def mpat(tr, seqs) :
dseqs = dict(seqs)
for n in getPostOrder(tr) :
data = n.data
if not n.succ :
data.seq = (None,dseqs[n.data.taxon])
else :
s1,s2 = [tr.node(x).data.seq for x in n.succ]
if s1[1] :
if s2[1] :
a = calign.globalAlign(s1[1],s2[1])
data.seq = (calign.createProfile(a),None)
else :
p1,p2 = calign.createProfile(s1[1:]), s2[0]
#assert all([sum(x)==sum(p1[0]) for x in p1])
#assert all([sum(x)==sum(p2[0]) for x in p2])
pa = calign.prof2profAlign(p1,p2)
data.seq = (trimp(pa),None)
#print len(pa)
else :
p1 = s1[0]
if s2[1] :
p2 = calign.createProfile(s2[1:])
else :
p2 = s2[0]
#assert all([sum(x)==sum(p1[0]) for x in p1])
#assert all([sum(x)==sum(p2[0]) for x in p2])
pa = calign.prof2profAlign(p1,p2)
data.seq = (trimp(pa),None)
#print len(pa)
#import pdb; pdb.set_trace()
assert n.id == tr.root
return n.data.seq[0]
def mpa(tr, seqs, scores = defaultMatchScores, trimEnd = None) :
dseqs = dict(seqs)
#scores = (None,None,gapPenalty,feg)
for n in getPostOrder(tr) :
data = n.data
if not n.succ :
data.seq = (None,dseqs[n.data.taxon.strip("'")])
else :
s1,s2 = [tr.node(x).data.seq for x in n.succ]
if s1[1] :
if s2[1] :
a = calign.globalAlign(s1[1],s2[1], scores = scores)
data.seq = (calign.createProfile(a),None)
else :
p1,p2 = calign.createProfile(s1[1:]), s2[0]
pa = calign.prof2profAlign(p1,p2, scores = scores)
data.seq = (trimendsp(pa, trimEnd) if trimEnd is not None else pa,None)
#print len(pa)
else :
p1 = s1[0]
if s2[1] :
p2 = calign.createProfile(s2[1:])
else :
p2 = s2[0]
pa = calign.prof2profAlign(p1,p2, scores = scores)
data.seq = (trimendsp(pa, trimEnd) if trimEnd is not None else pa,None)
#print len(pa)
#import pdb; pdb.set_trace()
assert n.id == tr.root
return n.data.seq[0]
def _populateTreeWithNodeToTipDistances(tree) :
for n in getPostOrder(tree) :
if not n.succ:
n.data.dtips = [[[n,0]],[],[]]
else :
ch = [tree.node(c) for c in n.succ]
n.data.dtips = [[[a[0],a[1]+x.data.branchlength] for a in x.data.dtips[0]] +
[[a[0],a[1]+x.data.branchlength] for a in x.data.dtips[1]]
for x in ch]
if n.id != tree.root :
n.data.dtips.append([])
_populateTipDistancesFromParent(tree, tree.node(tree.root), [])
def _UMTree_slsqp(tin, targetHeight = None, useDerives = True,
doInit = False, normed = True,
niter=1000, r0x = None, verb = 0, internals = False) :
t = copy.deepcopy(tin)
nmap = []
code = []
primecode = []
codeNodes = []
# Number of branches
nr = 2*(len(t.get_terminals()) - 1)
if targetHeight is None :
targetHeight = treeHeightEstimate(t)
assert float(targetHeight) == targetHeight and targetHeight > 0
for n in getPostOrder(t) :
if n.id != t.root :
if n.prev != t.root :
n.data.myindx = len(nmap)
n.data.hexpr = "%.14f * r[%d]" % (n.data.branchlength, n.data.myindx)
c = ["0"]*(nr+2)
c[n.data.myindx] = "%.14f" % n.data.branchlength
n.data.hpexpr = ",".join(c)
nmap.append(n.id)
else :
n.data.hexpr = ""
n.data.hpexpr = ""
if n.succ :
n.data.hexpr += ' + h%d' % n.succ[0]
n.data.hpexpr = cmbn(n.data.hpexpr, n.succ[0])
if n.prev != t.root and n.data.hexpr :
n.data.chexpr = "h%d = %s" % (n.id, n.data.hexpr)
code.append(n.data.chexpr)
primecode.append("hp%d = [%s]" % (n.id, n.data.hpexpr))
else :
# left,right rates, left right branch : len(nmap)-2 to len(nmap)+1
assert nr - 2 == len(nmap)
lft = t.node(n.succ[0])
lft.data.myindx = nr-2
lft.data.hexpr = "r[%d] " % (nr) + lft.data.hexpr
lft.data.chexpr = "h%d = %s" % (lft.id,lft.data.hexpr)
sss = ",".join(["0"]*(nr) + ["1","0"])
