alg/nway_paths.cpp

   1 //  This file is part of the Mussa source distribution.
   2 //  http://mussa.caltech.edu/
   3 //  Contact author: Tristan  De Buysscher, tristan@caltech.edu
   4
   5 // This program and all associated source code files are Copyright (C) 2005
   6 // the California Institute of Technology, Pasadena, CA, 91125 USA.  It is
   7 // under the GNU Public License; please see the included LICENSE.txt
   8 // file for more information, or contact Tristan directly.
   9
  10
  11 //                        ----------------------------------------
  12 //                         ---------- mussa_nway.cc  -----------
  13 //                        ----------------------------------------
  14
  15 #include <boost/filesystem/fstream.hpp>
  16 namespace fs = boost::filesystem;
  17
  18 #include "alg/nway_paths.hpp"
  19 #include "alg/conserved_path.hpp"
  20 #include "mussa_exceptions.hpp"
  21
  22 #include <iostream>
  23 #include <stdexcept>
  24
  25 using namespace std;
  26
  27 NwayPaths::NwayPaths()
  28 {
  29 }
  30
  31 void
  32 NwayPaths::setup(int w, int t)
  33 {
  34   threshold = t;
  35   soft_thres = threshold;
  36   win_size = w;
  37   pathz.clear();
  38
  39   //cout << "nway: thres = " << threshold
  40   //     << ", soft threo = " << soft_thres << endl;
  41 }
  42
  43 void
  44 NwayPaths::set_soft_thres(int sft_thres)
  45 {
  46   soft_thres = sft_thres;
  47 }
  48
  49 int NwayPaths::get_threshold() const
  50 {
  51   return threshold;
  52 }
  53
  54 int NwayPaths::get_window() const
  55 {
  56   return win_size;
  57 }
  58
  59 // dumbly goes thru and combines path windows exactly adjacent (ie + 1 index)
  60 // doesn't deal with interleaved adjacency
  61 void
  62 NwayPaths::simple_refine()
  63 {
  64   // ext_path remembers the first window set in an extending path
  65   ExtendedConservedPath ext_path, new_path;
  66   list<ConservedPath>::iterator cur_path, next_path;
  67   list<ConservedPath>::iterator pathz_i;
  68   int win_ext_len = 0;
  69   bool extending, end = false;
  70
  71   refined_pathz.clear();
  72
  73   //cout << "path number is: " << pathz.size() << endl;
  74   pathz_i = pathz.begin();
  75
  76   // only try to extend when pathz isn't empty.
  77   if (pathz_i != pathz.end())
  78   {
  79     ext_path = ExtendedConservedPath( win_size, *pathz_i);
  80
  81     while(pathz_i != pathz.end())
  82     {
  83       // keep track of current path and advance to next path
  84       cur_path = pathz_i;
  85       ++pathz_i;
  86
  87       if (pathz_i == pathz.end()) {
  88         end = true;
  89         extending = false;
  90       } else {
  91         next_path = pathz_i;
  92         // if node for each seq is equal to the next node+1 then for all
  93         // sequences then we are extending
  94         extending = cur_path->nextTo(*next_path);
  95       }
  96
  97       if (extending)
  98       {
  99         win_ext_len++;
 100       }
 101       else
 102       {
 103         // add the extend window length as first element and add as refined
 104         // now that we have the path to extend save it
 105         new_path = ext_path;
 106         new_path.extend(win_ext_len);
 107         refined_pathz.push_back(new_path);
 108         if (not end) {
 109           // reset stuff
 110           win_ext_len = 0;
 111           ext_path = ExtendedConservedPath( win_size, *next_path);
 112         }
 113       }
 114     }
 115   }
 116   //cout << "r_path number is: " << refined_pathz.size() << endl;
 117 }
 118
 119
 120 void
 121 NwayPaths::add_path(int threshold, vector<int>& loaded_path)
 122 {
 123   pathz.push_back(ConservedPath(threshold, loaded_path));
 124 }
 125
 126 void
 127 NwayPaths::add_path(ConservedPath loaded_path)
 128 {
 129   pathz.push_back(loaded_path);
 130 }
 131
 132
 133 void
 134 NwayPaths::save(fs::path save_file_path)
 135 {
 136   fs::fstream save_file;
