0
|
1 #include "Constraints.hpp"
|
|
2
|
|
3 // return whether this constraint is compatible with the subexons.
|
|
4 bool Constraints::ConvertAlignmentToBitTable( struct _pair *segments, int segCnt,
|
|
5 struct _subexon *subexons, int seCnt, int seStart, struct _constraint &ct )
|
|
6 {
|
|
7 int i, j, k ;
|
|
8 k = seStart ;
|
|
9 ct.vector.Init( seCnt ) ;
|
|
10 // Each segment of an alignment can cover several subexons.
|
|
11 // But the first and last segment can partially cover a subexon.
|
|
12 for ( i = 0 ; i < segCnt ; ++i )
|
|
13 {
|
|
14 int leftIdx, rightIdx ; // the range of subexons covered by this segment.
|
|
15 leftIdx = -1 ;
|
|
16 rightIdx = -1 ;
|
|
17
|
|
18 for ( ; k < seCnt ; ++k )
|
|
19 {
|
|
20 //if ( segments[0].b == 110282529 && segCnt == 2 )
|
|
21 // printf( "(%d:%d %d):(%d:%d %d)\n", i, (int)segments[i].a, (int)segments[i].b, k, (int)subexons[k].start, (int)subexons[k].end ) ;
|
|
22 if ( subexons[k].start > segments[i].b )
|
|
23 break ;
|
|
24 if ( segments[i].a > subexons[k].end )
|
|
25 continue ;
|
|
26
|
|
27 int relaxedWidth = 0 ;
|
|
28 if ( ( subexons[k].start >= segments[i].a && subexons[k].end <= segments[i].b )
|
|
29 || ( i == 0 && subexons[k].start - relaxedWidth < segments[i].a && subexons[k].end <= segments[i].b )
|
|
30 || ( i == segCnt - 1 && subexons[k].start >= segments[i].a && subexons[k].end + relaxedWidth > segments[i].b )
|
|
31 || ( i == 0 && i == segCnt - 1 && subexons[k].start - relaxedWidth < segments[i].a && subexons[k].end + relaxedWidth > segments[i].b ) )
|
|
32 {
|
|
33 if ( leftIdx == -1 )
|
|
34 leftIdx = k ;
|
|
35 rightIdx = k ;
|
|
36 ct.vector.Set( k ) ;
|
|
37 }
|
|
38 else
|
|
39 {
|
|
40 return false ;
|
|
41 }
|
|
42 }
|
|
43
|
|
44 if ( leftIdx == -1 )
|
|
45 return false ;
|
|
46
|
|
47 // The cover contradict the boundary.
|
|
48 if ( !( ( subexons[leftIdx].leftType == 0 || subexons[leftIdx].start <= segments[i].a )
|
|
49 && ( subexons[rightIdx].rightType == 0 || subexons[rightIdx].end >= segments[i].b ) ) )
|
|
50 return false ;
|
|
51
|
|
52 // The intron must exists in the subexon graph.
|
|
53 if ( i > 0 )
|
|
54 {
|
|
55 for ( j = 0 ; j < subexons[ ct.last ].nextCnt ; ++j )
|
|
56 if ( subexons[ct.last].next[j] == leftIdx )
|
|
57 break ;
|
|
58 if ( j >= subexons[ ct.last ].nextCnt )
|
|
59 return false ;
|
|
60 }
|
|
61
|
|
62 // The subexons must be consecutive
|
|
63 for ( j = leftIdx + 1 ; j <= rightIdx ; ++j )
|
|
64 if ( subexons[j].start > subexons[j - 1].end + 1 )
|
|
65 return false ;
|
|
66
|
|
67 if ( i == 0 )
|
|
68 ct.first = leftIdx ;
|
|
69 ct.last = rightIdx ;
|
|
70 }
|
|
71 return true ;
|
|
72 }
|
|
73
|
|
74 void Constraints::CoalesceSameConstraints()
|
|
75 {
|
|
76 int i, k ;
|
|
77 int size = constraints.size() ;
|
|
78 for ( i = 0 ; i < size ; ++i )
|
|
79 {
|
|
80 constraints[i].info = i ;
|
|
81 //printf( "constraints %d: %d %d %d\n", i, constraints[i].vector.Test( 0 ), constraints[i].vector.Test(1), constraints[i].support ) ;
|
|
82 }
|
|
83
|
|
84 std::vector<int> newIdx ;
|
|
85 newIdx.resize( size, 0 ) ;
|
|
86
|
|
87 // Update the constraints.
