int main( int argc, char *argv[] )
{
int i ;
int ret ;
Alignments alignments ;
Alignments clippedAlignments ;
Blocks blocks ;
Genome genome ;
char *genomeFile = NULL ;
if ( argc < 2 )
{
printf( "%s", usage ) ;
exit( 0 ) ;
}
minimumSupport = 2 ;
minimumEffectiveLength = 200 ;
kmerSize = 23 ;
breakN = 1 ;
minContigSize = 200 ;
prefix = NULL ;
VERBOSE = false ;
outputConnectionSequence = false ;
aggressiveMode = false ;
for ( i = 1 ; i < argc ; ++i )
{
if ( !strcmp( "-b", argv[i] ) )
{
alignments.Open( argv[i + 1]) ;
++i ;
}
else if ( !strcmp( "-o", argv[i] ) )
{
prefix = argv[i + 1] ;
++i ;
}
else if ( !strcmp( "-f", argv[i] ) )
{
genomeFile = argv[i + 1] ;
++i ;
}
else if ( !strcmp( "-ms", argv[i] ) )
{
minimumSupport = atoi( argv[i + 1] ) ;
++i ;
}
else if ( !strcmp( "-ml", argv[i] ) )
{
minimumEffectiveLength = atoi( argv[i + 1] ) ;
++i ;
}
else if ( !strcmp( "-k", argv[i] ) )
{
kmerSize = atoi( argv[i + 1] ) ;
++i ;
}
else if ( !strcmp( "-breakN", argv[i] ) )
{
breakN = atoi( argv[i + 1] ) ;
++i ;
}
else if ( !strcmp( "-minContigSize", argv[i] ) )
{
minContigSize = atoi( argv[i + 1] ) ;
++i ;
}
else if ( !strcmp( "-v", argv[i] ) )
{
VERBOSE = true ;
}
else if ( !strcmp( "-cs", argv[i] ) )
{
outputConnectionSequence = true ;
}
/*else if ( !strcmp( "-aggressive", argv[i] ) )
{
aggressiveMode = true ;
}*/
else if ( !strcmp( "-bc", argv[i] ) )
{
// So far, assume the input is from BWA mem
clippedAlignments.Open( argv[i + 1] ) ;
clippedAlignments.SetAllowSupplementary( true ) ;
++i ;
}
else
{
fprintf( stderr, "Unknown parameter: %s\n", argv[i] ) ;
exit( 1 ) ;
}
}
if ( !alignments.IsOpened() )
{
printf( "Must use -b to specify the bam file." ) ;
return 0 ;
}
if ( prefix != NULL )
{
char buffer[255] ;
sprintf( buffer, "%s.out", prefix ) ;
fpOut = fopen( buffer, "w" ) ;
}
else
{
char buffer[255] ;
prefix = strdup( "rascaf" ) ;
sprintf( buffer, "%s.out", prefix ) ;
fpOut = fopen( buffer, "w" ) ;
}
if ( genomeFile != NULL )
{
genome.Open( alignments, genomeFile ) ;
alignments.Rewind() ;
}
if ( outputConnectionSequence == true && genomeFile == NULL )
{
fprintf( stderr, "Must use -f to specify assembly file when using -cs\n" ) ;
exit( EXIT_FAILURE ) ;
}
// 74619
//printf( "%c\n", genome.GetNucleotide( 74619, 4 ) ) ;
//exit(0) ;
// Build the graph
ret = blocks.BuildExonBlocks( alignments, genome ) ;
alignments.Rewind() ;
fprintf( stderr, "Found %d exon blocks.\n", ret ) ;
if ( clippedAlignments.IsOpened() )
{
fprintf( stderr, "Extend exon blocks with clipped alignments.\n" ) ;
Blocks extendBlocks ;
extendBlocks.BuildExonBlocks( clippedAlignments, genome ) ;
clippedAlignments.Rewind() ;
ret = blocks.ExtendExonBlocks( extendBlocks ) ;
fprintf( stderr, "Found %d exon blocks after extension.\n", ret ) ;
}
blocks.GetAlignmentsInfo( alignments ) ;
alignments.Rewind() ;
ret = blocks.BuildGeneBlocks( alignments, genome ) ;
alignments.Rewind() ;
fprintf( stderr, "Found %d gene blocks.\n", ret ) ;
blocks.BuildGeneBlockGraph( alignments ) ;
if ( clippedAlignments.IsOpened() )
{
blocks.AddGeneBlockGraphByClippedAlignments( clippedAlignments ) ;
}
// Cleaning
blocks.CleanGeneBlockGraph( alignments, genome ) ;
// Scaffolding
Scaffold scaffold( blocks, genome ) ;
//scaffold.Init( blocks ) ;
int componentCnt = scaffold.BuildComponent() ;
fprintf( stderr, "Found %d non-trivial gene block components.\n", componentCnt ) ;
// Possible for parallelization
for ( i = 0 ; i < componentCnt ; ++i )
{
scaffold.ScaffoldComponent( i ) ;
}
scaffold.ScaffoldGenome() ;
// Output the command line
fprintf( fpOut, "command line: " ) ;
char *fullpath = (char *)malloc( sizeof( char ) * 4096 ) ;
for ( i = 0 ; i < argc ; ++i )
{
char c = ' ' ;
if ( i == argc - 1 )
c = '\n' ;
if ( i > 0 && !strcmp( argv[i - 1], "-b" ) )
{
if ( realpath( argv[i], fullpath ) == NULL )
{
fprintf( stderr, "Failed to resolve the path of file %s.\n", argv[i] ) ;
exit( 1 ) ;
}
fprintf( fpOut, "%s%c", fullpath, c ) ;
}
else if ( i > 0 && !strcmp( argv[i - 1], "-f" ) )
{
if ( realpath( argv[i], fullpath ) == NULL )
{
fprintf( stderr, "Failed to resolve the path of file %s.\n", argv[i] ) ;
exit( 1 ) ;
}
fprintf( fpOut, "%s%c", fullpath, c ) ;
}
else
fprintf( fpOut, "%s%c", argv[i], c ) ;
}
free( fullpath ) ;
scaffold.Output( fpOut, alignments ) ;
return 0 ;
}
int main( int argc, char *argv[] )
{
Alignments alignments ;
Genome genome ;
std::vector<int> rascafFileId ;
char line[2048] ;
char prefix[512] = "rascaf_scaffold" ;
int rawAssemblyInd = 1 ;
FILE *rascafFile ;
bool contigLevel = false ;
int i ;
FILE *outputFile ;
FILE *infoFile ;
breakN = 1 ;
if ( argc < 2 )
{
fprintf( stderr, "%s", usage ) ;
exit( 1 ) ;
}
for ( i = 1 ; i < argc ; ++i )
{
if ( !strcmp( "-o", argv[i] ) )
{
strcpy( prefix, argv[i + 1 ] ) ;
++i ;
}
else if ( !strcmp( "-ms", argv[i] ) )
{
minSupport = atoi( argv[i + 1] ) ;
++i ;
}
else if ( !strcmp( "-ignoreGap", argv[i] ) )
{
ignoreGap = true ;
}
else if ( !strcmp( "-r", argv[i] ) )
{
rascafFileId.push_back( i + 1 ) ;
++i ;
}
else
{
fprintf( stderr, "Unknown option: %s\n", argv[i] ) ;
exit( EXIT_FAILURE ) ;
}
}
if ( rascafFileId.size() == 0 )
{
fprintf( stderr, "Must use -r to specify rascaf output file.\n" ) ;
exit( EXIT_FAILURE ) ;
}
MAX_NEIGHBOR = 1 + rascafFileId.size() ;
// Get the bam file.
