def pair_f(f, rtimes, std_sizes, similarity, deviation=False):
""" match rtimes to std_sizes
return: [ (rtime, size), ... ] or
[ (rtime, size, f(rtime), dev), ... ]
"""
rtimes = list(reversed(rtimes))
std_sizes = list(reversed(std_sizes))
similarity = list(reversed(similarity))
S = generate_scores(std_sizes, rtimes, similarity, f)
result = dp(S, -5e-3)
matches = result['matches']
aligned_peaks = [(rtimes[j], std_sizes[i]) for i, j in matches]
if not deviation:
return aligned_peaks
peak_pairs = []
for (rtime, size) in aligned_peaks:
rtime_size = f(rtime)
peak_pairs.append((rtime, size, rtime_size, (size - rtime_size)**2))
return peak_pairs
def align_dp(rtimes, sizes, similarity, z, rss, order=3):
""" align ladders with peaks using dynamic programming (global alignment)
return (dpscore, RSS, Z, ladder_aligned_peaks)
"""
sizes = list(sorted(sizes, reverse=True))
rtimes = list(sorted(rtimes, reverse=True))
dpscore = -1
while True:
S = generate_scores(sizes, rtimes, similarity, np.poly1d(z))
result = dp(S, -5e-3)
cur_dpscore = result['D'][-1][-1]
matches = result['matches']
aligned_peaks = [(sizes[i], rtimes[j]) for i, j in matches]
# realign
std_size, peak_sizes = zip(*aligned_peaks)
cur_zres = estimate_z(peak_sizes, std_size, order)
if cur_dpscore < dpscore:
if is_verbosity(4):
cerr('W: dynamic programming did not converge!!')
break
if cur_dpscore == dpscore:
break
z = cur_zres.z
rss = cur_zres.rss
dpscore = cur_dpscore
sized_peaks = aligned_peaks
return DPResult(dpscore, rss, z, sized_peaks)