|
|
88 if ( size > 0 )
|
|
89 {
|
|
90 std::sort( constraints.begin(), constraints.end(), CompSortConstraints ) ;
|
|
91
|
|
92 k = 0 ;
|
|
93 newIdx[ constraints[0].info ] = 0 ;
|
|
94 for ( i = 1 ; i < size ; ++i )
|
|
95 {
|
|
96 if ( constraints[k].vector.IsEqual( constraints[i].vector ) )
|
|
97 {
|
|
98 constraints[k].weight += constraints[i].weight ;
|
|
99 constraints[k].support += constraints[i].support ;
|
|
100 constraints[k].uniqSupport += constraints[i].uniqSupport ;
|
|
101 constraints[k].maxReadLen = ( constraints[k].maxReadLen > constraints[i].maxReadLen ) ?
|
|
102 constraints[k].maxReadLen : constraints[i].maxReadLen ;
|
|
103 constraints[i].vector.Release() ;
|
|
104 }
|
|
105 else
|
|
106 {
|
|
107 ++k ;
|
|
108 if ( k != i )
|
|
109 constraints[k] = constraints[i] ;
|
|
110 }
|
|
111 newIdx[ constraints[i].info ] = k ;
|
|
112 }
|
|
113 constraints.resize( k + 1 ) ;
|
|
114 }
|
|
115
|
|
116 // Update the mate pairs.
|
|
117 size = matePairs.size() ;
|
|
118 if ( size > 0 )
|
|
119 {
|
|
120 for ( i = 0 ; i < size ; ++i )
|
|
121 {
|
|
122 //printf( "%d %d: %d => %d | %d =>%d\n", i, newIdx.size(), matePairs[i].i, newIdx[ matePairs[i].i ],
|
|
123 // matePairs[i].j, newIdx[ matePairs[i].j ] ) ;
|
|
124 matePairs[i].i = newIdx[ matePairs[i].i ] ;
|
|
125 matePairs[i].j = newIdx[ matePairs[i].j ] ;
|
|
126 }
|
|
127
|
|
128 std::sort( matePairs.begin(), matePairs.end(), CompSortMatePairs ) ;
|
|
129
|
|
130 k = 0 ;
|
|
131 for ( i = 1 ; i < size ; ++i )
|
|
132 {
|
|
133 if ( matePairs[i].i == matePairs[k].i && matePairs[i].j == matePairs[k].j )
|
|
134 {
|
|
135 matePairs[k].support += matePairs[i].support ;
|
|
136 matePairs[k].uniqSupport += matePairs[i].uniqSupport ;
|
|
137 }
|
|
138 else
|
|
139 {
|
|
140 ++k ;
|
|
141 matePairs[k] = matePairs[i] ;
|
|
142 }
|
|
143 }
|
|
144 //printf( "%s: %d\n", __func__, matePairs[1].i) ;
|
|
145 matePairs.resize( k + 1 ) ;
|
|
146 }
|
|
147
|
|
148 // Update the data structure for future mate pairs.
|
|
149 mateReadIds.UpdateIdx( newIdx ) ;
|
|
150 }
|
|
151
|
|
152 void Constraints::ComputeNormAbund( struct _subexon *subexons )
|
|
153 {
|
|
154 int i, j ;
|
|
155 int ctSize = constraints.size() ;
|
|
156 for ( i = 0 ; i < ctSize ; ++i )
|
|
157 {
|
|
158 // spanned more than 2 subexon
|
|
159 int readLen = constraints[i].maxReadLen ;
|
|
160 if ( constraints[i].first + 1 < constraints[i].last )
|
|
161 {
|
|
162 std::vector<int> subexonInd ;
|
|
163 constraints[i].vector.GetOnesIndices( subexonInd ) ;
|
|
164 int size = subexonInd.size() ;
|
|
165 for ( j = 1 ; j < size - 1 ; ++j )
|
|
166 {
|
|
167 int a = subexonInd[j] ;
|
|
168 readLen -= ( subexons[a].end - subexons[a].start + 1 ) ;
|
|
169 }
|
|
170 }
|
|
171
|
|
172 int effectiveLength ;
|
|
173 if ( constraints[i].first == constraints[i].last )
|
|
174 {
|
|
175 effectiveLength = ( subexons[ constraints[i].first ].end - readLen + 1 )- subexons[ constraints[i].first ].start + 1 ;
|
|
176 if ( effectiveLength <= 0 ) // this happens in the 3',5'-end subexon, where we trimmed the length
|
|
177 effectiveLength = ( subexons[ constraints[i].first ].end - subexons[ constraints[i].first ].start + 1 ) / 2 + 1 ;
|
|
178 }
|
|
179 else
|
|
180 {
|
|
181 int a = constraints[i].first ;
|
|
182 int b = constraints[i].last ;
|
|
183 int start, end ; // the range of the possible start sites of a read in subexons[a].