rascafFile = fopen( argv[ rascafFileId[0] ], "r" ) ;
while ( fgets( line, sizeof( line ), rascafFile ) != NULL )
{
if ( strstr( line, "command line:" ) )
{
char *p ;
char buffer[512] ;
p = strstr( line, "-breakN" ) ;
if ( p != NULL )
{
p += 7 ;
while ( *p == ' ' )
++p ;
for ( i = 0 ; *p && *p != ' ' ; ++p, ++i )
buffer[i] = *p ;
buffer[i] = '\0' ;
breakN = atoi( buffer ) ;
}
p = strstr( line, "-b" ) ;
if ( p == NULL )
{
fprintf( stderr, "Could not find the bam file specified by -b in Rascaf.\n" ) ;
exit( 1 ) ;
}
p += 2 ;
while ( *p == ' ' )
++p ;
for ( i = 0 ; *p && *p != ' ' ; ++p, ++i )
buffer[i] = *p ;
buffer[i] = '\0' ;
alignments.Open( buffer ) ;
p = strstr( line, "-f") ;
if ( p == NULL )
{
fprintf( stderr, "Could not find the raw assembly file specified by -f in Rascaf.\n" ) ;
exit( 1 ) ;
}
p += 2 ;
while ( *p == ' ' )
++p ;
for ( i = 0 ; *p && *p != ' ' && *p != '\n' ; ++p, ++i )
buffer[i] = *p ;
buffer[i] = '\0' ;
fprintf( stderr, "Found raw assembly file: %s\n", buffer ) ;
genome.Open( alignments, buffer ) ;
break ;
}
}
fclose( rascafFile ) ;
// Parse the input.
for ( unsigned int fid = 0 ; fid < rascafFileId.size() ; ++fid )
{
rascafFile = fopen( argv[ rascafFileId[fid] ], "r" ) ;
bool start = false ;
int tag ;
while ( fgets( line, sizeof( line ), rascafFile ) != NULL )
{
if ( strstr( line, "command line:" ) )
{
start = true ;
if ( strstr( line, "-f" ) )
{
contigLevel = true ;
}
continue ;
}
if ( !start )
continue ;
if ( !strcmp( line, "WARNINGS:\n" ) )
break ;
std::vector<struct _part> nparts ;
if ( line[0] >= '0' && line[0] <= '9' )
{
AddConnection( line, alignments, nparts ) ;
connects.push_back( nparts ) ;
tag = 0 ;
}
else if ( line[0] == '\t' || line[0] == ' ' )
{
// Break the nparts if the support is too low.
int num = 0 ;
for ( i = 0 ; line[i] < '0' || line[i] > '9' ; ++i )
;
for ( ; line[i] >= '0' && line[i] <= '9' ; ++i )
num = num * 10 + line[i] - '0' ;
++tag ;
if ( num < minSupport )
{
nparts = connects.back() ;
connects.pop_back() ;
int size = nparts.size() ;
std::vector<struct _part> newNParts ;
for ( i = 0 ; i < tag ; ++i )
newNParts.push_back( nparts[i] ) ;
if ( newNParts.size() > 1 )
connects.push_back( newNParts ) ;
newNParts.clear() ;
for ( ; i < size ; ++i )
newNParts.push_back( nparts[i] ) ;
if ( newNParts.size() > 1 )
connects.push_back( newNParts ) ;
tag = 0 ;
}
}
}
fclose( rascafFile ) ;
}
if ( contigLevel == false )
{
genome.SetIsOpen( contigLevel ) ;
}
// Build the graph
int contigCnt = genome.GetContigCount() ;
int edgeCnt = 0 ;
int csize = connects.size() ;
for ( i = 0 ; i < csize ; ++i )
edgeCnt += connects[i].size() ;
ContigGraph contigGraph( contigCnt, contigCnt + edgeCnt ) ;
for ( i = 0 ; i < contigCnt - 1 ; ++i )
{
if ( genome.GetChrIdFromContigId( i ) == genome.GetChrIdFromContigId( i + 1 ) )
{
contigGraph.AddEdge( i, 1, i + 1, 0 ) ;
}
}
for ( i = 0 ; i < csize ; ++i )
{
std::vector<struct _part> &parts = connects[i] ;
int size = parts.size() ;
for ( int j = 0 ; j < size - 1 ; ++j )
{
struct _part &a = parts[j] ;
struct _part &b = parts[j + 1] ;
// Two dummy nodes for each contig. Left is 0, right is 1
int dummyU = 0 ;
int dummyV = 0 ;
if ( a.strand == '+' )
dummyU = 1 ;
if ( b.strand == '-' )
dummyV = 1 ;
contigGraph.AddEdge( a.contigId, dummyU, b.contigId, dummyV, true ) ;
}
}
// Check the cycles in the contig graph. This may introduces when combining different rascaf outputs.
