def write_results(self, out_dir, bname):
for names_tup, (Imin, Imax) in sorted(
self.Imin_max_pairs_given_names.items()):
Imin_name, Imax_name = names_tup
Imin = misc.list_if_scalar(Imin, len(self.concentrations))
Imax = misc.list_if_scalar(Imax, len(self.concentrations))
out_fname = '{}_{}_{}_Kds_and_ABAs.txt'.format(
bname, Imin_name.replace(' ', '_'),
Imax_name.replace(' ', '_'))
out_fpath = os.path.join(out_dir, out_fname)
with open(out_fpath, 'w') as out:
out.write('# Target: {}\n'.format(self.target))
out.write('# Neg Control: {}\n'.format(
self.neg_control_target))
out.write('# Concentration\tImin\tImax\n')
for conc, imin, imax in zip(self.concentrations, Imin, Imax):
out.write(
'\t'.join(map(str, map(float, (conc, imin, imax)))) +
'\n')
out.write('\t'.join([
'# Seq', 'Kd (pM)', 'Kd error', 'ABA (kB T)', 'ABA error'
]) + '\n')
out_zipper = zip(self.IA.seqs, self.Kds[names_tup],
self.Kd_errors[names_tup],
self.ABAs[names_tup],
self.ABA_errors[names_tup])
out.write('\n'.join(
'\t'.join(map(str, [seq, Kd, Kd_err, ABA, ABA_err]))
for seq, Kd, Kd_err, ABA, ABA_err in out_zipper))
def plot_normalized_intensities(self,
ax,
seq,
Imin,
Imax,
max_clust=2000,
xvals=None):
if xvals is None:
xvals = self.course_trait_list
Imin = misc.list_if_scalar(Imin, self.course_len)
Imax = misc.list_if_scalar(Imax, self.course_len)
inten_lol = self.intensity_lol_given_seq[seq]
nclust = min(max_clust, len(inten_lol[0]))
alpha = self._path_alpha(nclust)
for i in range(nclust):
path = [(inten_list[i] - imn) /
(imx - imn) if inten_list[i] is not None else None
for inten_list, imn, imx in zip(inten_lol, Imin, Imax)]
ax.plot(xvals, path, 'b', alpha=alpha)
def all_normalized_trait_and_inten_vals_given_seq(self,
seq,
Imin,
Imax,
max_clust=None,
bootstrap=False):
"""
Returns all concentration/intensities pairs in read_names, return as two lists, with the
intensities adjusted by Imin and Imax to run typically between 0 and 1.
Returns:
:list: all_concentrations
:list: all_intensities
"""
Imin = misc.list_if_scalar(Imin, self.course_len)
Imax = misc.list_if_scalar(Imax, self.course_len)
assert len(Imin) == len(Imax) == len(self.course_trait_list), (Imin,
Imax)
all_trait_vals, all_intensities = [], []
for tval, imn, imx, inten_arr, inten_list in zip(
self.course_trait_list, Imin, Imax,
self.intensity_loarr_given_seq[seq],
self.intensity_lol_given_seq[seq]):
if len(inten_arr) == 0:
tmp_inten = np.array([])
elif bootstrap:
tmp_idx = np.random.choice(np.arange(len(inten_list)),
size=min(max_clust,
len(inten_list)),
replace=True)
tmp_inten = np.array([
inten_list[idx] for idx in tmp_idx
if inten_list[idx] != None
])
else:
tmp_inten = np.array(list(inten_arr[:max_clust]))
tmp_inten = list((tmp_inten - imn) / (imx - imn))
all_trait_vals.extend([tval] * len(tmp_inten))
all_intensities.extend(tmp_inten)
return all_trait_vals, all_intensities
def plot_raw_fit(self, ax, seq, Kd, Imin, Imax):
self.IA.plot_raw_intensities(ax, seq, xvals=self.nM_concentrations)
Imin = misc.list_if_scalar(Imin, self.IA.course_len)
Imax = misc.list_if_scalar(Imax, self.IA.course_len)
nM_Kd = Kd / 1000.0
ax.plot(self.nM_concentrations,
Imin,
'ko',
alpha=0.8,
label='$I_{min}$')
ax.plot(self.nM_concentrations,
Imax,
's',
color='darkgoldenrod',
alpha=0.8,
label='$I_{max}$')
def Iobs(x, Kd, Imin, Imax):
return (Imax - Imin) / (1.0 + (float(Kd) / x)) + Imin
fit_path = [
Iobs(conc, nM_Kd, imn, imx)
for conc, imn, imx in zip(self.nM_concentrations, Imin, Imax)
]
ax.plot(self.nM_concentrations, fit_path, 'k--')
x = np.logspace(np.log10(self.nM_concentrations[0]),
np.log10(self.nM_concentrations[-1]), 200)
y = [Iobs(xx, nM_Kd, self.Imin_const, self.Imax_const) for xx in x]
ax.plot(x, y, 'r')
ax.set_xscale('log')
ax.set_xlabel('Concentration (nM)', fontsize=18)
ax.set_axis_bgcolor('white')
ax.grid(False)
ax.set_ylabel('Intensity', fontsize=18)
def ML_fit_Kd(self,
seq,
Imin_name,
max_clust=None,
bootstrap=False,
*args,
**kw_args):
if max_clust is None:
max_clust = self.max_clust
if bootstrap:
len_inten_list = len(self.IA.intensity_lol_given_seq[seq][0])
bootstrap_idxs = np.random.choice(np.arange(len_inten_list),
size=min(len_inten_list,
max_clust),
replace=True)
else:
bootstrap_idxs = None
Imin = getattr(self, Imin_name)
Imin_list = misc.list_if_scalar(Imin, len(self.concentrations))
def neg_log_L(Kd):
return -self.model_logL([seq], [Kd],
Imin_list,
self.Imax_ML[Imin_name],
self.sigma_consts[Imin_name],
self.sigma_I[Imin_name],
bootstrap_idxs=bootstrap_idxs)
res = minimize(neg_log_L,
x0=20,
method='powell',
options=dict(maxiter=1000000, maxfev=1000000))
if not res.success:
print '\nWarning: Failure on {} ({})'.format(
seq, seqtools.mm_names(self.target, seq))
return float(res.x)
def fit_Imax_ML_given_Imin(self, ML_seqs, Imin):
idx1 = len(ML_seqs)
idx2 = idx1 + self.IA.course_len
idx3 = idx2 + self.IA.course_len
Imin_list = misc.list_if_scalar(Imin, len(self.concentrations))
def neg_log_L(params):
params = map(abs, params)
Kds = params[:idx1]
Imax_list = params[idx1:idx2]
sig_cs = params[idx2:idx3]
sigI = params[idx3]
return -self.model_logL(ML_seqs, Kds, Imin_list, Imax_list, sig_cs,
sigI)
x0 = list([
self.curve_fit_Kd(seq, self.Imin_const, self.Imax_const)
for seq in ML_seqs
] + self.Imax_adjusted + [1] * self.IA.course_len +
[(self.Imax_const - self.Imin_const) / 10])
assert len(x0) == idx3 + 1
res = minimize(neg_log_L,
x0=x0,
method='powell',
options=dict(maxiter=1000000, maxfev=1000000,
disp=True))
if not res.success:
print '\nWarning: Failure on {} ({})'.format(
seq, seqtools.mm_names(self.target, seq))
params = map(abs, res.x)
Kds = params[:idx1]
Imax_list = params[idx1:idx2]
sig_cs = params[idx2:idx3]
sigI = params[idx3]
return Kds, Imax_list, sig_cs, sigI
