def single_pH(filename, t0, ref_filename=None):
"""Perform fitting of Series in one file and calculate derived quantities.
Parameters
----------
filename : file with peak height vs time data at one pH*
ref_filename : file with SPHERE reference exchange rates at one pH
Returns
-------
DataFrame indexed by residue, with decay rate, error,
and derived quantities for each residue
"""
data = nmrfn.parse_hx(filename, t0)
fits = [fit_decay(data[res], a0=4e7, k0=1e-6) for res in data]
fits = pd.concat(fits, axis=1, keys=data.columns).T
if ref_filename is None:
# Make a handy default.
ref_filename = extract_pH(filename) + "_sphere"
ref = nmrfn.sphere_file(ref_filename)
R = 1.986e-3 # Universal gas constant in kcal K^(-1) mol^(-1)
T = 298.0 # Temperature in K
fits["dg"] = -R * T * np.log(fits["k_obs"] / ref) # Log here is natural log
fits["dger"] = R * T * fits["k_error"] / fits["k_obs"]
# fits['lpf'] = np.log10(ref/fits['k_obs']) # Log_10(protectionfactor)
return fits
def hdx_simulation(f, t0="manual"):
"""Simulation of H-D exchange decay to assess accuracy of fitting.
Generates data from exponential, adds random noise, fits noisy series.
Prints results as a table to inspect visually.
Parameters
==========
f : string filename (Sparky rh table), provides realistic time values
t0 : string, default 'manual' applies if timepoints in seconds are already
encoded in the Sparky rh file columns; otherwise should be a timestamp
in the format '%Y-%m-%d_%H:%M'
"""
t_series = nmrfn.parse_hx(f, t0)
k_input_names = (
["5e-05"] * 4 + ["1e-05"] * 4 + ["5e-06"] * 4 + ["1e-06"] * 4 + ["5e-07"] * 4 + ["1e-07"] * 4 + ["5e-08"] * 4
)
noise_scale_names = ["0.01", "0.05", "0.1", "0.2"] * 7
arrays = [k_input_names, noise_scale_names]
tuples = zip(*arrays)
index = pd.MultiIndex.from_tuples(tuples, names=["k_in", "noiz"])
columns = ["rat_avg", "k_avg", "ker_avg", "k_std", "er_rat", "dg_sd"]
df = pd.DataFrame(columns=columns, index=index)
k_in = {"5e-05": 5e-5, "1e-05": 1e-5, "5e-06": 5e-6, "1e-06": 1e-6, "5e-07": 5e-7, "1e-07": 1e-7, "5e-08": 5e-8}
noise_in = {"0.01": 0.01, "0.05": 0.05, "0.1": 0.1, "0.2": 0.2}
a = 7e7
for kkey, kval in k_in.items():
cln = nmrfn.decay(t_series, a, kval) # clean data
for nkey, nval in noise_in.items():
ratios = []
k_fits = []
k_errs = []
while len(k_fits) < 1000:
noisy = cln + np.random.normal(loc=0, scale=nval * a, size=len(cln))
ratios.append(noisy[-6:-1].mean() / noisy[0:4])
a0 = 4e7
k0 = 1e-6
popt, pcov = curve_fit(nmrfn.decay, xdata=t_series, ydata=noisy, p0=[a0, k0])
k_fits.append(popt[1])
k_errs.append(np.sqrt(pcov.diagonal()[1]))
ratios = np.array(ratios)
k_fits = np.array(k_fits)
k_errs = np.array(k_errs)
k_avg = k_fits.mean()
k_er = k_errs.mean()
k_sd = k_fits.std()
RT = 1.986e-3 * 298.0
df.ix[kkey, nkey]["rat_avg"] = "{0:.2f}".format(ratios.mean())
df.ix[kkey, nkey]["k_avg"] = "{0:.2e}".format(k_avg)
df.ix[kkey, nkey]["ker_avg"] = "{0:.2e}".format(k_er)
df.ix[kkey, nkey]["er_rat"] = "{0:.2f}".format(k_er / k_avg)
df.ix[kkey, nkey]["k_std"] = "{0:.2e}".format(k_sd)
df.ix[kkey, nkey]["dg_sd"] = "{0:.2f}".format(RT * k_sd / k_avg)
print(df)