 137   list<ExtendedConservedPath >::iterator path_i, paths_end;
 138
 139   save_file.open(save_file_path, ios::out);
 140
 141   save_file << "<Mussa type=flp seq_count=" << sequence_count();
 142   save_file << " win=" << win_size;
 143   // add a function para new_thres defaults to -1 to later deal with
 144   // reanalysis with higher thres - if statement whether to record base thres
 145   // or new thres (ie if -1, then base)
 146   save_file << " thres=" << threshold << " >\n";
 147
 148   path_i = refined_pathz.begin();
 149   paths_end = refined_pathz.end();
 150   //path_i = pathz.begin();
 151   //paths_end = pathz.end();
 152   while (path_i != paths_end)
 153   {
 154     ExtendedConservedPath& a_path = *path_i;
 155     //cout << a_path.size() << endl;
 156     //first entry is the window length of the windows in the path
 157     save_file << a_path.window_size << ":";
 158     for(size_t i = 0; i != sequence_count(); ++i)
 159     {
 160       save_file << a_path[i];
 161       if (i != sequence_count())
 162         save_file << ",";
 163     }
 164     save_file << endl;
 165     ++path_i;
 166   }
 167
 168   save_file << "</Mussa>\n";
 169   save_file.close();
 170 }
 171
 172
 173 size_t
 174 NwayPaths::sequence_count()
 175 {
 176   if (refined_pathz.begin() == refined_pathz.end() )
 177     return 0;
 178   else
 179     return refined_pathz.begin()->size();
 180 }
 181
 182
 183 void
 184 NwayPaths::load(fs::path load_file_path)
 185 {
 186   fs::fstream load_file;
 187   string file_data_line, header_data, data, path_node, path_width;
 188   int space_split_i, equal_split_i, comma_split_i, colon_split_i;
 189   vector<int> loaded_path;
 190
 191   load_file.open(load_file_path, ios::in);
 192
 193   if (!load_file)
 194   {
 195     throw mussa_load_error("Sequence File: " + load_file_path.string() + " not found");
 196   }
 197   else
 198   {
 199     // get header data
 200     // grab mussa tag - discard for now...maybe check in future...
 201     getline(load_file,file_data_line);
 202     space_split_i = file_data_line.find(" ");
 203     file_data_line = file_data_line.substr(space_split_i+1);
 204     // grab type tag - need in future to distinguish between flp and vlp paths
 205     space_split_i = file_data_line.find(" ");
 206     file_data_line = file_data_line.substr(space_split_i+1);
 207     // get species/seq number
 208     space_split_i = file_data_line.find(" ");
 209     header_data = file_data_line.substr(0,space_split_i);
 210     equal_split_i = header_data.find("=");
 211     data = file_data_line.substr(equal_split_i+1);
 212     unsigned int species_num = atoi (data.c_str());
 213     file_data_line = file_data_line.substr(space_split_i+1);
 214     // get window size
 215     space_split_i = file_data_line.find(" ");
 216     header_data = file_data_line.substr(0,space_split_i);
 217     equal_split_i = header_data.find("=");
 218     data = file_data_line.substr(equal_split_i+1);
 219     win_size = atoi (data.c_str());
 220     file_data_line = file_data_line.substr(space_split_i+1);
 221     // get threshold size
 222     space_split_i = file_data_line.find(" ");
 223     header_data = file_data_line.substr(0,space_split_i);
 224     equal_split_i = header_data.find("=");
 225     data = file_data_line.substr(equal_split_i+1);
 226     threshold = atoi (data.c_str());
 227     file_data_line = file_data_line.substr(space_split_i+1);
 228
 229     //cout << "seq_num=" << species_num << " win=" << win_size;
 230     //cout << " thres=" << threshold << endl;
 231
 232     // clear out the current data
 233     refined_pathz.clear();
 234
 235     int temp;
 236
 237     getline(load_file,file_data_line);
 238     while ( (!load_file.eof()) && (file_data_line != "</Mussa>") )
 239     {
 240       if (file_data_line != "")
 241       {
 242         loaded_path.clear();
 243         colon_split_i = file_data_line.find(":");
 244         // whats our window size?