|
|
184 start = subexons[a].end + 1 - ( readLen - 1 ) ;
|
|
185 if ( start < subexons[a].start )
|
|
186 start = subexons[a].start ;
|
|
187
|
|
188 if ( subexons[b].end - subexons[b].start + 1 >= readLen - 1 || subexons[b].rightType == 0 )
|
|
189 end = subexons[a].end ;
|
|
190 else
|
|
191 {
|
|
192 end = subexons[a].end + 1 - ( readLen - ( subexons[b].end - subexons[b].start + 1 ) ) ;
|
|
193 }
|
|
194
|
|
195 if ( end < start ) // when we trimmed the subexon.
|
|
196 end = subexons[a].start ;
|
|
197
|
|
198 effectiveLength = end - start + 1 ;
|
|
199 }
|
|
200 //printf( "%d: effectiveLength=%d support=%d\n", i, effectiveLength, constraints[i].support ) ;
|
|
201 constraints[i].normAbund = (double)constraints[i].weight / (double)effectiveLength ;
|
|
202
|
|
203 if ( ( subexons[ constraints[i].first ].leftType == 0 && subexons[ constraints[i].first ].end - subexons[ constraints[i].first ].start + 1 >= 8 * pAlignments->readLen )
|
|
204 || ( subexons[ constraints[i].last ].rightType == 0 && subexons[ constraints[i].last ].end - subexons[ constraints[i].last ].start + 1 >= 8 * pAlignments->readLen ) ) // some random elongation of the sequence might make unnecessary long effective length.
|
|
205 {
|
|
206 constraints[i].normAbund *= 2 ;
|
|
207 }
|
|
208 constraints[i].abundance = constraints[i].normAbund ;
|
|
209 }
|
|
210
|
|
211 ctSize = matePairs.size() ;
|
|
212 for ( i = 0 ; i < ctSize ; ++i )
|
|
213 {
|
|
214 double a = constraints[ matePairs[i].i ].normAbund ;
|
|
215 double b = constraints[ matePairs[i].j ].normAbund ;
|
|
216
|
|
217 matePairs[i].normAbund = a < b ? a : b ;
|
|
218
|
|
219 if ( matePairs[i].i != matePairs[i].j )
|
|
220 {
|
|
221 if ( subexons[ constraints[ matePairs[i].i ].first ].leftType == 0
|
|
222 && constraints[ matePairs[i].i ].first == constraints[ matePairs[i].i ].last
|
|
223 && a < b )
|
|
224 {
|
|
225 matePairs[i].normAbund = b ;
|
|
226 }
|
|
227 else if ( subexons[ constraints[ matePairs[i].j ].last ].rightType == 0
|
|
228 && constraints[ matePairs[i].j ].first == constraints[ matePairs[i].j ].last
|
|
229 && a > b )
|
|
230 {
|
|
231 matePairs[i].normAbund = a ;
|
|
232 }
|
|
233 }
|
|
234 //matePairs[i].normAbund = sqrt( a * b ) ;
|
|
235
|
|
236 matePairs[i].abundance = matePairs[i].normAbund ;
|
|
237 }
|
|
238 }
|
|
239
|
|
240 int Constraints::BuildConstraints( struct _subexon *subexons, int seCnt, int start, int end )
|
|
241 {
|
|
242 int i ;
|
|
243 int tag = 0 ;
|
|
244 int coalesceThreshold = 16384 ;
|
|
245 Alignments &alignments = *pAlignments ;
|
|
246 // Release the memory from previous gene.
|
|
247 int size = constraints.size() ;
|
|
248
|
|
249 if ( size > 0 )
|
|
250 {
|
|
251 for ( i = 0 ; i < size ; ++i )
|
|
252 constraints[i].vector.Release() ;
|
|
253 std::vector<struct _constraint>().swap( constraints ) ;
|
|
254 }
|
|
255 std::vector<struct _matePairConstraint>().swap( matePairs ) ;
|
|
256 mateReadIds.Clear() ;
|
|
257
|
|
258 // Start to build the constraints.
