 = root[root2]; if (root1 < root2) root[root2] = root1; else if (root1 > root2) root[root1] = root2; } } /* End loop on hits from a chunk */ last_seq = info[num_hits-1].id1; } /* End loop on chunks of hits */ /* Create the cluster structures and print out. */ if (outfp != NULL) { global_fp = outfp; BlastClusterNeighbours(num_queries, seq_len, id_list, gi_list, used_id_index); } *gilp = gi_list; *idlp = id_list; *seqlp = seq_len; MemFree(id_string); MemFree(info); return last_seq + 1; } /** Calculates percent identity given a gapped alignment block. * @param HSP gapped alignment structure, returned from a gapped alignment * routine. [in] * @return percent of identical matches in this alignment. */ static FloatHi GapAlignPercentIdentity(GapAlignBlkPtr gabp) { GapXEditScriptPtr esp = gabp->edit_block->esp; Int4 identical, total, q_index, s_index, i; FloatHi perc_identity; Uint1Ptr query, subject; q_index = gabp->query_start; s_index = gabp->subject_start; query = gabp->query + q_index; subject = gabp->subject + s_index; identical = total = 0; while (esp) { if (esp->op_type == GAPALIGN_SUB || esp->op_type == GAPALIGN_DECLINE) { for (i=0; inum; i++) { if (*query == *subject) identical++; query++; subject++; } } else if (esp->op_type == GAPALIGN_DEL) subject += esp->num; else if (esp->op_type == GAPALIGN_INS) query += esp->num; total += esp->num; esp = esp->next; } perc_identity = ((FloatHi)identical) / total * 100; return perc_identity; } /** Line length for configuration file with advanced options. */ #define BUFFER_SIZE 80 /** Callback for printing out HSP data into the hits file. * @param ptr BlastSearchBlkPtr, containing current list of HSPs. [in] */ static int LIBCALLBACK PrintNeighbors(VoidPtr ptr) { BLAST_HitListPtr current_hitlist; BLAST_HSPPtr hsp; Int4 index; BlastSearchBlkPtr search; Int4 id1, id2, root1, root2, hspcnt; Int4 subject_length; Uint1Ptr subject, query; ClusterLogInfoPtr loginfo = NULL; BlastClustHspInfoPtr PNTR hsp_info_array; Int4 query_count; if (ptr == NULL) return 0; search = (BlastSearchBlkPtr) ptr; if (search->current_hitlist == NULL || search->current_hitlist->hspcnt <= 0) { /* No hits to save. */ search->subject_info = BLASTSubjectInfoDestruct(search->subject_info); return 0; } current_hitlist = search->current_hitlist; if (search->prog_number == blast_type_blastp) subject_length = readdb_get_sequence(search->rdfp, search->subject_id, &subject); else { /* Mega BLAST saves ncbi4na-encoded sequence */ subject = search->subject->sequence_start + 1; subject_length = search->subject->length; } hspcnt = current_hitlist->hspcnt; if (search->prog_number == blast_type_blastp) id1 = SeqId2OrdinalId(search->rdfp, search->query_id); else id1 = -1; id2 = search->subject_id; /* For blastp, save only HSPs with first sequence index smaller than second. */ if (id1 < id2) { #define BUF_CHUNK_SIZE 1024 Int4 query_length, q_length, s_length; FloatHi length_coverage, bit_score, score_coverage, perc_identity; BLAST_KarlinBlkPtr kbp; GapAlignBlkPtr gap_align = search->gap_align; Int2 context; Int4 max_score = 0; if (search->prog_number == blast_type_blastp) { query = search->context[0].query->sequence; query_length = search->context[0].query->length; } query_count = 1; /* For blastn, the current hit list contains HSPs from multiple "query" sequences. Create an array of HSP information structures, and find highest scoring HSP for every pair of sequences present in the current list. */ if (search->prog_number == blast_type_blastn) { hsp_info_array = (BlastClustHspInfoPtr PNTR) Malloc(hspcnt*sizeof(BlastClustHspInfoPtr)); for (index=0; indexhsp_array[index]; hsp_info_array[index]->id = SeqId2OrdinalId(search->rdfp, search->qid_array[hsp->context/2]); hsp_info_array[index]->score = hsp->score; hsp_info_array[index]->index = index; } HeapSort(hsp_info_array, hspcnt, sizeof(BlastClustHspInfoPtr), hsp_info_id_score_compare); /* Leave only one highest scoring hsp per "query" sequence. Deallocate memory for all others. */ for (index=1; indexid < id2 && hsp_info_array[index]->id != hsp_info_array[query_count-1]->id) { hsp_info_array[query_count] = hsp_info_array[index]; query_count++; } else hsp_info_array[index] = MemFree(hsp_info_array[index]); } hsp = current_hitlist->hsp_array[hsp_info_array[0]->index]; } else { /* For blastp, just find the highest-scoring HSP. */ for (index=0; indexhsp_array[index]->score > max_score) { hsp = current_hitlist->hsp_array[index]; max_score = hsp->score; } } } if (global_parameters->logfp) loginfo = (ClusterLogInfoPtr) MemNew(query_count*sizeof(ClusterLogInfo)); index = 0; while (hsp) { /* Calculate the coverage values for this HSP. */ if (search->prog_number == blast_type_blastn) { context = hsp->context; id1 = hsp_info_array[index]->id; query_length = search->query_context_offsets[context+1] - search->query_context_offsets[context] - 1; } q_length = hsp->query.length; s_length = hsp->subject.length; if (global_parameters->bidirectional) length_coverage = MIN(((FloatHi)q_length) / query_length, ((FloatHi)s_length) / subject_length); else length_coverage = MAX(((FloatHi)q_length) / query_length, ((FloatHi)s_length) / subject_length); if (search->prog_number == blast_type_blastp) kbp = search->sbp->kbp_gap[search->first_context]; else kbp = search->sbp->kbp[context]; bit_score = ((hsp->score*kbp->Lambda) - kbp->logK)/NCBIMATH_LN2; /* Calculate percent identity, performing traceback search, if necessary. */ if (search->prog_number == blast_type_blastp) { gap_align->query_frame = ContextToFrame(search, hsp->context); gap_align->subject_frame = hsp->subject.frame; gap_align->q_start = hsp->query.gapped_start; gap_align->s_start = hsp->subject.gapped_start; PerformGappedAlignmentWithTraceback(gap_align); perc_identity = GapAlignPercentIdentity(gap_align); gap_align->state_struct = GapXDropStateDestroy(gap_align->state_struct); gap_align->edit_block = GapXEditBlockDelete(gap_align->edit_block); } else { Int4Ptr length, start; Uint1Ptr strands; Int4 align_length = 0, numseg, i; GapXEditScriptPtr esp; query = search->context[context].query->sequence; esp = hsp->gap_info->esp; for (numseg=0; esp; esp = esp->next, numseg++); GXECollectDataForSeqalign(hsp->gap_info, hsp->gap_info->esp, numseg, &start, &length, &strands, &hsp->query.offset, &hsp->subject.offset); perc_identity = 0; for (i=0; iscore_threshold < 3.0) score_coverage = bit_score / (MAX(q_length, s_length)); else score_coverage = perc_identity; /* If hits are being saved, save this HSP's information. */ if (global_parameters->logfp) { loginfo[index].id1 = id1; loginfo[index].id2 = id2; loginfo[index].hsp_length1 = q_length; loginfo[index].hsp_length2 = s_length; loginfo[index].bit_score = bit_score; loginfo[index].perc_identity = perc_identity; } /* If this HSP satisfies the coverage criteria, update the cluster roots. */ if (length_coverage >= global_parameters->length_threshold && score_coverage >= global_parameters->score_threshold) { root1 = id1; NlmMutexLockEx(&root_mutex); while (root[root1] != root1) root1 = root[root1]; root2 = id2; while (root[root2] != root2) root2 = root[root2]; if (root1 < root2) root[root2] = root1; else if (root1 > root2) root[root1] = root2; NlmMutexUnlock(root_mutex); } /* For blastn, get to the highest scoring hsp for the next "query" */ if (search->prog_number == blast_type_blastn && ++index < query_count) hsp = current_hitlist->hsp_array[hsp_info_array[index]->index]; else hsp = NULL; } /* Deallocate memory for the auxiliary array of HSP info structures. Also free all gap information structures, because they are not freed when hit list is cleaned. */ if (search->prog_number == blast_type_blastn) { for (index=0; indexhspcnt; ++index) { hsp = current_hitlist->hsp_array[index]; hsp->gap_info = GapXEditBlockDelete(hsp->gap_info); } } if (global_parameters->logfp && loginfo) { FileWrite(loginfo, sizeof(ClusterLogInfo), query_count, global_parameters->logfp); loginfo = MemFree(loginfo); fflush(global_parameters->logfp); } } else { /* This can't happen in normal situation. If it does, the most likely reason is presense of non-unique identifiers in the input file */ ErrPostEx(SEV_FATAL, 1, 0, "Blastclust cannot process input files with" " non-unique sequence identifiers\n"); return 1; } return 0; } #ifdef NUMARG #undef NUMARG #endif #define NUMARG (sizeof(myargs)/sizeof(myargs[0])) static Args myargs [] = { { "FASTA input file (program will format the database and remove files in the end)", /* 0 */ "stdin", NULL, NULL, FALSE, 'i', ARG_FILE_IN, 0.0, 0, NULL}, { "Number of CPU's to use", /* 1 */ "1", NULL, NULL, FALSE, 'a', ARG_INT, 0.0, 0, NULL}, { "Output file for list of clusters", /* 2 */ "stdout", NULL, NULL, FALSE, 'o', ARG_FILE_OUT, 0.0, 0, NULL}, { "Length coverage threshold", /* 3 */ "0.9", NULL, NULL, FALSE, 'L', ARG_FLOAT, 0.0, 0, NULL}, { "Score coverage threshold (bit score / length if < 3.0, percentage of identities otherwise)", /* 4 */ "1.75", NULL, NULL, FALSE, 'S', ARG_FLOAT, 0.0, 0, NULL}, { "Require coverage on both neighbours?", /* 5 */ "TRUE", NULL, NULL, FALSE, 'b', ARG_BOOLEAN, 0.0, 0, NULL}, { "File to save all neighbours", /* 6 */ NULL, NULL, NULL, TRUE, 's', ARG_FILE_OUT, 0.0, 0, NULL}, { "File to restore neighbors for reclustering", /* 7 */ NULL, NULL, NULL, TRUE, 'r', ARG_FILE_IN, 0.0, 0, NULL}, { "Input as a database", /* 8 */ NULL, NULL, NULL, TRUE, 'd', ARG_FILE_IN, 0.0, 0, NULL}, { "Print progress messages (verbose mode)", /* 9 */ "stdout", NULL, NULL, FALSE, 'v', ARG_FILE_OUT, 0.0, 0, NULL}, { "Complete unfinished clustering", /* 10 */ "FALSE", NULL, NULL, FALSE, 'C', ARG_BOOLEAN, 0.0, 0, NULL}, { "Restrict reclustering to id list", /* 11 */ NULL, NULL, NULL, TRUE, 'l', ARG_FILE_IN, 0.0, 0, NULL}, { "Is input proteins?", /* 12 */ "TRUE", NULL, NULL, FALSE, 'p', ARG_BOOLEAN, 0.0, 0, NULL}, { "Enable id parsing in database formatting?", /* 13 */ "TRUE", NULL, NULL, FALSE, 'e', ARG_BOOLEAN, 0.0, 0, NULL}, { "Configuration file with advanced options", /* 14 */ NULL, NULL, NULL, TRUE, 'c', ARG_FILE_IN, 0.0, 0, NULL}, { "Word size to use (0 for default: proteins 3, nucleotides 32)",/* 15 */ "0", NULL, NULL, FALSE, 'W', ARG_INT, 0.0, 0, NULL} }; /** Print progress messages after how many processed sequences? */ #define PROGRESS_INTERVAL 1000 /** Maximal number of sequences to concatenate for a blastn search. */ #define MAX_NUM_QUERIES 16383 /* == 1/2 INT2_MAX */ /** Maximal total length of concatenated sequences in a blastn search. */ #define MAX_TOTAL_LENGTH 5000000 Int2 Main (void) { BLAST_OptionsBlkPtr options; BlastSearchBlkPtr