int *visitTime = new int[contigCnt] ;
struct _pair *neighbors = new struct _pair[ MAX_NEIGHBOR ] ;
bool *isInCycle = new bool[contigCnt] ;
std::vector<int> cycleNodes ;
memset( visitTime, -1, sizeof( int ) * contigCnt ) ;
memset( isInCycle, false, sizeof( bool ) * contigCnt ) ;
for ( i = 0 ; i < contigCnt ; ++i )
{
if ( isInCycle[i] )
continue ;
if ( contigGraph.IsInCycle( i, cycleNodes, visitTime ) )
{
int cnt = cycleNodes.size() ;
//printf( "===\n") ;
for ( int j = 0 ; j < cnt ; ++j )
{
//printf( "In cycle %d\n", cycleNodes[j] ) ;
isInCycle[ cycleNodes[j] ] = true ;
}
}
}
//exit( 1 ) ;
// Remove the connected edges involving the nodes in the cycle
for ( i = 0 ; i < contigCnt ; ++i )
{
if ( isInCycle[i] )
{
for ( int dummy = 0 ; dummy <= 1 ; ++dummy )
{
int ncnt = contigGraph.GetNeighbors( i, dummy, neighbors, MAX_NEIGHBOR ) ;
for ( int j = 0 ; j < ncnt ; ++j )
{
if ( neighbors[j].a == i + 2 * dummy - 1 && neighbors[j].b != dummy
&& genome.GetChrIdFromContigId( i ) == genome.GetChrIdFromContigId( neighbors[j].a ) )
continue ; // the connection created by the raw assembly
else
contigGraph.RemoveEdge( i, dummy, neighbors[j].a, neighbors[j].b ) ;
}
}
}
}
//delete[] isInCycle ;
//printf( "hi: %d %d\n", __LINE__, contigCnt ) ;
//printf( "%d %d\n", contigGraph.GetNeighbors( 0, 0, neighbors, MAX_NEIGHBOR ), contigGraph.GetNeighbors( 0, 1, neighbors, MAX_NEIGHBOR ) ) ;
// Sort the scaffolds from fasta file, so that longer scaffold come first
int scafCnt = genome.GetChrCount() ;
struct _pair *scafInfo = new struct _pair[scafCnt] ;
memset( scafInfo, -1, sizeof( struct _pair) * scafCnt ) ;
for ( i = 0 ; i < contigCnt ; ++i )
{
int chrId = genome.GetChrIdFromContigId( i ) ;
if ( scafInfo[chrId].a == -1 )
{
scafInfo[ chrId ].a = i ;
scafInfo[ chrId ].b = genome.GetChrLength( chrId ) ;
}
}
qsort( scafInfo, scafCnt, sizeof( struct _pair ), CompScaffold ) ;
// Merge the branches and build the scaffold
ContigGraph scaffold( contigCnt, 2 * contigCnt ) ;
// Use a method similar to topological sort
bool *used = new bool[contigCnt] ;
int *degree = new int[2 *contigCnt] ;
int *danglingVisitTime = new int[contigCnt] ;
int *counter = new int[contigCnt] ;
int *visitDummy = new int[ contigCnt ] ;
int *buffer = new int[contigCnt] ;
int *buffer2 = new int[contigCnt] ;
bool *isInQueue = new bool[ contigCnt ] ;
int *chosen = new int[contigCnt] ;
int chosenCnt ;
memset( isInCycle, false, sizeof( bool ) * contigCnt ) ;
memset( visitTime, -1, sizeof( int ) * contigCnt ) ;
memset( visitDummy, -1, sizeof( int ) * contigCnt ) ;
memset( counter, -1, sizeof( int ) * contigCnt ) ;
// Use those memory to remove triangular cycles
for ( i = 0 ; i < scafCnt ; ++i )
{
int from, to ;
if ( scafInfo[i].a == -1 )
continue ;
genome.GetChrContigRange( genome.GetChrIdFromContigId( scafInfo[i].a ), from, to ) ;
ForwardSearch( from, 0, i, visitTime, counter, visitDummy, contigGraph ) ;
chosenCnt = 0 ;
BackwardSearchForTriangularCycle( to, 1, i, visitTime, counter, visitDummy, contigGraph, chosen, chosenCnt ) ;
for ( int j = 0 ; j < chosenCnt ; ++j )
{
//printf( "%d\n", chosen[j] ) ;
isInCycle[ chosen[j] ] = true ;
}
}
for ( i = 0 ; i < contigCnt ; ++i )
{
if ( isInCycle[i] )
{
for ( int dummy = 0 ; dummy <= 1 ; ++dummy )
{
int ncnt = contigGraph.GetNeighbors( i, dummy, neighbors, MAX_NEIGHBOR ) ;
for ( int j = 0 ; j < ncnt ; ++j )
{
if ( neighbors[j].a == i + 2 * dummy - 1 && neighbors[j].b != dummy
&& genome.GetChrIdFromContigId( i ) == genome.GetChrIdFromContigId( neighbors[j].a ) )
continue ; // the connection created by the raw assembly
else
contigGraph.RemoveEdge( i, dummy, neighbors[j].a, neighbors[j].b ) ;
}
}
}
}
memset( used, false, sizeof( bool ) * contigCnt ) ;
memset( visitTime, -1, sizeof( int ) * contigCnt ) ;
memset( visitDummy, -1, sizeof( int ) * contigCnt ) ;
memset( danglingVisitTime, -1, sizeof( int ) * contigCnt ) ;
memset( counter, -1, sizeof( int ) * contigCnt ) ;
memset( isInQueue, false, sizeof( bool ) * contigCnt ) ;
ContigGraph newGraph( contigCnt, edgeCnt ) ;
// Compute the gap size
int *gapSize = new int[contigCnt] ;
for ( i = 0 ; i < contigCnt - 1 ; ++i )
{
if ( genome.GetChrIdFromContigId( i ) == genome.GetChrIdFromContigId( i + 1 ) )
{
struct _contig c1 = genome.GetContigInfo( i ) ;
struct _contig c2 = genome.GetContigInfo( i + 1 ) ;
gapSize[i] = c2.start - c1.end - 1 ;
}
else
gapSize[i] = -1 ;
}
// Start search
int ncnt ;
struct _pair *queue = new struct _pair[ contigCnt ] ;
int head = 0, tail ;
int danglingTime = 0 ;
// Pre-allocate the subgraph.
