int CAlnVec::CalculateScore(TNumrow row1, TNumrow row2) const
{
TNumrow numrows = m_NumRows;
TNumrow index1 = row1, index2 = row2;
TSignedSeqPos start1, start2;
string buff1, buff2;
bool isAA1, isAA2;
int score = 0;
TSeqPos len;
isAA1 = GetBioseqHandle(row1).GetBioseqCore()
->GetInst().GetMol() == CSeq_inst::eMol_aa;
isAA2 = GetBioseqHandle(row2).GetBioseqCore()
->GetInst().GetMol() == CSeq_inst::eMol_aa;
CSeqVector& seq_vec1 = x_GetSeqVector(row1);
TSeqPos size1 = seq_vec1.size();
CSeqVector & seq_vec2 = x_GetSeqVector(row2);
TSeqPos size2 = seq_vec2.size();
for (TNumseg seg = 0; seg < m_NumSegs; seg++) {
start1 = m_Starts[index1];
start2 = m_Starts[index2];
if (start1 >=0 && start2 >= 0) {
len = m_Lens[seg];
if (IsPositiveStrand(row1)) {
seq_vec1.GetSeqData(start1,
start1 + len,
buff1);
} else {
seq_vec1.GetSeqData(size1 - (start1 + len),
size1 - start1,
buff1);
}
if (IsPositiveStrand(row2)) {
seq_vec2.GetSeqData(start2,
start2 + len,
buff2);
} else {
seq_vec2.GetSeqData(size2 - (start2 + len),
size2 - start2,
buff2);
}
score += CalculateScore(buff1, buff2, isAA1, isAA2);
}
index1 += numrows;
index2 += numrows;
}
return score;
}
CSeqVector& CAlnVec::x_GetSeqVector(TNumrow row) const
{
TSeqVectorCache::iterator iter = m_SeqVectorCache.find(row);
CRef<CSeqVector> seq_vec;
if (iter != m_SeqVectorCache.end()) {
seq_vec = iter->second;
}
else {
CBioseq_Handle h = GetBioseqHandle(row);
CSeqVector vec = h.GetSeqVector
(CBioseq_Handle::eCoding_Iupac,
IsPositiveStrand(row) ?
CBioseq_Handle::eStrand_Plus :
CBioseq_Handle::eStrand_Minus);
seq_vec.Reset(new CSeqVector(vec));
m_SeqVectorCache[row] = seq_vec;
}
if ( seq_vec->IsNucleotide() ) {
if (m_NaCoding != CSeq_data::e_not_set) {
seq_vec->SetCoding(m_NaCoding);
}
else {
seq_vec->SetIupacCoding();
}
}
else if ( seq_vec->IsProtein() ) {
if (m_AaCoding != CSeq_data::e_not_set) {
seq_vec->SetCoding(m_AaCoding);
}
else {
seq_vec->SetIupacCoding();
}
}
return *seq_vec;
}
CBioseq_Handle CSimpleOM::GetBioseqHandle(TGi gi)
{
CSeq_id id;
id.SetGi(gi);
return GetBioseqHandle(id);
}
CBioseq_Handle CSimpleOM::GetBioseqHandle(const string& id_string)
{
CSeq_id id(id_string);
return GetBioseqHandle(id);
}
CSeqVector CSimpleOM::GetSeqVector(const CSeq_id_Handle& id, ENa_strand strand)
{
return GetBioseqHandle(id).GetSeqVector(strand);
}
CSeqVector CSimpleOM::GetSeqVector(TGi gi, ENa_strand strand)
{
return GetBioseqHandle(gi).GetSeqVector(strand);
}
CSeqVector CSimpleOM::GetSeqVector(const string& id_string, ENa_strand strand)
{
return GetBioseqHandle(id_string).GetSeqVector(strand);
}
void CAlnVec::CreateConsensus(vector<string>& consens) const
{
bool isNucleotide = GetBioseqHandle(0).IsNucleotide();
const int numBases = isNucleotide ? 4 : 26;
int base_count[26]; // must be a compile-time constant for some compilers
// determine what the number of segments required for a gapped consensus
// segment is. this must be rounded to be at least 50%.
int gap_seg_thresh = m_NumRows - m_NumRows / 2;
for (size_t j = 0; j < (size_t)m_NumSegs; ++j) {
// evaluate for gap / no gap
int gap_count = 0;
for (size_t i = 0; i < (size_t)m_NumRows; ++i) {
if (m_Starts[ j*m_NumRows + i ] == -1) {
++gap_count;
}
}
// check to make sure that this seg is not a consensus
// gap seg
if ( gap_count > gap_seg_thresh )
continue;
// the base threshold for being considered unique is at least
// 70% of the available sequences
int base_thresh =
((m_NumRows - gap_count) * 7 + 5) / 10;
{
// we will build a segment with enough bases to match
consens[j].resize(m_Lens[j]);
// retrieve all sequences for this segment
vector<string> segs(m_NumRows);
RetrieveSegmentSequences(j, segs);
TransposeSequences(segs);
typedef multimap<int, unsigned char, greater<int> > TRevMap;
//
// evaluate for a consensus
//
for (size_t i = 0; i < m_Lens[j]; ++i) {
if (isNucleotide) {
CollectNucleotideFrequences(segs[i], base_count, numBases);
} else {
CollectProteinFrequences(segs[i], base_count, numBases);
}
// we create a sorted list (in descending order) of
// frequencies of appearance to base
// the frequency is "global" for this position: that is,
// if 40% of the sequences are gapped, the highest frequency
// any base can have is 0.6
TRevMap rev_map;
for (int k = 0; k < numBases; ++k) {
// this gets around a potentially tricky idiosyncrasy
// in some implementations of multimap. depending on
// the library, the key may be const (or not)
TRevMap::value_type p(base_count[k], isNucleotide ? (1<<k) : k);
rev_map.insert(p);
}
// now, the first element here contains the best frequency
// we scan for the appropriate bases
if (rev_map.count(rev_map.begin()->first) == 1 &&
rev_map.begin()->first >= base_thresh) {
consens[j][i] = isNucleotide ?