search; Boolean db_is_na, query_is_na; Int4 qsize, dbsize, first_seq; ReadDBFILEPtr rdfp; Uint1 align_type; SeqIdPtr sip; BioseqPtr query_bsp = NULL, PNTR query_bsp_array; CharPtr blast_program, blast_inputfile, blast_outputfile, blast_database, progress_file = NULL; CharPtr logfile, info_file, input_name; Int4 total_id_len = 0; FILE *outfp, *infofp, *progressfp, *idfp; Int4 index, i, num_queries, num_bsps, num_left; Int8 total_length; Int4Ptr gi_list = NULL, seq_len = NULL; CharPtr PNTR id_list = NULL, id_string = NULL; Boolean db_formatted = FALSE; Boolean numeric_id_type = TRUE; Char db_file[BUFFER_SIZE]; Boolean print_progress, finish_incomplete, is_prot, parse_mode; FDB_optionsPtr fdb_options; Char timestr[24]; CharPtr tmpdir; Int4 total_query_length; CharPtr title = NULL; Char buf[256] = { '\0' }; StringCpy(buf, "blastclust "); StringNCat(buf, BlastGetVersionNumber(), sizeof(buf)-StringLen(buf)-1); if (! GetArgs (buf, NUMARG, myargs)) return (1); UseLocalAsnloadDataAndErrMsg (); if (! SeqEntryLoad()) return 1; ErrSetMessageLevel(SEV_WARNING); global_parameters = (ClusterParametersPtr) MemNew(sizeof(ClusterParameters)); global_parameters->length_threshold = myargs[3].floatvalue; global_parameters->score_threshold = myargs[4].floatvalue; global_parameters->bidirectional = (Boolean) myargs[5].intvalue; finish_incomplete = (Boolean) myargs[10].intvalue; print_progress = (Boolean) StrCmp(myargs[9].strvalue, "F"); if (print_progress) progress_file = myargs[9].strvalue; if (progress_file != NULL && (progressfp = FileOpen(progress_file, "w")) == NULL) { ErrPostEx(SEV_FATAL, 1, 0, "blastclust: Unable to open progress file %s\n", progress_file); return (1); } blast_outputfile = myargs[2].strvalue; outfp = NULL; if (blast_outputfile != NULL) { if ((outfp = FileOpen(blast_outputfile, "w")) == NULL) { ErrPostEx(SEV_FATAL, 1, 0, "blastclust: Unable to open output file %s\n", blast_outputfile); return (1); } } info_file = myargs[7].strvalue; if (info_file) { /* Non-empty string means only retrieve neighbors for reclustering */ if ((infofp = FileOpen(info_file, "rb")) == NULL) { ErrPostEx(SEV_FATAL, 1, 0, "blastclust: Unable to open neighbors file %s for reading\n", info_file); return (1); } if (myargs[11].strvalue) { if ((idfp = FileOpen(myargs[11].strvalue, "r")) == NULL) { ErrPostEx(SEV_FATAL, 1, 0, "blastclust: Unable to open id file %s for reading\n", myargs[11].strvalue); return (1); } } else idfp = NULL; /* No need for another search, simply get all the neighbours and reculster them using new thresholds */ ReclusterFromFile(infofp, outfp, &gi_list, &id_list, &seq_len, idfp); MemFree(gi_list); MemFree(id_list); MemFree(seq_len); FileClose(infofp); FileClose(outfp); return 0; } else infofp = NULL; is_prot = (Boolean) myargs[12].intvalue; parse_mode = (Boolean) myargs[13].intvalue; if (is_prot) blast_program = StringSave("blastp"); else blast_program = StringSave("blastn"); if (myargs[8].strvalue) db_formatted = TRUE; if (db_formatted) { blast_database = myargs[8].strvalue; } else { /* Need to format the database */ blast_inputfile = myargs[0].strvalue; input_name = FileNameFind(blast_inputfile); if (tmpdir = getenv("TMPDIR")) { blast_database = Malloc(StringLen(input_name) + StringLen(tmpdir) + 2); sprintf(blast_database, "%s/%s", tmpdir, input_name); } else blast_database = blast_inputfile; fdb_options = FDBOptionsNew(blast_inputfile, is_prot, NULL, FALSE, FALSE, FALSE, FALSE, TRUE, parse_mode, blast_database, NULL, 0, 0, FORMATDB_VER, FALSE, FALSE); FastaToBlastDB(fdb_options, 