ContigGraph subgraph( contigCnt, 3 * contigCnt ) ;
for ( i = 0 ; i < scafCnt ; ++i )
{
//if ( used[144281] == true )
// printf( "changed %d %d\n", i, scafInfo[i - 1].a ) ;
if ( scafInfo[i].a == -1 )
continue ;
int from, to ;
genome.GetChrContigRange( genome.GetChrIdFromContigId( scafInfo[i].a ), from, to ) ;
//printf( "%d: %d %d %d\n", i, scafInfo[i].b, from, to ) ;
ForwardSearch( from, 0, i, visitTime, counter, visitDummy, contigGraph ) ;
chosenCnt = 0 ;
BackwardSearch( to, 1, i, visitTime, counter, contigGraph, chosen, chosenCnt ) ;
/*printf( "%s %d (%d %d) %d\n", alignments.GetChromName( genome.GetChrIdFromContigId( scafInfo[i].a ) ), i, from, to, chosenCnt ) ;
if ( chosenCnt > 1 )
{
printf( "=== " ) ;
for ( int j = 0 ; j < chosenCnt ; ++j )
printf( "%d ", chosen[j] ) ;
printf( "\n" ) ;
}*/
for ( int j = 0 ; j < chosenCnt ; ++j )
{
ncnt = contigGraph.GetNeighbors( chosen[j], 0, neighbors, MAX_NEIGHBOR ) ;
//printf( "%d %d %d: %d %d %d\n", j, chosen[j], ncnt, neighbors[0].a, visitTime[ neighbors[0].a ],
// counter[neighbors[0].a ] ) ;
for ( int k = 0 ; k < ncnt ; ++k )
{
//if ( i == 639 )
// printf( "Neighbor from 0 %d: %d %d\n", k, neighbors[k].a, neighbors[k].b ) ;
if ( visitTime[ neighbors[k].a ] == 2 * i + 1 && counter[neighbors[k].a ] == 2 )
{
subgraph.AddEdge( chosen[j], 0, neighbors[k].a, neighbors[k].b, true ) ;
//printf( "subgraph: (%d %d)=>(%d %d)\n", chosen[j], 0, neighbors[k].a, neighbors[k].b ) ;
}
}
ncnt = contigGraph.GetNeighbors( chosen[j], 1, neighbors, MAX_NEIGHBOR ) ;
for ( int k = 0 ; k < ncnt ; ++k )
{
//if ( i == 639 )
// printf( "Neighbor from 1 %d: %d %d\n", k, neighbors[k].a, neighbors[k].b ) ;
if ( visitTime[ neighbors[k].a ] == 2 * i + 1 && counter[neighbors[k].a ] == 2 )
{
subgraph.AddEdge( chosen[j], 1, neighbors[k].a, neighbors[k].b, true ) ;
//printf( "subgraph: (%d %d)=>(%d %d)\n", chosen[j], 1, neighbors[k].a, neighbors[k].b ) ;
}
}
}
// Initialize the degree counter
for ( int j = 0 ; j < chosenCnt ; ++j )
{
for ( int l = 0 ; l < 2 ; ++l )
{
/*if ( i == 6145 )
{
std::vector<struct _pair> neighbors ;
ncnt = subgraph.GetNeighbors( chosen[j], l, neighbors ) ;
printf( "%d ncnt=%d\n", l, ncnt ) ;
}*/
ncnt = subgraph.GetNeighbors( chosen[j], l, neighbors, MAX_NEIGHBOR ) ;
degree[ 2 * chosen[j] + l ] = ncnt ;
}
}
// "topological" sort
head = 0 ;
isInQueue[from] = true ;
queue[0].a = from ;
queue[0].b = 0 ;
tail = 1 ;
int prevTag = -1 ;
int *prevAdd = buffer ; // reuse counter to save some memory.
int *nextAdd = buffer2 ;
int firstAdd = -1 ;
while ( head < tail )
{
int tailTag = tail ;
for ( int j = head ; j < tailTag ; ++j )
{
nextAdd[j] = -1 ;
if ( !used[ queue[j].a ] )
{
used[ queue[j].a ] = true ;
if ( prevTag != -1 )
{
scaffold.AddEdge( queue[ prevTag].a, 1 - queue[prevTag].b, queue[j].a, queue[j].b ) ;
nextAdd[ prevTag ] = j ;
/*if ( i == 639 )
printf( "(%lld %lld)=>(%lld %lld)\n", queue[ prevTag].a, 1 - queue[prevTag].b, queue[j].a, queue[j].b ) ;*/
}
else
firstAdd = j ;
prevTag = j ;
}
prevAdd[j] = prevTag ; // the most recent(<=) queue id when added to scaffold.
ncnt = subgraph.GetNeighbors( queue[j].a, 1 - queue[j].b, neighbors, MAX_NEIGHBOR ) ;
for ( int k = 0 ; k < ncnt ; ++k )
{
--degree[ 2 * neighbors[k].a + neighbors[k].b ] ;
if ( degree[ 2 * neighbors[k].a + neighbors[k].b ] == 0 && !isInQueue[neighbors[k].a] )
{
isInQueue[ neighbors[k].a ] = true ;
queue[ tail ] = neighbors[k] ; // Interesting assignment, I think.