|
|
259 bool callNext = false ; // the last used alignment
|
|
260 if ( alignments.IsAtBegin() )
|
|
261 callNext = true ;
|
|
262 while ( !alignments.IsAtEnd() )
|
|
263 {
|
|
264 if ( callNext )
|
|
265 {
|
|
266 if ( !alignments.Next() )
|
|
267 break ;
|
|
268 }
|
|
269 else
|
|
270 callNext = true ;
|
|
271
|
|
272 if ( alignments.GetChromId() < subexons[0].chrId )
|
|
273 continue ;
|
|
274 else if ( alignments.GetChromId() > subexons[0].chrId )
|
|
275 break ;
|
|
276 // locate the first subexon in this region that overlapps with current alignment.
|
|
277 for ( ; tag < seCnt && subexons[tag].end < alignments.segments[0].a ; ++tag )
|
|
278 ;
|
|
279
|
|
280 if ( tag >= seCnt )
|
|
281 break ;
|
|
282 if ( alignments.segments[ alignments.segCnt - 1 ].b < subexons[tag].start )
|
|
283 continue ;
|
|
284
|
|
285 int uniqSupport = 0 ;
|
|
286 if ( usePrimaryAsUnique )
|
|
287 uniqSupport = alignments.IsPrimary() ? 1 : 0 ;
|
|
288 else
|
|
289 uniqSupport = alignments.IsUnique() ? 1 : 0 ;
|
|
290
|
|
291 struct _constraint ct ;
|
|
292 ct.vector.Init( seCnt ) ;
|
|
293 //printf( "%s %d: %lld-%lld | %d-%d\n", __func__, alignments.segCnt, alignments.segments[0].a, alignments.segments[0].b, subexons[tag].start, subexons[tag].end ) ;
|
|
294 ct.weight = 1.0 / alignments.GetNumberOfHits() ;
|
|
295 if ( alignments.IsGCRich() )
|
|
296 ct.weight *= 10 ;
|
|
297 ct.normAbund = 0 ;
|
|
298 ct.support = 1 ;
|
|
299 ct.uniqSupport = uniqSupport ;
|
|
300 ct.maxReadLen = alignments.GetRefCoverLength() ;
|
|
301
|
|
302 if ( alignments.IsPrimary() && ConvertAlignmentToBitTable( alignments.segments, alignments.segCnt,
|
|
303 subexons, seCnt, tag, ct ) )
|
|
304 {
|
|
305
|
|
306 //printf( "%s ", alignments.GetReadId() ) ;
|
|
307 //ct.vector.Print() ;
|
|
308
|
|
309
|
|
310 // If the alignment has clipped end or tail. We only keep those clipped in the 3'/5'-end
|
|
311 bool validClip = true ;
|
|
312 if ( alignments.HasClipHead() )
|
|
313 {
|
|
314 if ( ( ct.first < seCnt - 1 && subexons[ct.first].end + 1 == subexons[ct.first + 1].start )
|
|
315 || subexons[ct.first].prevCnt > 0
|
|
316 || alignments.segments[0].b - alignments.segments[0].a + 1 <= alignments.GetRefCoverLength() / 3.0 )
|
|
317 validClip = false ;
|
|
318 }
|
|
319 if ( alignments.HasClipTail() )
|
|
320 {
|
|
321 int tmp = alignments.segCnt - 1 ;
|
|
322 if ( ( ct.last > 0 && subexons[ct.last].start - 1 == subexons[ct.last - 1].end )
|
|
323 || subexons[ct.last].nextCnt > 0
|
|
324 || alignments.segments[tmp].b - alignments.segments[tmp].a + 1 <= alignments.GetRefCoverLength() / 3.0 )
|
|
325 validClip = false ;
|
|
326 }
|
|
327
|
|
328 if ( validClip )
|
|
329 {
|
|
330 constraints.push_back( ct ) ; // if we just coalesced but the list size does not decrease, this will force capacity increase.
|
|
331 //if ( !strcmp( alignments.GetReadId(), "ERR188021.8489052" ) )
|
|
332 // ct.vector.Print() ;
|
|
333 // Add the mate-pair information.