if lft.data.hpexpr :
lft.data.hpexpr = "x1+y1 for x1,y1 in zip([%s], %s)" % \
(sss,lft.data.hpexpr)
else :
lft.data.hpexpr = sss
rht = t.node(n.succ[1])
rht.data.myindx = nr-1
brr = (lft.data.branchlength+rht.data.branchlength)
rht.data.hexpr = "r[%d] " % (nr+1) + rht.data.hexpr
rht.data.chexpr = "h%d = %s" % (rht.id,rht.data.hexpr)
code.append(rht.data.chexpr)
code.append("ch1 = h%d - %.14f" % (rht.id, targetHeight))
sss = ",".join(["0"]*(nr) + ["0","1"])
if rht.data.hpexpr :
rht.data.hpexpr = "x2+y2 for x2,y2 in zip([%s], %s)" % \
(sss, rht.data.hpexpr)
else :
rht.data.hpexpr = sss
nmap.extend(n.succ)
code.append(lft.data.chexpr)
primecode.append("cph1 = hp%d = [%s]" % (rht.id, rht.data.hpexpr))
primecode.append("cph2 = hp%d = [%s]" % (lft.id, lft.data.hpexpr))
code.append("ch2 = h%d - %.14f" % (lft.id, targetHeight))
brr = rht.data.branchlength + lft.data.branchlength
code.append("crt = (r[%d] * r[%d] + r[%d] * r[%d]) - %.15f" % (nr-2,nr,nr-1,nr+1,brr))
sss = ",".join(["0"]*(nr-2)) + ",r[%d],r[%d],r[%d],r[%d]" % (nr,nr+1,nr-2,nr-1)
primecode.append("cprt = [%s]" % sss)
if n.succ and n.id != t.root:
code.append("c%d = h%d - h%d" % ((n.id,) + tuple(n.succ)))
primecode.append("cp%d = [x-y for x,y in zip(hp%d, hp%d)]" % ((n.id,) + tuple(n.succ)))
codeNodes.append(n.id)
cd = ["def fx(r) :"]
cd.extend([" " + x for x in code])
ccs = ["c%d" % k for k in codeNodes] + ['ch1','ch2','crt']
hhs = ["h%d" % k for k in nmap]
cd.append(" return [" + ",".join(ccs) + "], [" + ",".join(hhs) + "]")
exec ( "\n".join(cd) ) in globals()
cdp = ["def fxp(r) :"]
cdp.extend([" " + x for x in primecode])
ccs = ["cp%d" % k for k in codeNodes] + ['cph1','cph2','cprt']
cdp.append(" return [" + ",".join(ccs) + "]")
exec ( "\n".join(cdp) ) in globals()
# optimization target
if r0x is not None :
r0 = r0x
else :
if doInit:
calcNoderateMultipliers(t, targetHeight)
cl,cr = t.node(t.root).succ
btl,btr = (t.node(cl).data.timebranchlength , t.node(cr).data.timebranchlength)
r0 = [t.node(c).data.timebranchlength/t.node(c).data.branchlength for c in nmap] + \
[btl, btr]
r0[-3] = 1/r0[-3]
r0[-4] = 1/r0[-4]
else :
r0 = [1]*(nr) + [targetHeight*.1]*2
# most are constant, can improve
#derv = array(fxp(0))
assert nr == len(nmap)
slsqp = scipy.optimize.slsqp.fmin_slsqp
if useDerives :
fm,fmp = (_targetNormedVar,_targetNormedVarDerivatives) if normed \
else (_targetVar, _targetVarDerivatives)
re = slsqp(fm, r0,
fprime = lambda x : array(fmp(x)),
f_eqcons = lambda x : array(fx(x)[0]),
fprime_eqcons = lambda x : array(fxp(x)),
bounds = [(1e-10,10)]*nr + [(0,targetHeight)]*2,
iter = niter, iprint=verb, full_output=1)
else :
re = slsqp(fm, r0,
f_eqcons = lambda x : array(fx(x)[0]),
bounds = [(1e-10,10)]*nr + [(0,targetHeight)]*2,
iprint=verb, full_output=1)
r = re[0]
if re[3] != 0 :
if not (re[3] == 8 and (abs(targetHeight - r[-2]) < 1e-9 or
abs(targetHeight - r[-1]) < 1e-9)) :
ok = re[3] in [8,9] and all([abs(x) < 1e-8 for x in fx(r)[0]])
if not ok :
import pdb ; pdb.set_trace()
raise RuntimeError(re[4])
hs = fx(r)[1]
for k,i in enumerate(nmap) :
n = t.node(i)
n.data.ph = hs[k]
n.data.subsbranchlength = n.data.branchlength
n.data.attributes = {'clockrate' : r[n.data.myindx]}
if n.succ :
n.data.branchlength = n.data.ph - t.node(n.succ[0]).data.ph
else :
n.data.branchlength = n.data.ph
n.data.branchlength = max(n.data.branchlength, 0)
if internals :
return t, fm(r), r, (fx, fxp, fm, fmp)
return t, fm(r)