++tail ;
/*if ( i == 639 )
printf( "pushed in queue: %d\n", neighbors[k].a ) ;*/
// Put the consecutive contigs together.
struct _pair testNeighbors[ MAX_NEIGHBOR ] ;
struct _pair tag ;
tag = neighbors[k] ;
while ( 1 )
{
if ( contigGraph.GetNeighbors( tag.a, 1 - tag.b, testNeighbors, MAX_NEIGHBOR ) != 1 )
break ;
int n = subgraph.GetNeighbors( tag.a, 1 - tag.b, testNeighbors, MAX_NEIGHBOR ) ;
if ( n != 1 )
break ;
//printf( "%d %d\n", n, testNeighbors[0].a ) ;
struct _pair backNeighbors[ MAX_NEIGHBOR ] ;
if ( contigGraph.GetNeighbors( testNeighbors[0].a, testNeighbors[0].b, backNeighbors, MAX_NEIGHBOR ) != 1 )
break ;
n = subgraph.GetNeighbors( testNeighbors[0].a, testNeighbors[0].b, backNeighbors, MAX_NEIGHBOR ) ;
if ( n != 1 )
break ;
isInQueue[ testNeighbors[0].a ] = true ;
queue[tail] = testNeighbors[0] ;
++tail ;
/*if ( i == 639 )
printf( "pushed in queue: %d\n", testNeighbors[0].a ) ;*/
tag = testNeighbors[0] ;
}
}
}
}
head = tailTag ;
}
// Remove the effect on the subgraph.
/*if ( tail != chosenCnt )
{
printf( "WARNING: not matched\n" ) ;
exit( 1 ) ;
}*/
for ( int j = 0 ; j < tail ; ++j )
{
visitDummy[ queue[j].a ] = -1 ;
counter[ queue[j].a ] = -1 ;
subgraph.RemoveAdjacentEdges( queue[j].a ) ;
isInQueue[ queue[j].a ] = false ;
}
subgraph.ResetEdgeUsed() ;
// no point is picked
if ( prevTag == -1 )
{
continue ;
}
// Update the gap size
prevTag = -1 ;
for ( int j = 0 ; j < tail - 1 ; ++j )
{
if ( genome.GetChrIdFromContigId( queue[j].a ) == genome.GetChrIdFromContigId( from ) )
prevTag = queue[j].a ;
else if ( prevTag != -1 )
{
struct _contig c = genome.GetContigInfo( queue[j].a ) ;
gapSize[prevTag] -= ( c.end - c.start + 1) ;
}
}
// Add the dangling contigs. Use the fact that the queue holding the contigs in the same order as in the scaffold.
// 5'->3' dangling
int *chosenDummy = degree ;
for ( int j = tail - 1 ; j >= 0 ; --j )
{
//if ( j < tail - 1 )
// continue ;
chosenCnt = 0 ;
//if ( queue[j].a == 0 )
// printf( "Dummy: %d %d %d\n", j, queue[j].b, 1 - queue[j].b ) ;
SearchDangling( queue[j].a, queue[j].b, used, danglingTime, danglingVisitTime, contigGraph, false, chosen, chosenDummy, chosenCnt, genome ) ;
++danglingTime ;
int prevTag = prevAdd[j] ;
/*if ( queue[j].a == 0 )
{
struct _pair neighbors[5] ;
int ncnt = contigGraph.GetNeighbors( queue[j].a, 1 - queue[j].b, neighbors, 5 ) ;
printf( "%d %d %d %d: %d %d\n", queue[j].b, chosenCnt, prevTag, ncnt, neighbors[0].a, used[ neighbors[0].a ] ) ;
}*/
if ( prevTag == -1 )
break ;
// Trim the dangling list
int k = chosenCnt - 1 ;
if ( j > 0 && j < tail - 1 )
{
for ( k = chosenCnt - 1 ; k >= 1 ; --k )
if ( genome.GetChrIdFromContigId( chosen[k] ) != genome.GetChrIdFromContigId( chosen[k - 1] ) )
break ;
}
// Test the gap size
int len = 0 ;
for ( int l = 0 ; l <= k ; ++l )
{
struct _contig c = genome.GetContigInfo( chosen[k] ) ;
len += c.end - c.start + 1 ;
}
if ( j < tail - 1 )
{
int l ;
for ( l = j ; l >= 0 ; --l )
if ( genome.GetChrIdFromContigId( queue[l].a ) == genome.GetChrIdFromContigId( from ) )
break ;
if ( !ignoreGap && len >= gapSize[ queue[l].a ] + 100 )
continue ;
else
gapSize[ queue[l].a ] -= len ;
}
for ( ; k >= 0 ; --k )
{
used[ chosen[k] ] = true ;
//printf( "Dangling 1: %d=>%d\n", queue[prevTag].a, chosen[k] ) ;
scaffold.InsertNode( queue[ prevTag ].a, 1 - queue[ prevTag ].b, chosen[k], chosenDummy[k] ) ;
}
}
// 3'->5' dangling
for ( int j = 0 ; j < tail ; ++j )
{
//if ( j > 0 )
// continue ;
chosenCnt = 0 ;
SearchDangling( queue[j].a, 1 - queue[j].b, used, danglingTime, danglingVisitTime, contigGraph, false, chosen, chosenDummy, chosenCnt, genome ) ;
++danglingTime ;
int prevTag = prevAdd[j] ;
int nextTag ;
if ( prevTag == -1 || j <= firstAdd )
nextTag = firstAdd ;
else if ( j == prevTag )
nextTag = j ;
else
nextTag = nextAdd[ prevTag ] ;
if ( nextTag == -1 )
break ;
/*if ( queue[j].a == 37549 )
{
struct _pair neighbors[5] ;
int ncnt = contigGraph.GetNeighbors( queue[j].a, queue[j].b, neighbors, 5 ) ;
