// Stitch provides a function that may be used by polymer types to implement Stitcher.
// It makes use of reflection and so may be slower than type-specific implementations.
// This is the reference implementation and should be used to compare type-specific
// implementation against in testing.
func Stitch(pol interface{}, offset int, f feat.FeatureSet) (s interface{}, err error) {
t := interval.NewTree()
var i *interval.Interval
for _, feature := range f {
i, err = interval.New(emptyString, feature.Start, feature.End, 0, nil)
if err != nil {
return
} else {
t.Insert(i)
}
}
pv := reflect.ValueOf(pol)
pLen := pv.Len()
end := pLen + offset
span, err := interval.New(emptyString, offset, end, 0, nil)
if err != nil {
panic("Sequence.End() < Sequence.Start()")
}
fs, _ := t.Flatten(span, 0, 0)
l := 0
for _, seg := range fs {
l += util.Min(seg.End(), end) - util.Max(seg.Start(), offset)
}
tv := reflect.MakeSlice(pv.Type(), 0, l)
for _, seg := range fs {
tv = reflect.AppendSlice(tv, pv.Slice(util.Max(seg.Start()-offset, 0), util.Min(seg.End()-offset, pLen)))
}
return tv.Interface(), nil
}
// Join segments of the sequence, returning any error.
func (self *Seq) Compose(f feat.FeatureSet) (err error) {
l := 0
for _, seg := range f {
if seg.End < seg.Start {
return bio.NewError("Feature end < start", 0, seg)
}
l += util.Min(seg.End, self.End()) - util.Max(seg.Start, self.Start())
}
t := &Seq{}
*t = *self
t.S = &Packing{Letters: make([]alphabet.Pack, 0, (l+3)/4)}
var tseg seq.Sequence
for _, seg := range f {
tseg, err = self.Subseq(util.Max(seg.Start, self.Start()), util.Min(seg.End, self.End()))
if err != nil {
return
}
tseg := tseg.(*Seq)
if seg.Strand == -1 {
tseg.RevComp()
}
tseg.S.Align(seq.Start)
t.S.Align(seq.End)
t.S.Letters = append(t.S.Letters, tseg.S.Letters...)
t.S.RightPad = tseg.S.RightPad
}
*self = *t
return
}
func (self *Interval) adjustRange() {
if self.left != nil && self.right != nil {
self.minStart = util.Min(self.start, self.left.minStart)
self.maxEnd = util.Max(self.end, self.left.maxEnd, self.right.maxEnd)
} else if self.left != nil {
self.minStart = util.Min(self.start, self.left.minStart)
self.maxEnd = util.Max(self.end, self.left.maxEnd)
} else if self.right != nil {
self.minStart = util.Min(self.start, self.right.minStart)
self.maxEnd = util.Max(self.end, self.right.maxEnd)
}
}
func (self *Multi) Stitch(f feat.FeatureSet) (err error) {
tr := interval.NewTree()
var i *interval.Interval
for _, feature := range f {
i, err = interval.New(emptyString, feature.Start, feature.End, 0, nil)
if err != nil {
return
} else {
tr.Insert(i)
}
}
span, err := interval.New(emptyString, self.Start(), self.End(), 0, nil)
if err != nil {
panic("Sequence.End() < Sequence.Start()")
}
fs, _ := tr.Flatten(span, 0, 0)
ff := feat.FeatureSet{}
for _, seg := range fs {
ff = append(ff, &feat.Feature{
Start: util.Max(seg.Start(), self.Start()),
End: util.Min(seg.End(), self.End()),
})
}
return self.Compose(ff)
}
func (self *Seq) stitch(f []*interval.Interval) (ts []byte) {
for _, seg := range f {
ts = append(ts, self.Seq[util.Max(seg.Start()-self.Offset, 0):util.Min(seg.End()-self.Offset, len(self.Seq))]...)