 245         path_width = file_data_line.substr(0,colon_split_i);
 246         file_data_line = file_data_line.substr(colon_split_i+1);
 247         for(size_t i = 0; i < species_num; i++)
 248         {
 249           comma_split_i = file_data_line.find(",");
 250           path_node = file_data_line.substr(0, comma_split_i);
 251           temp = atoi (path_node.c_str());
 252           loaded_path.push_back(temp);
 253           file_data_line = file_data_line.substr(comma_split_i+1);
 254         }
 255         assert (loaded_path.size() == species_num );
 256         refined_pathz.push_back(ExtendedConservedPath(atoi(path_width.c_str()),
 257                                                       threshold,
 258                                                       loaded_path));
 259       }
 260       getline(load_file,file_data_line);
 261     }
 262     load_file.close();
 263   }
 264 }
 265
 266
 267 void
 268 NwayPaths::path_search(vector<vector<FLPs> > all_comparisons, ConservedPath path, size_t depth)
 269 {
 270   list<int> new_nodes, trans_check_nodes;
 271   list<int>::iterator new_nodes_i, new_nodes_end;
 272   bool trans_check_good;
 273
 274   new_nodes = all_comparisons[depth - 1][depth].match_locations(path[depth-1]);
 275   new_nodes_i = new_nodes.begin();
 276   new_nodes_end = new_nodes.end();
 277   while(new_nodes_i != new_nodes_end)
 278   {
 279     //cout << "    * species " << depth << " node: " << *new_nodes_i << endl;
 280     // check transitivity with previous nodes in path
 281     trans_check_good = true;
 282     for(size_t i = 0; i < depth - 1; i++)
 283     {
 284       trans_check_nodes = all_comparisons[i][depth].match_locations(path[i]);
 285       if ( (trans_check_nodes.end() == find(trans_check_nodes.begin(),
 286                                             trans_check_nodes.end(),
 287                                             *new_nodes_i) ) &&
 288            (trans_check_nodes.end() == find(trans_check_nodes.begin(),
 289                                            trans_check_nodes.end(),
 290                                            *new_nodes_i * -1) ) )
 291         trans_check_good = false;
 292     }
 293
 294     if (trans_check_good)
 295     {
 296       // this makes sure path nodes are recorded with RC status relative to
 297       // the base species
 298       if ( path[depth-1] >= 0)
 299         path.push_back(*new_nodes_i);
 300       else
 301         path.push_back(*new_nodes_i * -1);
 302
 303       if (depth < all_comparisons.size() - 1)
 304         path_search(all_comparisons, path, depth + 1);
 305       else
 306         pathz.push_back(path);
 307       path.pop_back();
 308     }
 309     ++new_nodes_i;
 310   }
 311 }
 312 /* use this if I ever get the friggin seqcomp match lists to sort...