fprintf( stderr, "%d %d %d: %d %d %d: %d %d %d\n", j, queue[j].a, queue[j].b, chosenCnt, nextTag, ncnt, chosen[0], chosenDummy[0], used[ chosen[0] ] ) ;
}*/
// trim the danling list
int k = chosenCnt - 1 ;
if ( j < tail - 1 && j > 0 )
{
for ( k = chosenCnt - 1 ; k >= 1 ; --k )
if ( genome.GetChrIdFromContigId( chosen[k] ) != genome.GetChrIdFromContigId( chosen[k - 1] ) )
break ;
}
// Test the gap size
int len = 0 ;
for ( int l = 0 ; l <= k ; ++l )
{
struct _contig c = genome.GetContigInfo( chosen[k] ) ;
len += c.end - c.start + 1 ;
}
if ( j > 0 )
{
int l ;
for ( l = j - 1 ; l >= 0 ; --l ) // Notice the j-1 here, because we want the gap strictly before current contig
if ( genome.GetChrIdFromContigId( queue[l].a ) == genome.GetChrIdFromContigId( from ) )
break ;
if ( !ignoreGap && len >= gapSize[ queue[l].a ] + 100 )
continue ;
else
gapSize[ queue[l].a ] -= len ;
}
for ( ; k >= 0 ; --k )
{
used[ chosen[k] ] = true ;
scaffold.InsertNode( queue[nextTag].a, queue[nextTag].b, chosen[k], chosenDummy[k] ) ;
//printf( "Dangling 2: %d<=%d\n", queue[nextTag].a, chosen[k] ) ;
//if ( chosen[k] == 10246 )
// printf( "hi %d %d %d %d\n", j, queue[j].a, k, chosen[k] ) ;
}
}
}
//return 0 ;
// Output the scaffold
int id = 0 ;
char infoFileName[512] ;
char outputFileName[512] ;
sprintf( infoFileName, "%s.info", prefix ) ;
sprintf( outputFileName, "%s.fa", prefix ) ;
outputFile = fopen( outputFileName, "w" ) ;
infoFile = fopen( infoFileName, "w") ;
memset( used, false, sizeof( bool ) * contigCnt ) ;
for ( i = 0 ; i < contigCnt ; ++i )
{
//printf( "%d (%s)\n", i, alignments.GetChromName( genome.GetChrIdFromContigId( i ) ) ) ; fflush( stdout ) ;
/*if ( i == 10246 )
{
std::vector<struct _pair> neighbors ;
scaffold.GetNeighbors( i, 0, neighbors ) ;
printf( "%u\n", neighbors.size() ) ;
}*/
if ( used[i] )
continue ;
int ncnt1 = scaffold.GetNeighbors( i, 0, neighbors, MAX_NEIGHBOR ) ;
int ncnt2 = scaffold.GetNeighbors( i, 1, neighbors, MAX_NEIGHBOR ) ;
if ( ncnt1 == 0 || ncnt2 == 0 ) // The end of a scaffold
{
fprintf( outputFile, ">scaffold_%d\n", id) ;
fprintf( infoFile, ">scaffold_%d", id ) ;
++id ;
int p = i ;
int dummyP = 1 ;
if ( ncnt1 == 0 )
dummyP = 0 ;
used[i] = true ;
genome.PrintContig( outputFile, i, dummyP ) ;
fprintf( infoFile, " (%s %d %c)", alignments.GetChromName( genome.GetChrIdFromContigId( p ) ), p, dummyP == 0 ? '+' : '-' ) ;
while ( 1 )
{
ncnt = scaffold.GetNeighbors( p, 1 - dummyP, neighbors, MAX_NEIGHBOR ) ;
if ( ncnt == 0 )
break ;
// ncnt must be 1
int insertN = 17 ;
if ( genome.GetChrIdFromContigId( p ) == genome.GetChrIdFromContigId( neighbors[0].a ) )
{
struct _contig cp, cna ;
cp = genome.GetContigInfo( p ) ;
cna = genome.GetContigInfo( neighbors[0].a ) ;
if ( p < neighbors[0].a )
insertN = cna.start - cp.end - 1 ;
else if ( p > neighbors[0].a )
insertN = cp.start - cna.end - 1 ;
}
p = neighbors[0].a ;
dummyP = neighbors[0].b ;
for ( int j = 0 ; j < insertN ; ++j )
fprintf( outputFile, "N" ) ;
used[p] = true ;
genome.PrintContig( outputFile, p, dummyP ) ;
fprintf( infoFile, " (%s %d %c)", alignments.GetChromName( genome.GetChrIdFromContigId( p ) ), p, dummyP == 0 ? '+' : '-' ) ;
}
fprintf( outputFile, "\n" ) ;
fprintf( infoFile, "\n" ) ;
}
}
for ( i = 0 ; i < contigCnt ; ++i )
if ( !used[i] )
{
fprintf( stderr, "Unreported contig %d.\n", i ) ;
}
fclose( outputFile ) ;
fclose( infoFile ) ;
delete[] buffer ;
delete[] buffer2 ;
delete[] chosen ;
delete[] queue ;
delete[] counter ;
delete[] visitTime ;
delete[] used ;
delete[] scafInfo ;
delete[] isInQueue ;
delete[] gapSize ;
//fclose( rascafFile ) ;
return 0 ;
}
int main( int argc, char *argv[] )
{
int i, j, k ;
FILE *fp ;
char buffer[10100] ;
std::vector<std::string> fields ;
int binSize = 50 ;
const char *backboneName = NULL ;
// the compatilibity between the alignment and allele
// The likelihood of this read from this allele
struct _compatible **compatibility ;
int **snpAllele ; // whether a snp showed up in the allele.
