def get_temp_stats(self, percent_skip=0, stride=1):
"""
Returns temperature stats in string format (used for cmdline printout)
for all logfiles combined
"""
temps = self.get_temps(percent_skip=percent_skip, stride=stride)
tt, tf, tf_solu, tf_solv = temps.get_columns(
("T_tot", "T_free", "T_free_solute", "T_free_solvent"))
tt_mean, tt_std = np.mean(tt), np.std(tt)
tf_mean, tf_std = np.mean(tf), np.std(tf)
tf_solu_mean, tf_solu_std = np.mean(tf_solu), np.std(tf_solu)
tf_solv_mean, tf_solv_std = np.mean(tf_solv), np.std(tf_solv)
tt_max_dev = max(map(lambda x: abs(x - tt_mean), tt))
tf_max_dev = max(map(lambda x: abs(x - tf_mean), tf))
tf_solu_max_dev = max(map(lambda x: abs(x - tf_solu_mean), tf_solu))
tf_solv_max_dev = max(map(lambda x: abs(x - tf_solv_mean), tf_solv))
outstr = """\
Temperature stats:
{0:20s}{1:>20s}{2:>20s}{3:>20s}
{4:20s}{5:>20.2f}{6:>20.2f}{7:>20.2f}
{8:20s}{9:>20.2f}{10:>20.2f}{11:>20.2f}
{12:20s}{13:>20.2f}{14:>20.2f}{15:>20.2f}
{16:20s}{17:>20.2f}{18:>20.2f}{19:>20.2f}
""".format("", "Mean", "Stdev", "Max.Abs.Dev.", "T_total", tt_mean, tt_std,
tt_max_dev, "T_free", tf_mean, tf_std, tf_max_dev, "T_free_solute",
tf_solu_mean, tf_solu_std, tf_solu_max_dev, "T_free_solvent",
tf_solv_mean, tf_solv_std, tf_solv_max_dev)
return outstr
def stats_str(self):
"""Free energy stats in string format."""
dgas = self.dgas.values()
dg0s = self.dg0s.values()
dgs_fep = self.dgs_fep.values()
allres = {}
allres["calc_type"] = self._subcalc_key or ""
allres["dg_n"] = len(dgas)
allres["dga"] = (np.mean(dgas), np.std(dgas), np.median(dgas),
np.std_error(dgas))
allres["dg0"] = (np.mean(dg0s), np.std(dg0s), np.median(dg0s),
np.std_error(dg0s))
allres["dg_fep_n"] = len(dgs_fep)
allres["dg_fep"] = (np.mean(dgs_fep), np.std(dgs_fep),
np.median(dgs_fep), np.std_error(dgs_fep))
return """\
# {calc_type:<15} Mean Std.dev Median Std.error N
dG* {dga[0]:10.2f} {dga[1]:10.2f} {dga[2]:10.2f} {dga[3]:10.2f} {dg_n:10}
dG0 {dg0[0]:10.2f} {dg0[1]:10.2f} {dg0[2]:10.2f} {dg0[3]:10.2f} {dg_n:10}
dG_lambda {dg_fep[0]:10.2f} {dg_fep[1]:10.2f} {dg_fep[2]:10.2f} \
{dg_fep[3]:10.2f} {dg_fep_n:10}
""".format(**allres)
def _getmeans(self):
# called by fit_to_reference and _do_iteration
# calls map_all()
# analyses the outputs and returns dG* and dG0 means
# raises QMapperError on total failure
self.mapall(_supress_info=True)
for (qfo, err) in self.failed.iteritems():
logger.info("Failed to map '{}': {}".format(qfo, err))
if not self.mapped:
raise QMapperError("All directories failed to map! Try changing "
"the initial-guess values (Hij and alpha) "
"or step_size... Also, check the mapping "
"parameters (skip, bins, ...).")
dga, dg0 = [], []
for mapdir, (_, qfo_str) in self.mapped.iteritems():
try:
qfo = QFepOutput(qfo_str)
dga.append(qfo.part3.dga)
dg0.append(qfo.part3.dg0)
except QFepOutputError as error_msg:
logger.info("Failed to analyse '{}': {}"
"".format(mapdir, error_msg))
except Exception as error_msg:
logger.warning("Uncaught exception when analysing '{}': {}"
"".format(mapdir, error_msg))
if not dga or not dg0:
raise QMapperError("All directories failed to analyse! Try "
"changing the initial-guess values (Hij and "
"alpha) or step_size...")