 313       if (binary_search(trans_check_nodes.begin(), trans_check_nodes.end(),
 314                         *new_nodes_i))
 315 */
 316
 317 void
 318 NwayPaths::find_paths_r(vector<vector<FLPs> > all_comparisons)
 319 {
 320   ConservedPath path;
 321   int win_i, window_num;
 322   list<int> new_nodes;
 323   list<int>::iterator new_nodes_i, new_nodes_end;
 324
 325   pathz.clear();
 326   window_num = all_comparisons[0][1].size();
 327   // loop thru all windows in first species
 328   for (win_i = 0; win_i < window_num; win_i++)
 329   {
 330     path.clear();
 331     path.push_back(win_i);
 332     new_nodes = all_comparisons[0][1].match_locations(path[0]);
 333     new_nodes_i = new_nodes.begin();
 334     new_nodes_end = new_nodes.end();
 335     //if (new_nodes_i != new_nodes_end)
 336     //cout << "* species 0 node: " << win_i << endl;
 337     path.push_back(0);
 338     while(new_nodes_i != new_nodes_end)
 339     {
 340       //cout << "  * species 1 node: " << *new_nodes_i << endl;
 341       path[1] = *new_nodes_i;
 342       path_search(all_comparisons, path, 2);
 343       ++new_nodes_i;
 344     }
 345   }
 346 }
 347
 348
 349 void
 350 NwayPaths::save_old(fs::path save_file_path)
 351 {
 352   fs::fstream save_file;
 353   list<ConservedPath >::iterator path_i, paths_end;
 354   size_t i;
 355
 356   save_file.open(save_file_path, ios::app);
 357
 358   path_i = pathz.begin();
 359   paths_end = pathz.end();
 360   while(path_i != paths_end)
 361   {
 362     ConservedPath& a_path = *path_i;
 363     //cout << a_path.size() << endl;
 364     for(i = 0; i < path_i->size(); ++i)
 365       save_file << i << "," << a_path[i] << " ";
 366     save_file << endl;
 367     ++path_i;
 368   }
 369   save_file.close();
 370 }
 371
 372
 373 /*
 374 void
 375 NwayPaths::find_paths(vector<vector<FLPs> > all_comparisons)
 376 {
 377   int win_i, sp_i;
 378   vector<list<list<int> > > path_src_tree;
 379   list<int> new_nodes;
 380   list<int>::iterator node_i, node_end;
 381   <list<list<int> > >::iterator branch_i, branch_end;
 382
 383   pathz.clear();
 384   path_src_tree.reserve(all_comparisons.size()- 1);
 385
 386
 387   // loop thru all windows in first species
 388   for (win_i = 0; win_i < window_num; win_i++)
 389   {
 390     // clear the path search tree
 391     for(i = 0; i < all_comparisons.size(); i++)
 392       path_src_tree[i].clear();
 393
 394     // top level kept empty even tho implicity has one entry of the first
 395     // species at this window - why bother, adds a little speed
 396
 397     // get connection list for first species, creating a list of nodes
 398     // of second species connected to the first species at this window
 399     new_nodes = all_comparisons[0][1];
 400     path_src_tree[1].push_back(new_nodes);
 401
 402     // loop thru rest of species for this window to see if any paths of matches
 403     // go across all species
 404     // if path search tree becomes empty, break out of loop, no reason to search further
 405     sp_i = 1;
 406     while ((sp_i < all_comparisons.size()) && (path tree not empty))
 407     {
 408       branch_i = path_src_tree[1].begin();
 409       branch_end = path_src_tree[1].end();
 410       while (branch_i != branch_end)
 411       {
 412         node_i = branch_i->begin();
 413         node_end = branch_i->end();
 414       }
 415
 416
 417       // loop over all current nodes
 418          // get connection list for each node
 419          // loop over each previous node in list
 420             // get those nodes connection list
 421             // intersect previous node connections with current
 422
 423       ++sp_i;
 424     }
 425
 426     // insert any of the paths found into the master list of paths
 427
 428     // add no paths if tmp_pathz is empty...
 429   }
 430 }
 431
 432 void NwayPaths::refine()
 433 {
 434 }
 435 */
 436
 437
 438 void NwayPaths::print(list<vector<int> >& dump_path)
 439 {
 440   list<vector<int> >::iterator pathz_i;
 441   vector<int>::iterator path_i;
 442
 443   cout << "printing list of lists\n";
 444   for (pathz_i = dump_path.begin(); pathz_i != dump_path.end(); ++pathz_i)
 445   {
 446     for (path_i = pathz_i->begin(); path_i != pathz_i->end(); ++path_i)
 447       cout << *path_i << " ";
 448     cout << endl;
 449   }
 450 }