Alignments alignments ;
alignments.Open( argv[2] ) ;
for ( i = 3 ; i < argc ; ++i )
{
if ( !strcmp( argv[i], "-b" ) )
{
backboneName = argv[i + 1] ;
alignments.OnlyChrom( backboneName ) ;
++i ;
}
else
{
fprintf( stderr, "Unknown argument %s.\n", argv[i] ) ;
exit( 1 ) ;
}
}
// Parse the files associate with the snps
// Firstly, read in the snp list have the information
sprintf( buffer, "%s.snp", argv[1] ) ;
fp = fopen( buffer, "r" ) ;
k = 0 ;
while ( fgets( buffer, sizeof( buffer ), fp ) )
{
Split( buffer, '\t', fields ) ;
if ( backboneName && strcmp( fields[2].c_str(), backboneName ) )
continue ;
snpNameToId[ fields[0] ] = k ;
struct _snpInfo info ;
info.type = fields[1][0] ;
if ( info.type == 'd' )
{
info.position = atoi( fields[3].c_str() ) ;
info.length = atoi( fields[4].c_str() ) ;
}
else if ( info.type == 'i' )
{
info.position = atoi( fields[3].c_str() ) ;
info.length = strlen( fields[4].c_str() ) ;
}
else
{
info.position = atoi( fields[3].c_str() ) ; // notice that the snp file is 0-based index.
info.length = 1 ;
info.nucleotide = fields[4][0] ;
}
snpInfo.push_back( info ) ;
std::vector< int > tmpList ;
snpLink.push_back( tmpList ) ;
for ( int p = 0 ; p < info.length ; ++p )
{
if ( info.type != 'i' || p == 0 )
{
if ( positionToSnp.find( info.position + p ) == positionToSnp.end() )
{
positionToSnp[ info.position + p] = tmpList ;
}
positionToSnp[ info.position + p].push_back( k ) ;
}
}
++k ;
}
fclose( fp ) ;
// Read in the link file. Determine the id of alleles and the association
// of alleles and snps.
// TODO: obtain the length of each allele and take the length into account in the statistical model
// Add the id for the backbound
int backboneLength = 0 ;
sprintf( buffer, "%s_backbone.fa", argv[1] ) ;
fp = fopen( buffer, "r" ) ;
/*for ( i = 1 ; buffer[i] && buffer[i] != ' ' && buffer[i] != '\n' ; ++i )
;
buffer[i] = '\0' ;
std::string backboneName( buffer + 1 ) ;
alleleNameToId[ backboneName ] = 0 ;
alleleIdToName.push_back( backboneName ) ;*/
bool start = false ;
while ( fgets( buffer, sizeof( buffer ), fp ) )
{
if ( buffer[0] == '>' )
{
for ( i = 1 ; buffer[i] && buffer[i] != ' ' && buffer[i] != '\n' ; ++i )
;
buffer[i] = '\0' ;
if ( !strcmp( backboneName, buffer + 1 ) )
{
start = true ;
}
else if ( start )
break ;
}
if ( start && buffer[0] != '>' )
{
int len = strlen( buffer ) ;
if ( buffer[len - 1 ] == '\n' )
backboneLength += len - 1 ;
else
backboneLength += len ;
}
}
fclose( fp ) ;
/*k = 0 ;
if ( k == 0 )
{
std::vector<int> tmpList ;
alleleSnpList.push_back( tmpList ) ;
alleleLength.push_back( backboneLength ) ;
}*/
// scanning the link file
sprintf( buffer, "%s.link", argv[1] ) ;
fp = fopen( buffer, "r" ) ;
k = 0 ;
while ( fgets( buffer, sizeof( buffer ), fp ) )
{
std::vector<std::string> tmpFields ;
Split( buffer, '\t', tmpFields ) ;
// skip the snps from other backbones
if ( snpNameToId.find( tmpFields[0] ) == snpNameToId.end() )
continue ;
int snpId = snpNameToId[ tmpFields[0] ] ;
Split( tmpFields[1].c_str(), ' ', fields ) ;
int size = fields.size() ;
for ( i = 0 ; i < size ; ++i )
{
if ( alleleNameToId.find( fields[i] ) == alleleNameToId.end() )
{
//printf( "%s %d\n", fields[i].c_str(), k ) ;
alleleNameToId[ fields[i] ] = k ;
alleleIdToName.push_back( fields[i] ) ;
std::vector<int> tmpList ;
alleleSnpList.push_back( tmpList ) ;
alleleLength.push_back( backboneLength ) ;
++k ;
}
int alleleId = alleleNameToId[ fields[i] ] ;
//if ( snpId == 118 )
// printf( "%s: %s %d\n", tmpFields[0].c_str(), fields[i].c_str(), alleleId ) ;
snpLink[ snpId ].push_back( alleleId ) ;
alleleSnpList[ alleleId ].push_back( snpId ) ;
if ( snpInfo[ snpId ].type == 'd' )
{
alleleLength[ alleleId ] -= snpInfo[ snpId ].length ;
}
else if ( snpInfo[ snpId ].type == 'i' )
{
alleleLength[ alleleId ] += snpInfo[ snpId ].length ;
}
}
}
fclose( fp ) ;
int numOfAllele = alleleIdToName.size() ;
int numOfSnps = snpLink.size() ;
snpAllele = new int* [numOfSnps] ;
for ( i = 0 ; i < numOfSnps ; ++i )
{
snpAllele[i] = new int[numOfAllele] ;
memset( snpAllele[i], 0, sizeof( int ) * numOfAllele ) ;
}
for ( i = 0 ; i < numOfSnps ; ++i )
{
int size = snpLink[i].size() ;
for ( j = 0 ; j < size ; ++j )
{
snpAllele[i][ snpLink[i][j] ] = 1 ;
}
}
// Compute the compatbility score for each alignment and the allele
// Get the number of alignment
int numOfAlignments = 0 ;
while ( alignments.Next() )
++numOfAlignments ;
alignments.Rewind() ;
compatibility = new struct _compatible*[numOfAlignments] ;
for ( i = 0 ; i < numOfAlignments ; ++i )
{
compatibility[i] = new struct _compatible[ numOfAllele ] ;
for ( j = 0 ; j < numOfAllele ; ++j )
{
compatibility[i][j].value = 0 ;//-log( (double)alleleLength[j] ) / log( 10.0 );
}
}
i = 0 ;
bool *snpHit = new bool[ numOfSnps ] ;
while ( alignments.Next() )
{
struct _pair coord = alignments.segments[0] ;
alignmentCoords.push_back( coord ) ;
memset( snpHit, 0, sizeof( bool ) * numOfSnps ) ;
Split( alignments.GetFieldZ( "Zs" ), ',', fields ) ;
int size = fields.size() ;
for ( k = 0 ; k < size ; ++k )
{
std::vector<std::string> subfields ;
Split( fields[k].c_str(), '|', subfields ) ;
int snpId = snpNameToId[ subfields[2] ] ;
snpHit[ snpId ] = true ;
}
for ( k = coord.a ; k <= coord.b ; ++k )
{
int size = positionToSnp[k].size() ;
for ( int l = 0 ; l < size ; ++l )
{
// if this SNP is hit. Then other allele don't have this snp
// will deduct its likelihood
//TODO: the deduction can be based on the quality score of the read
int tag = 0 ;
int snpId = positionToSnp[k][l] ;
if ( snpHit[ snpId ] )
{
tag = 0 ;
}
else
{
// if this SNP is not hit, then every allele containing this snp
// will deduct its likelihood
tag = 1 ;
}
for ( j = 0 ; j < numOfAllele ; ++j )
{
if ( snpAllele[ snpId ][j] == tag )
{
int v = -2 ;
//if ( snpInfo[ snpId ].type == 'd' || snpInfo[ snpId ].type == 'i' )
// v = -4 * snpInfo[ snpId ].length ;
if ( snpInfo[ snpId ].type == 'd' && snpInfo[ snpId ].position < k
&& k != coord.a )
{
// The penality has already been subtracted.