|
|
334 int mateChrId ;
|
|
335 int64_t matePos ;
|
|
336 alignments.GetMatePosition( mateChrId, matePos ) ;
|
|
337 if ( alignments.GetChromId() == mateChrId )
|
|
338 {
|
|
339 if ( matePos < alignments.segments[0].a )
|
|
340 {
|
|
341 int mateIdx = mateReadIds.Query( alignments.GetReadId(), alignments.segments[0].a ) ;
|
|
342 if ( mateIdx != -1 )
|
|
343 {
|
|
344 struct _matePairConstraint nm ;
|
|
345 nm.i = mateIdx ;
|
|
346 nm.j = constraints.size() - 1 ;
|
|
347 nm.abundance = 0 ;
|
|
348 nm.support = 1 ;
|
|
349 nm.uniqSupport = uniqSupport ;
|
|
350 nm.effectiveCount = 2 ;
|
|
351 matePairs.push_back( nm ) ;
|
|
352 }
|
|
353 }
|
|
354 else if ( matePos > alignments.segments[0].a )
|
|
355 {
|
|
356 mateReadIds.Insert( alignments.GetReadId(), alignments.segments[0].a, constraints.size() - 1, matePos ) ;
|
|
357 }
|
|
358 else // two mates have the same coordinate.
|
|
359 {
|
|
360 if ( alignments.IsFirstMate() )
|
|
361 {
|
|
362 struct _matePairConstraint nm ;
|
|
363 nm.i = constraints.size() - 1 ;
|
|
364 nm.j = constraints.size() - 1 ;
|
|
365 nm.abundance = 0 ;
|
|
366 nm.support = 1 ;
|
|
367 nm.uniqSupport = uniqSupport ;
|
|
368 nm.effectiveCount = 2 ;
|
|
369 matePairs.push_back( nm ) ;
|
|
370 }
|
|
371 }
|
|
372 }
|
|
373 }
|
|
374 else
|
|
375 ct.vector.Release() ;
|
|
376
|
|
377 // Coalesce if necessary.
|
|
378 size = constraints.size() ;
|
|
379 if ( (int)size > coalesceThreshold && size == (int)constraints.capacity() )
|
|
380 {
|
|
381
|
|
382 //printf( "start coalescing. %d\n", constraints.capacity() ) ;
|
|
383 CoalesceSameConstraints() ;
|
|
384
|
|
385 // Not coalesce enough
|
|
386 if ( constraints.size() >= constraints.capacity() / 2 )
|
|
387 {
|
|
388 coalesceThreshold *= 2 ;
|
|
389 }
|
|
390 }
|
|
391 }
|
|
392 else
|
|
393 {
|
|
394 //printf( "not compatible\n" ) ;
|
|
395 ct.vector.Release() ;
|
|
396 }
|
|
397 }
|
|
398 //printf( "start coalescing. %d %d\n", constraints.size(), matePairs.size() ) ;
|
|
399 CoalesceSameConstraints() ;
|
|
400 //printf( "after coalescing. %d %d\n", constraints.size(), matePairs.size() ) ;
|
|
401 //for ( i = 0 ; i < matePairs.size() ; ++i )
|
|
402 // printf( "matePair: %d %d %d\n", matePairs[i].i, matePairs[i].j, matePairs[i].support ) ;
|
|
403 // single-end data set
|
|
404 //if ( matePairs.size() == 0 )
|
|
405 if ( alignments.fragStdev == 0 )
|
|
406 {
|
|
407 int size = constraints.size() ;
|
|
408 matePairs.clear() ;
|
|
409
|
|
410 for ( i = 0 ; i < size ; ++i )
|
|
411 {
|
|
412 struct _matePairConstraint nm ;
|
|
413 nm.i = i ;
|
|
414 nm.j = i ;
|
|
415 nm.abundance = 0 ;
|
|
416 nm.support = constraints[i].support ;
|
|
417 nm.uniqSupport = constraints[i].uniqSupport ;
|
|
418 nm.effectiveCount = 1 ;
|
|
419
|
|
420 matePairs.push_back( nm ) ;
|
|
421 }
|
|
422 }
|
|
423
|
|
424 ComputeNormAbund( subexons ) ;
|
|
425
|
|
426 /*for ( i = 0 ; i < constraints.size() ; ++i )
|
|
427 {
|
|
428 printf( "constraints %d: %lf %d %d %d ", i, constraints[i].normAbund, constraints[i].first, constraints[i].last, constraints[i].support ) ;
|
|
429 constraints[i].vector.Print() ;
|
|
430 }
|
|
431
|
|
432 for ( i = 0 ; i < matePairs.size() ; ++i )
|
|
433 {
|
|
434 printf( "mates %d: %lf %d %d %d %d\n", i, matePairs[i].normAbund, matePairs[i].i, matePairs[i].j, matePairs[i].support, matePairs[i].uniqSupport ) ;
|
|
435 }*/
|
|
436
|
|
437 return 0 ;
|
|
438 }
|