ToIupac(rev_map.begin()->second) :
(rev_map.begin()->second+'A');
} else {
// now we need to make some guesses based on IUPACna
// notation
int count;
unsigned char c = 0x00;
int freq = 0;
TRevMap::iterator curr = rev_map.begin();
TRevMap::iterator prev = rev_map.begin();
for (count = 0;
curr != rev_map.end() &&
(freq < base_thresh || prev->first == curr->first);
++curr, ++count) {
prev = curr;
freq += curr->first;
if (isNucleotide) {
c |= curr->second;
} else {
unsigned char cur_char = curr->second+'A';
switch (c) {
case 0x00:
c = cur_char;
break;
case 'N': case 'D':
c = (cur_char == 'N' || cur_char == 'N') ? 'B' : 'X';
break;
case 'Q': case 'E':
c = (cur_char == 'Q' || cur_char == 'E') ? 'Z' : 'X';
break;
case 'I': case 'L':
c = (cur_char == 'I' || cur_char == 'L') ? 'J' : 'X';
break;
default:
c = 'X';
}
}
}
//
// catchall
//
if (count > 2) {
consens[j][i] = isNucleotide ? 'N' : 'X';
} else {
consens[j][i] = isNucleotide ? ToIupac(c) : c;
}
}
}
}
}
}
//
// CreateConsensus()
//
// compute a consensus sequence given a particular alignment
// the rules for a consensus are:
// - a segment is consensus gap if > 50% of the sequences are gap at this
// segment. 50% exactly is counted as sequence
// - for a segment counted as sequence, for each position, the most
// frequently occurring base is counted as consensus. in the case of
// a tie, the consensus is considered muddied, and the consensus is
// so marked
//
CRef<CDense_seg>
CAlnVec::CreateConsensus(int& consensus_row, CBioseq& consensus_seq,
const CSeq_id& consensus_id) const
{
consensus_seq.Reset();
if ( !m_DS || m_NumRows < 1) {
return CRef<CDense_seg>();
}
bool isNucleotide = GetBioseqHandle(0).IsNucleotide();
size_t i;
size_t j;
// temporary storage for our consensus
vector<string> consens(m_NumSegs);
CreateConsensus(consens);
//
// now, create a new CDense_seg
// we create a new CBioseq for our data and
// copy the contents of the CDense_seg
//
string data;
TSignedSeqPos total_bases = 0;
CRef<CDense_seg> new_ds(new CDense_seg());
new_ds->SetDim(m_NumRows + 1);
new_ds->SetNumseg(m_NumSegs);
new_ds->SetLens() = m_Lens;
new_ds->SetStarts().reserve(m_Starts.size() + m_NumSegs);
if ( !m_Strands.empty() ) {
new_ds->SetStrands().reserve(m_Strands.size() +
m_NumSegs);
}
for (i = 0; i < consens.size(); ++i) {
// copy the old entries
for (j = 0; j < (size_t)m_NumRows; ++j) {
int idx = i * m_NumRows + j;
new_ds->SetStarts().push_back(m_Starts[idx]);
if ( !m_Strands.empty() ) {
new_ds->SetStrands().push_back(m_Strands[idx]);
}
}
// add our new entry
// this places the consensus as the last sequence
// it should preferably be the first, but this would mean adjusting
// the bioseq handle and seqvector caches, and all row numbers would
// shift
if (consens[i].length() != 0) {
new_ds->SetStarts().push_back(total_bases);
} else {
new_ds->SetStarts().push_back(-1);
}
if ( !m_Strands.empty() ) {
new_ds->SetStrands().push_back(eNa_strand_unknown);
}
total_bases += consens[i].length();
data += consens[i];
}
// copy our IDs
for (i = 0; i < m_Ids.size(); ++i) {
new_ds->SetIds().push_back(m_Ids[i]);
}
// now, we construct a new Bioseq
{{
// sequence ID
CRef<CSeq_id> id(new CSeq_id());
id->Assign(consensus_id);
consensus_seq.SetId().push_back(id);
new_ds->SetIds().push_back(id);
// add a description for this sequence
CSeq_descr& desc = consensus_seq.SetDescr();
CRef<CSeqdesc> d(new CSeqdesc);
desc.Set().push_back(d);
d->SetComment("This is a generated consensus sequence");
// the main one: Seq-inst
CSeq_inst& inst = consensus_seq.SetInst();
inst.SetRepr(CSeq_inst::eRepr_raw);
inst.SetMol(isNucleotide ? CSeq_inst::eMol_na : CSeq_inst::eMol_aa);
inst.SetLength(data.length());
CSeq_data& seq_data = inst.SetSeq_data();
if (isNucleotide) {
CIUPACna& na = seq_data.SetIupacna();
na = CIUPACna(data);
} else {
CIUPACaa& aa = seq_data.SetIupacaa();
aa = CIUPACaa(data);
}
}}
consensus_row = new_ds->GetIds().size() - 1;
return new_ds;
}
CBioseq_Handle CTSE_Handle::GetBioseqHandle(const CSeq_id& id) const
{
return GetBioseqHandle(CSeq_id_Handle::GetHandle(id));
}