v = 0 ;
}
compatibility[i][j].value += v ;
/*if ( i == 8 && j == 78 )
{
printf( "Bad snp %d: %d %d\n", tag, k, positionToSnp[k][l] ) ;
}*/
}
}
}
}
++i ;
}
//printf( "%d %d\n", numOfAlignments, numOfAllele ) ;
// Now, let's consider every pair of alleles, and compute its log likelihood
double **logLikelihood ;
logLikelihood = new double *[ numOfAllele] ;
for ( j = 0 ; j < numOfAllele ; ++j )
{
logLikelihood[j] = new double[ numOfAllele ] ;
//memset( logLikelihood[j], 0, sizeof( double ) * numOfAllele ) ;
for ( k = 0 ; k < numOfAllele ; ++k )
logLikelihood[j][k] = 0 ;
}
int prevBin = -1 ;
double assignJBin = 0 ;
double assignKBin = 0 ;
for ( j = 0 ; j < numOfAllele ; ++j )
{
for ( k = j ; k < numOfAllele ; ++k )
{
double binAdjust = 0 ;
double averageRead = ( (double)numOfAlignments ) / (double)( alleleLength[j] + alleleLength[k] ) * binSize ;
for ( i = 0 ; i < numOfAlignments ; ++i )
{
double vj = compatibility[i][j].value ;
double vk = compatibility[i][k].value ;
double weightJ = 0, weightK = 0 ;
if ( vj == vk )
{
weightJ = weightK = 0.5 ;
}
else if ( vj == vk + 2 )
{
if ( vj == 0 )
{
weightJ = 1 ;
}
else
{
weightJ = 0.99 ;
weightK = 0.01 ;
}
}
else if ( vk == vj + 2 )
{
if ( vk == 0 )
{
weightK = 1 ;
}
else
{
weightJ = 0.01 ;
weightK = 0.99 ;
}
}
else
{
if ( vk > vj )
weightK = 1 ;
else
weightJ = 1 ;
}
double l = weightJ * compatibility[i][j].value + weightK * compatibility[i][k].value ;
if ( alignmentCoords[i].a / binSize != prevBin )
{
if ( prevBin != -1 &&
( assignJBin > averageRead + 4 * sqrt( averageRead )
|| assignKBin > averageRead + 4 * sqrt( averageRead ) ) )
{
//if ( j == 8 && k == 78 )
// printf( "%lf: %lf %lf %d %d\n", averageRead, assignJBin, assignKBin, alleleLength[j], alleleLength[k] ) ;
binAdjust -= 4 ;
}
prevBin = alignmentCoords[i].a / binSize ;
assignJBin = 0 ;
assignKBin = 0 ;
}
assignJBin += weightJ ;
assignKBin += weightK ;
/*if ( j == 8 && k == 78 && l < 0 )
{
printf( "Bad alignment %d (%s %s). %lf %lf: %lf\n", i,
alleleIdToName[j].c_str(), alleleIdToName[k].c_str(),
compatibility[i][j].value, compatibility[i][k].value, l ) ;
}*/
logLikelihood[j][k] += l ;
}
logLikelihood[j][k] += ( -log( (double)alleleLength[j] ) / log(10.0 ) -
log( (double)alleleLength[k] ) / log(10.0) ) ;
logLikelihood[j][k] += binAdjust ;
}
}
// Find the result
double max ;
int maxj = -1 ;
int maxk = -1 ;
std::vector< struct _result > results ;
for ( j = 0 ; j < numOfAllele ; ++j )
{
for ( k = j ; k < numOfAllele ; ++k )
{
if ( maxj == -1 || logLikelihood[j][k] > max )
{
maxj = j ;
maxk = k ;
max = logLikelihood[j][k] ;
}
struct _result r ;
r.a = j ;
r.b = k ;
r.logLikelihood = logLikelihood[j][k] ;
results.push_back( r ) ;
}
}
//printf( "%s %s %lf\n", alleleIdToName[ maxj ].c_str(), alleleIdToName[ maxk ].c_str(), max) ;
//printf( "%lf\n", logLikelihood[124][128] ) ;
if ( results.size() == 0 )
{
printf( "-1 -1 -1\n" ) ;
exit( 1 ) ;
}
std::sort( results.begin(), results.end(), CompResult ) ;
i = 0 ;
printf( "%s %s %lf\n", alleleIdToName[ results[i].a ].c_str(), alleleIdToName[ results[i].b ].c_str(),
results[i].logLikelihood ) ;
k = results.size() ;
for ( i = 1 ; i < k ; ++i )
{
if ( results[i].logLikelihood != results[0].logLikelihood )
break ;
printf( "%s %s %lf\n", alleleIdToName[ results[i].a ].c_str(), alleleIdToName[ results[i].b ].c_str(),
results[i].logLikelihood ) ;
}
return 0 ;
}