}
return
}
// Merge a range of intervals provided by r. Returns merged intervals in a slice and
// intervals contributing to merged intervals groups in a slice of slices.
func (self *Interval) flatten(r chan *Interval, tolerance int) (flat []*Interval, rich [][]*Interval) {
flat = []*Interval{}
rich = [][]*Interval{{}}
min, max := util.MaxInt, util.MinInt
var last *Interval
for current := range r {
if last != nil && current.start-tolerance > max {
n, _ := New(current.seg, min, max, 0, nil)
flat = append(flat, n)
min = current.start
max = current.end
rich = append(rich, []*Interval{})
} else {
min = util.Min(min, current.start)
max = util.Max(max, current.end)
}
rich[len(rich)-1] = append(rich[len(rich)-1], current)
last = current
}
n, _ := New(last.seg, min, max, 0, nil)
flat = append(flat, n)
return
}
func (self *Quality) stitch(fs []*interval.Interval) (tq []Qsanger) {
for _, seg := range fs {
tq = append(tq, self.Qual[util.Max(seg.Start()-self.Offset, 0):util.Min(seg.End()-self.Offset, len(self.Qual))]...)
}
return
}
// Join sequentially order disjunct segments of the sequence, returning any error.
func (self *Seq) Stitch(f feat.FeatureSet) (err error) {
tr := interval.NewTree()
var i *interval.Interval
for _, feature := range f {
i, err = interval.New(emptyString, feature.Start, feature.End, 0, nil)
if err != nil {
return
} else {
tr.Insert(i)
}
}
span, err := interval.New(emptyString, self.offset, self.End(), 0, nil)
if err != nil {
panic("packed: Sequence.End() < Sequence.Start()")
}
fs, _ := tr.Flatten(span, 0, 0)
l := 0
for _, seg := range fs {
l += util.Min(seg.End(), self.End()) - util.Max(seg.Start(), self.Start())
}
t := &Seq{}
*t = *self
t.S = &Packing{Letters: make([]alphabet.Pack, 0, (l+3)/4)}
var tseg seq.Sequence
for _, seg := range fs {
tseg, err = self.Subseq(util.Max(seg.Start(), self.Start()), util.Min(seg.End(), self.End()))
if err != nil {
return
}
s := tseg.(*Seq).S
s.Align(seq.Start)
t.S.Align(seq.End)
t.S.Letters = append(t.S.Letters, s.Letters...)
t.S.RightPad = s.RightPad
}
*self = *t
return
}
// Compose provides a function that may be used by polymer types to implement Composer.
// It makes use of reflection and so may be slower than type-specific implementations.
// This is the reference implementation and should be used to compare type-specific
// implementation against in testing.
func Compose(pol interface{}, offset int, f feat.FeatureSet) (s []interface{}, err error) {
pv := reflect.ValueOf(pol)
pLen := pv.Len()
end := pLen + offset
tv := make([]reflect.Value, len(f))
for i, seg := range f {
if seg.End < seg.Start {
return nil, bio.NewError("Feature End < Start", 0, f)
}
l := util.Min(seg.End, end) - util.Max(seg.Start, offset)
tv[i] = reflect.MakeSlice(pv.Type(), l, l)
reflect.Copy(tv[i], pv.Slice(util.Max(seg.Start-offset, 0), util.Min(seg.End-offset, pLen)))
}
s = make([]interface{}, len(tv))
for i := range tv {
s[i] = tv[i].Interface()
}
return
}
func (self *Interval) merge(i *Interval, overlap int) (inserted *Interval, removed []*Interval) {
r := make(chan *Interval)
removed = []*Interval{}
wait := make(chan struct{})
go func() {
defer close(wait)
min, max := util.MaxInt, util.MinInt
for old := range r {
min, max = util.Min(min, old.start), util.Max(max, old.end)
removed = append(removed, old)
}
i.start, i.end = util.Min(i.start, min), util.Max(i.end, max)
inserted = i
// TODO: Do something sensible when only one interval is found and the only action is to extend or ignore
}()
self.intersect(i, overlap, r)
close(r)
<-wait
return
}
// Create a new KmerRainbow defined by the rectangle r, kmerindex index and background color.