return np.mean(dga), np.mean(dg0)
def lra_stats(self):
"""Calculate average and st.dev of LRA and reorg energies."""
average_lras = DataContainer([
"E_type", "(E2-E1)_10_mean", "(E2-E1)_10_std", "(E2-E1)_01_mean",
"(E2-E1)_01_std", "LRA_mean", "LRA_std", "REORG_mean", "REORG_std"
])
allvals = []
for lra in self.lras.values():
rows = lra.get_rows()
for irow, row in enumerate(rows):
try:
allvals[irow].append(row)
except IndexError:
allvals.append([
row,
])
# allvals now looks like this:
# [
# [
# ["EQtot", EQtot_de_st1_1, EQtot_de_st2_1, EQtot_lra_1, EQtot_reorg_1],
# ["EQtot", EQtot_de_st1_2, EQtot_de_st2_2, ...], ...
# ],
# [
# ["EQbond", EQbond_de_st1_1, EQbond_de_st2_1, EQbond_lra_1, EQbond_reorg_1],
# ["EQbond", EQbond_de_st1_2, EQbond_de_st2_2, ...], ...
# ]
# ]
#
for values in allvals:
# transpose to get [ ["EQtot","EQtot"...],
# [ EQtot_de_st1_1, EQtot_de_st1_2,...],
# [ EQtot_de_st2_1, EQtot_de_st2_2,...], ...]
values = zip(*values)
# now they can be easily averaged and std-ed
e_type = values[0][0]
de_st1_mean = np.mean(values[1])
de_st2_mean = np.mean(values[2])
lra_mean = np.mean(values[3])
reo_mean = np.mean(values[4])
de_st1_std = np.std(values[1])
de_st2_std = np.std(values[2])
lra_std = np.std(values[3])
reo_std = np.std(values[4])
average_lras.add_row([
e_type, de_st1_mean, de_st1_std, de_st2_mean, de_st2_std,
lra_mean, lra_std, reo_mean, reo_std
])
return average_lras
def calcall(self):
"""Run the GC calcs, update .gcs, .failed and .gcs_stats.
"""
semaphore = threading.BoundedSemaphore(self._nthreads)
self._qcalc_io.clear()
self.gcs.clear()
self.gcs_stats.delete_rows()
self.failed.clear()
threads = []
for calcdir in self._calcdirs:
threads.append(_QGroupContribThread(self, semaphore, calcdir))
threads[-1].start()
for t in threads:
while t.isAlive():
t.join(1.0)
if self.kill_event.is_set():
try:
t.qcalc.process.terminate()
except Exception as e:
pass
return
if t.error:
self.failed[t.calcdir] = t.error
else:
self._qcalc_io[t.calcdir] = (t.qinps, t.qouts)
# parse the output for results and
# calculate LRAs for each dir
for _dir, (_, qouts) in self._qcalc_io.iteritems():
gcs = []
failed_flag = False
for qout in qouts:
try:
qco = QCalcOutput(qout)
res = qco.results["gc"]
if not self.qcalc_version:
self.qcalc_version = qco.qcalc_version
except (QCalcError, KeyError) as error_msg:
self.failed[_dir] = error_msg
failed_flag = True
break
gc = {}
for row in res.get_rows():
resid, vdw, el = int(row[0]), float(row[1]), float(row[2])
gc[resid] = {"vdw": vdw, "el": el}
gcs.append(gc)
if failed_flag:
continue
resids = sorted(gcs[0].keys())
resnames = [
self._pdb_qstruct.residues[ri - 1].name for ri in resids
]
# do the LRA thingy
# LRA = 0.5*(<E2-E1>_conf1+<E2-E1>_conf2)
# REORG = <E2-E1>_conf1 - LRA
e2e1_st1_vdw = [
gcs[1][key]["vdw"] - gcs[0][key]["vdw"] for key in resids
]
e2e1_st1_el = [
gcs[1][key]["el"] - gcs[0][key]["el"] for key in resids
]
e2e1_st2_vdw = [