0); fdb_options = FDBOptionsFree(fdb_options); } logfile = myargs[6].strvalue; if (logfile) { /* Empty string means do not write log information */ if (finish_incomplete) { if ((global_parameters->logfp = FileOpen(logfile, "ab+")) == NULL) { ErrPostEx(SEV_FATAL, 1, 0, "blast: Unable to open log file %s for appending\n", logfile); return (1); } } else { if ((global_parameters->logfp = FileOpen(logfile, "wb+")) == NULL) { ErrPostEx(SEV_FATAL, 1, 0, "blast: Unable to open log file %s for writing\n", logfile); return (1); } } } global_fp = outfp; align_type = BlastGetTypes(blast_program, &query_is_na, &db_is_na); rdfp = readdb_new(blast_database, is_prot); options = BLASTOptionNew(blast_program, TRUE); if (options == NULL) return 3; /* Set the default search options. */ options->use_real_db_size = TRUE; options->sort_gi_list = FALSE; options->expect_value = 1e-6; qsize = 300; dbsize = (20*1000*1000); options->searchsp_eff = (FloatHi) qsize * (FloatHi) dbsize; options->perform_culling = FALSE; options->do_not_reevaluate = TRUE; options->do_sum_stats = FALSE; options->number_of_cpus = myargs[1].intvalue; if (!is_prot) { options->is_megablast_search = TRUE; options->gap_open = options->gap_extend = 0; options->wordsize = 32; options->block_width = 0; options->window_size = 0; } if (myargs[15].intvalue != 0) options->wordsize = myargs[15].intvalue; if (myargs[14].strvalue) { CharPtr string_options = NULL; Int2 string_options_len = 0; CharPtr error = NULL; FILE *config_file; Char opt_line[BUFFER_SIZE+1]; if ((config_file = FileOpen(myargs[14].strvalue, "r")) == NULL) { ErrPostEx(SEV_FATAL, 1, 0, "Cannot open advanced options configuration file %s\n", myargs[14].strvalue); } /* Read advanced options from a configuration file, one line at a time. */ while (fgets(opt_line, BUFFER_SIZE, config_file) != 0) { if (*opt_line == '#') continue; if (!string_options) { string_options = (CharPtr) MemNew(BUFFER_SIZE+1); string_options_len = BUFFER_SIZE; } else { if (StrLen(string_options) + StrLen(opt_line) + 1 > string_options_len) { CharPtr tmp = (CharPtr) Realloc(string_options, 2*string_options_len); if (tmp) { string_options = tmp; string_options_len *= 2; } } StringCat(string_options, " "); } StringCat(string_options, opt_line); } if (string_options) parse_blast_options(options, string_options, &error, NULL, NULL, NULL); if (error) { ErrPostEx(SEV_WARNING, 0, 0, "blastclust: parse_blast_options: %s\n", error); } } readdb_get_totals_ex(rdfp, &total_length, &num_queries, TRUE); root = (Int4Ptr) Malloc(num_queries*sizeof(Int4)); /* If this is a continuation of a previous search that has not been completed, read the previous search information and start from there. */ if (is_prot && finish_incomplete) { first_seq = ReclusterFromFile(global_parameters->logfp, NULL, &gi_list, &id_list, &seq_len, NULL); } else { Uint4 header_size = 0; Uint4 header_numeric_id_type = 0; first_seq = 0; /* Put each sequence in its own one-element cluster. */ for (index=0; indexchoice != SEQID_GENERAL || StringCmp(((DbtagPtr)sip->data.ptrvalue)->db, "BL_ORD_ID")) { numeric_id_type &= GetAccessionFromSeqId(sip, &gi_list[index], &id_list[index]); } else { CharPtr dummy_ptr = NULL; numeric_id_type = FALSE; if (title && *title != NULLB) { id_list[index] = StringSave(StringTokMT(title, " ", &dummy_ptr)); } else { id_list[index] = (CharPtr) Malloc(BUFFER_SIZE+1); SeqIdWrite(sip, id_list[index], PRINTID_FASTA_SHORT, BUFFER_SIZE); } } title = MemFree(title); sip = SeqIdSetFree(sip); } if (numeric_id_type) { id_list = MemFree(id_list); header_size = num_queries; header_numeric_id_type = 1; } else { total_id_len = 0; /* Check if some ids were gis and convert them to strings */ for (i=0; i 0) { id_list[i] = (CharPtr) MemNew(10); sprintf(id_list[i], "%ld", gi_list[i]); } total_id_len += StringLen(id_list[i]) + 1; } gi_list = MemFree(gi_list); id_string = (CharPtr) MemNew(total_id_len+1); for (i=0; ilogfp); FileWrite(&header_size, sizeof(Uint4), 1, global_parameters->logfp); if (numeric_id_type) FileWrite(gi_list, sizeof(Int4), num_queries, global_parameters->logfp); else { FileWrite(id_string, sizeof(Char), total_id_len, global_parameters->logfp); MemFree(id_string); } FileWrite(seq_len, sizeof(Int4), num_queries, global_parameters->logfp); fflush(global_parameters->logfp); } /* Print the first progress message, if necessary. */ if (print_progress) { DayTimeStr(timestr, TRUE, TRUE); if (finish_incomplete) fprintf(progressfp, "%s Finish clustering of %ld queries, starting from query %ld\n", timestr, num_queries, first_seq); else fprintf(progressfp, "%s Start clustering of %ld queries\n", timestr, num_queries); } if (!is_prot) { SeqAlignPtr PNTR head; /* Will not be used */ query_bsp_array = (BioseqPtr PNTR) Malloc((MIN(num_queries, MAX_NUM_QUERIES)+1)*sizeof(BioseqPtr)); num_left = num_queries; /* For blastn, concatenate as many queries as possible and run the search engine. */ while (num_left>0) { num_bsps = MIN(num_left, MAX_NUM_QUERIES); total_query_length = 0; for (index=0; indexlength; if (total_query_length > MAX_TOTAL_LENGTH) break; } num_bsps = index; query_bsp_array[num_bsps] = NULL; head = BioseqMegaBlastEngine(query_bsp_array, blast_program, blast_database, options, NULL, NULL, NULL, NULL, NULL, 0, PrintNeighbors); for (index=0; indexfirst_db_seq = index + 1; /* Set up search. */ search = BLASTSetUpSearchWithReadDbInternal(NULL, query_bsp, blast_program, seq_len[index], blast_database, options, NULL, NULL, NULL, 0, rdfp); if (search != NULL && !search->query_invalid) { search->handle_results = PrintNeighbors; /* Run BLAST. */ do_the_blast_run(search); } else if (search) { BlastErrorPrint(search->error_return); ErrPostEx(SEV_ERROR, 1, 0, "Failed to process query number %ld", (long) index); return 1; } search = BlastSearchBlkDestruct(search); query_bsp = BioseqFree(query_bsp); if (print_progress && (index + 1)%PROGRESS_INTERVAL == 0) { DayTimeStr(timestr, TRUE, TRUE); fprintf(progressfp, "%s Finished processing of %ld queries\n", timestr, index+1); } } /* End of loop on queries */ } rdfp = readdb_destruct(rdfp); /* Create the clusters and print them out. */ BlastClusterNeighbours(num_queries, seq_len, id_list, gi_list, NULL); MemFree(seq_len); if (numeric_id_type) MemFree(gi_list); else { for (i=0; ilogfp) FileClose(global_parameters->logfp); /* Remove the temporary BLAST database files. */ if (!db_formatted && StringLen(blast_database) > 0) { Char p_or_n = (is_prot) ? 'p' : 'n'; sprintf(db_file, "%s.%chr", blast_database, p_or_n); FileRemove(db_file); sprintf(db_file, "%s.%cin", blast_database, p_or_n); FileRemove(db_file); sprintf(db_file, "%s.%csq", blast_database, p_or_n); FileRemove(db_file); if (parse_mode) { sprintf(db_file, "%s.%cnd", blast_database, p_or_n); FileRemove(db_file); sprintf(db_file, "%s.%cni", blast_database, p_or_n); FileRemove(db_file); sprintf(db_file, "%s.%csd", blast_database, p_or_n); FileRemove(db_file); sprintf(db_file, "%s.%csi", blast_database, p_or_n); FileRemove(db_file); } } MemFree(blast_database); return 0; }