void CSplignTrim::CutToMatchRight(TSeg& s)
{
size_t pos = s.m_details.rfind('M');
if(pos == string::npos) {
s.SetToGap();
return;
}
size_t len = s.m_details.length() - pos - 1;//length to cut
if(len > 0) {
CutFromRight(len, s);
}
}
void CSplignTrim::Cut50FromRight(TSeg& s)
{
int score = 0, maxscore = -2;
string::iterator irs0, irs1, irs, irs_max;
irs0 = s.m_details.begin();
irs1 = s.m_details.end();
irs_max = irs0;
for(irs = irs0; irs != irs1; ++irs) {
if(*irs == 'M') {
++score;
} else {
--score;
}
if(score >= maxscore) {
maxscore = score;
irs_max = irs;
}
}
int len = irs1 - irs_max - 1;
if(len > 0) {
CutFromRight(len, s);
}
}
//trims exons around internal alignment gaps to complete codons
//if CDS can be retrieved from bioseq
void CSplignTrim::TrimHolesToCodons(TSegs& segments, CBioseq_Handle& mrna_bio_handle, bool mrna_strand, size_t mrna_len)
{
if( mrna_bio_handle ) {
//collect CDS intervals (could be more than one in a case of ribosomal slippage)
vector<TSeqRange> tr;
for(CFeat_CI ci(mrna_bio_handle, SAnnotSelector(CSeqFeatData::e_Cdregion)); ci; ++ci) {
for(CSeq_loc_CI slit(ci->GetLocation()); slit; ++slit) {
TSeqRange r, ori;
ori = slit.GetRange();
if( mrna_strand ) {
r = ori;
} else {//reverse
r.SetFrom(mrna_len - ori.GetTo() - 1);
r.SetTo(mrna_len - ori.GetFrom() - 1);
}
tr.push_back(r);
}
}
if(tr.empty()) return;// CDS not found
//trim
AdjustGaps(segments);//make sure there is no adjacent gaps
size_t pos1 = 0, pos2 = 2;
for(; pos2 < segments.size(); ++pos1, ++pos2) {
if( segments[pos1].m_exon && !segments[pos1+1].m_exon && segments[pos2].m_exon ) {//candidate for trimming
//trim left exon
TSeqPos p1 = segments[pos1].m_box[1];
ITERATE(vector<TSeqRange>, it, tr) {
if( p1 >= it->GetFrom() && p1 <= it->GetTo() ) {
TSeqPos cut_mrna_len = (p1 + 1 - it->GetFrom()) % 3, cnt = 0;
string transcript = segments[pos1].m_details;
int i = (int)transcript.size() - 1;
for(; i>=0; --i) {
if( cnt%3 == cut_mrna_len && transcript[i] == 'M' ) { //cut point
CutFromRight(transcript.size() - i - 1, segments[pos1]);
break;
}
if( transcript[i] != 'I' ) ++cnt;
}
if( i < 0 ) {// exon should not be so bad
NCBI_THROW(CAlgoAlignException, eInternal, g_msg_InvalidRange);
}
break;
}
}
//trim right exon
TSeqPos p2 = segments[pos2].m_box[0];
ITERATE(vector<TSeqRange>, it, tr) {
if( p2 >= it->GetFrom() && p2 <= it->GetTo() ) {
TSeqPos cut_mrna_len = ( 3 - ( p2 - it->GetFrom()) % 3 ) %3, cnt = 0;
string transcript = segments[pos2].m_details;
int i = 0;
for( ; i < (int)transcript.size(); ++i) {
if( cnt%3 == cut_mrna_len && transcript[i] == 'M' ) { //cut point
CutFromLeft(i, segments[pos2]);
break;
}
if( transcript[i] != 'I' ) ++cnt;
}
if( i == (int)transcript.size() ) {// exon should not be so bad
NCBI_THROW(CAlgoAlignException, eInternal, g_msg_InvalidRange);
}
break;
}
}
}
}
AdjustGaps(segments);
}
// try improving the segment by cutting it from the left, 20/20 rule
void CSplignTrim::ImproveFromRight(TSeg& s)
{
CutToMatchRight(s);
Cut50FromRight(s);
if(ThrowAwayShortExon(s)) return;
int len_total = (int)s.m_details.size();
if(len_total <= 20) return;//two short
//compute number of matches
int match_total = 0;
string::reverse_iterator irs0 = s.m_details.rbegin(),
irs1 = s.m_details.rend(), irs;
for(irs = irs0; irs != irs1; ++irs) {
if(*irs == 'M') {
++match_total;
}
}
//find the left boundary, 20/20 rule
{{
size_t pos = max(20, len_total/5) - 1;
pos = s.m_details.find('M', pos);
if( pos == string::npos ) return;//no M found.
pos = s.m_details.find_first_not_of('M', pos);
if( pos == string::npos ) return;// 100% id on the right, nothing to trim
irs1 = irs1 - pos;
}}
//after 20/20 *ir1s is M, irs1+1 is not M and eventually irs1 is a left boundary for trimming
string::reverse_iterator irs_tr = s.m_details.rend(); //trimming point
int match = 0, len = 0;
for(irs = irs0; irs != irs1; ++irs) {
if(*irs == 'M') {
++match;
}
++len;
double rid = match / (double)len;
double lid = (match_total - match) / (double)(len_total - len);
//dropoff check
double epsilon = 1e-10;
if( lid - rid - m_MaxPartExonIdentDrop > epsilon ) {
irs_tr = irs;
//do not count trimmed part, adjust values
match_total -= match;
len_total -= len;
match = 0;
len = 0;
}
}
if( irs_tr == s.m_details.rend() ) return;//no trimming point found
//actual trimming
CutFromRight( irs_tr - irs0 + 1 , s );
ThrowAwayShortExon(s);
}