func NewKmerRainbow(r image.Rectangle, index *kmerindex.Index, background color.HSVA) *KmerRainbow { // should generalise the BG color
h := r.Dy()
kmers := make([]int, h)
kmask := util.Pow4(index.GetK())
kmaskf := float64(kmask)
f := func(index *kmerindex.Index, _, kmer int) {
kmers[int(float64(kmer)*float64(h)/kmaskf)]++
}
index.ForEachKmerOf(index.Seq, 0, index.Seq.Len(), f)
max := util.Max(kmers...)
return &KmerRainbow{
RGBA: image.NewRGBA(r),
Index: index,
Max: max,
BackGround: background,
}
}
// Method to align two sequences using the Smith-Waterman algorithm. Returns an alignment or an error
// if the scoring matrix is not square.
func (a *Aligner) Align(reference, query *seq.Seq) (aln seq.Alignment, err error) {
gap := len(a.Matrix) - 1
for _, row := range a.Matrix {
if len(row) != gap+1 {
return nil, bio.NewError("Scoring matrix is not square.", 0, a.Matrix)
}
}
r, c := reference.Len()+1, query.Len()+1
table := make([][]int, r)
for i := range table {
table[i] = make([]int, c)
}
max, maxI, maxJ := 0, 0, 0
var (
score int
scores [3]int
)
for i := 1; i < r; i++ {
for j := 1; j < c; j++ {
if rVal, qVal := a.LookUp.ValueToCode[reference.Seq[i-1]], a.LookUp.ValueToCode[query.Seq[j-1]]; rVal < 0 || qVal < 0 {
continue
} else {
scores[diag] = table[i-1][j-1] + a.Matrix[rVal][qVal]
scores[up] = table[i-1][j] + a.Matrix[rVal][gap]
scores[left] = table[i][j-1] + a.Matrix[gap][qVal]
score = util.Max(scores[:]...)
if score < 0 {
score = 0
}
if score >= max { // greedy so make farthest down and right
max, maxI, maxJ = score, i, j
}
table[i][j] = score
}
}
}
refAln := &seq.Seq{ID: reference.ID, Seq: make([]byte, 0, reference.Len())}
queryAln := &seq.Seq{ID: query.ID, Seq: make([]byte, 0, query.Len())}
for i, j := maxI, maxJ; table[i][j] != 0 && i > 0 && j > 0; {
if rVal, qVal := a.LookUp.ValueToCode[reference.Seq[i-1]], a.LookUp.ValueToCode[query.Seq[j-1]]; rVal < 0 || qVal < 0 {
continue
} else {
scores[diag] = table[i-1][j-1] + a.Matrix[rVal][qVal]
scores[up] = table[i-1][j] + a.Matrix[gap][qVal]
scores[left] = table[i][j-1] + a.Matrix[rVal][gap]
switch d := maxIndex(scores[:]); d {
case diag:
i--
j--
refAln.Seq = append(refAln.Seq, reference.Seq[i])
queryAln.Seq = append(queryAln.Seq, query.Seq[j])
case up:
i--
refAln.Seq = append(refAln.Seq, reference.Seq[i])
queryAln.Seq = append(queryAln.Seq, a.GapChar)
case left:
j--
refAln.Seq = append(refAln.Seq, a.GapChar)
queryAln.Seq = append(queryAln.Seq, query.Seq[j])
}
}
}
for i, j := 0, len(refAln.Seq)-1; i < j; i, j = i+1, j-1 {
refAln.Seq[i], refAln.Seq[j] = refAln.Seq[j], refAln.Seq[i]
}
for i, j := 0, len(queryAln.Seq)-1; i < j; i, j = i+1, j-1 {
queryAln.Seq[i], queryAln.Seq[j] = queryAln.Seq[j], queryAln.Seq[i]
}
aln = seq.Alignment{refAln, queryAln}
return
}