gcs[3][key]["vdw"] - gcs[2][key]["vdw"] for key in resids
]
e2e1_st2_el = [
gcs[3][key]["el"] - gcs[2][key]["el"] for key in resids
]
# super efficient stuff here
vdw_lra = [
0.5 * (a + b) for a, b in zip(e2e1_st1_vdw, e2e1_st2_vdw)
]
el_lra = [0.5 * (a + b) for a, b in zip(e2e1_st1_el, e2e1_st2_el)]
vdw_reorg = [
0.5 * (a - b) for a, b in zip(e2e1_st1_vdw, e2e1_st2_vdw)
]
el_reorg = [
0.5 * (a - b) for a, b in zip(e2e1_st1_el, e2e1_st2_el)
]
# scale the ionized residues
if abs(self._scale_ionized - 1.0) > 1e-7:
for i, resname in enumerate(resnames):
if resname in ("ARG", "LYS", "HIP", "ASP", "GLU"):
e2e1_st1_el[i] = e2e1_st1_el[i] / self._scale_ionized
e2e1_st2_el[i] = e2e1_st2_el[i] / self._scale_ionized
el_lra[i] = el_lra[i] / self._scale_ionized
el_reorg[i] = el_reorg[i] / self._scale_ionized
# write the DataContainer
lambda1_st1 = self._lambdas_A[0]
lambda2_st1 = self._lambdas_B[0]
gc_lra = DataContainer([
"Residue_id", "Residue name",
"<E2-E1>1_VdW(l={:5.4f}->l={:5.4f})"
"".format(lambda1_st1, lambda2_st1),
"<E2-E1>1_El(l={:5.4f}->l={:5.4f})_(iscale={})"
"".format(lambda1_st1, lambda2_st1, self._scale_ionized),
"<E2-E1>2_VdW(l={:5.4f}->l={:5.4f})"
"".format(lambda1_st1, lambda2_st1),
"<E2-E1>2_El(l={:5.4f}->l={:5.4f})_(iscale={})"
"".format(lambda1_st1, lambda2_st1,
self._scale_ionized), "LRA_VdW(l={:5.4f}->l={:5.4f})"
"".format(lambda1_st1, lambda2_st1),
"LRA_El(l={:5.4f}->l={:5.4f})_(iscale={})"
"".format(lambda1_st1, lambda2_st1, self._scale_ionized),
"REORG_VdW(l={:5.4f}->l={:5.4f})"
"".format(lambda1_st1, lambda2_st1),
"REORG_El(l={:5.4f}->l={:5.4f})_(iscale={})"
"".format(lambda1_st1, lambda2_st1, self._scale_ionized)
])
for row in zip(resids, resnames, e2e1_st1_vdw, e2e1_st1_el,
e2e1_st2_vdw, e2e1_st2_el, vdw_lra, el_lra,
vdw_reorg, el_reorg):
gc_lra.add_row(row)
self.gcs[_dir] = gc_lra
# get GC stats over all directories
self.gcs_stats.delete_rows()
gcs = {}
for _, gc in self.gcs.iteritems():
for row in gc.get_rows():
resid, resname = row[0:2]
res_key = "{}.{}".format(resid, resname)
values = [[
val,
] for val in row[2:]]
if not gcs.has_key(res_key):
gcs[res_key] = values
else:
for i, val in enumerate(gcs[res_key]):
val.extend(values[i])
# iterate through each residue and calculate
# means and stdevs
# (sort by residue index)
for res_key in sorted(gcs.keys(), key=lambda x: int(x.split(".")[0])):
rc = gcs[res_key]
resid, resname = res_key.split(".")
# get mean and stdev
rc_stats = [
int(resid),
resname,
len(rc[0]),
np.mean(rc[0]),
np.std(rc[0]), # <E2-E1>1 vdw
np.mean(rc[1]),
np.std(rc[1]), # <E2-E1>1 el
np.mean(rc[2]),
np.std(rc[2]), # <E2-E1>2 vdw
np.mean(rc[3]),
np.std(rc[3]), # <E2-E1>2 el
np.mean(rc[4]),
np.std(rc[4]), # LRA vdw
np.mean(rc[5]),
np.std(rc[5]), # LRA el
np.mean(rc[6]),
np.std(rc[6]), # REORG vdw
np.mean(rc[7]),
np.std(rc[7])
] # REORG el
self.gcs_stats.add_row(rc_stats)
def test_mean(self):
vals = [i**2 for i in range(1, 21)]
assert is_close(np.mean(vals), 143.5)