for region in config['selected_data_regions']:
_, start_ppm, y0, end_ppm, y1 = region
if start_ppm < min_ppm:
start_ppm = min_ppm
if end_ppm > max_ppm:
end_ppm = max_ppm
if start_ppm < end_ppm:
start_ppm, end_ppm = end_ppm, start_ppm
# Convert ppm to nearest index
start_idx = scale.index(min(scale, key=lambda x: abs(x - start_ppm)))
end_idx = scale.index(min(scale, key=lambda x: abs(x - end_ppm)))
if start_idx > end_idx:
start_idx, end_idx = end_idx, start_idx
index_mask = index_mask[
~np.logical_and(index_mask > start_idx, index_mask < end_idx)]
regions.append((start_ppm, ymax, end_ppm, ymin))
output_data = output_data.iloc[:, index_mask]
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra
View = spectra(output_data, styles=styles, regions=regions)
import numpy as np minima = input_data[input_data>0].min().min() / 2 # # Get the dso filtered by class input_data[input_data <= 0] = minima output_data = input_data # Generate simple result figure (using pathomx libs) from pathomx.figures import spectra View = spectra(output_data);
output_data.index = pd.MultiIndex.from_tuples(
zip(output_data.index.values, range(1, len(output_data.index.values) + 1)),
names=['Sample'])
if not isinstance(output_data.columns, pd.MultiIndex):
output_data.columns = pd.MultiIndex.from_tuples(
zip(output_data.columns.values, range(1, len(output_data.columns.values) + 1)),
names=['Label', 'Measurement'])
# Fill in header defaults where items are missing
if any([c is None for c in output_data.index.names]):
labels = config['row_header_defaults'].split(',')
output_data.index.names = [l if l is not None else labels[n] for n, l in enumerate(output_data.index.names)]
if any([c is None for c in output_data.columns.names]):
labels = config['column_header_defaults'].split(',')
output_data.columns.names = [l if l is not None else labels[n] for n, l in enumerate(output_data.columns.names)]
# If we get here and don't have a sample Class entry on the index, create it
if 'Class' not in output_data.index.names:
output_data['Class'] = [''] * output_data.shape[0]
output_data.set_index(['Class'], append=True, inplace=True)
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra, heatmap
Spectra = spectra(output_data, styles=styles)
Heatmap = heatmap(output_data)
if len(row)>0 and row[0] == 'metabolite':
sample_nos = row[1:-2]
break
else:
continue # Escape to next file
num_of_samples = len(sample_nos)
# Pass the open f to read_csv to get the rest
# I hate MATLAB strings
dataset = pd.read_csv(f, header=None, index_col=[0], sep=r'\x00*\t+', converters={n+1:np.float64 for n in range(num_of_samples)})
#dataset.convert_objects(convert_numeric=True)
# Bottom two columns are the metabolite id info, chop off
dataset = dataset.T
dataset = dataset[:-2]
dataset = dataset.astype(float)
# We've only got sample items, need to add a class column
sample_nos = [f.replace('(expno_','').strip(')') for f in sample_nos]
dataset.index = pd.MultiIndex.from_tuples(zip(sample_nos,['']*len(sample_nos)), names=['Sample','Class'])
vars()[target] = dataset
del dataset
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra
Raw = spectra(raw, styles=styles);
PQN = spectra(pqn, styles=styles);
TSA = spectra(tsa, styles=styles);
scores.columns = columns
weights = pd.DataFrame(pca.components_)
weights.columns = input_data.columns
dso_pc = {}
weightsi = []
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra, scatterplot, plot_point_cov
for n in range(0, pca.components_.shape[0]):
pcd = pd.DataFrame(weights.values[n:n + 1, :])
pcd.columns = input_data.columns
vars()['PC%d' % (n + 1)] = spectra(pcd, styles=styles)
weightsi.append("PC %d" % (n + 1))
weights.index = weightsi
if config['plot_sample_numbers']:
label_index = 'Sample'
else:
label_index = None
# Build scores plots for all combinations up to n
score_combinations = list(
set([(a, b) for a in range(0, n) for b in range(a + 1, n + 1)]))
for sc in score_combinations:
real = input_data imag = input_data real.loc[:] = real.values.real imag.loc[:] = imag.values.imag # Generate simple result figure (using pathomx libs) from pathomx.figures import spectra Real = spectra(real); Imag = spectra(imag);
# Generate simple result figure (using pathomx libs) from pathomx.figures import spectra View = spectra(input_data, styles=styles)
scoresl.append( 'Latent Variable %d' % (n+1) ) #, plsr.y_weights_[0][n])
scores.columns = scoresl
weights = pd.DataFrame( plsr.x_weights_.T )
weights.columns = input_data.columns
dso_lv = {}
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra, scatterplot
weightsdc=[]
for n in range(0, plsr.x_weights_.shape[1] ):
lvd = pd.DataFrame( plsr.x_weights_[:,n:n+1].T )
lvd.columns = input_data.columns
vars()['LV%d' % (n+1)] = spectra(lvd, styles=styles)
#weightsdl.append("Weights on LV %s" % (n+1))
weightsdc.append("LV %s" % (n+1))
weights.index = weightsdc
Scores = scatterplot(scores, styles=styles)
weightsd = None; # Clean up
lvd = None; # Clean up
Scores
med_sf = config.get('med_sf')
med_sigma = config.get('med_sigma')
# Cbf pc algorithm vars
cbf_last_pc = config.get('cbf_last_pc')
# Cbf explicit algorithm vars
cbf_explicit_start = config.get('cbf_explicit_start')
cbf_explicit_end = config.get('cbf_explicit_start')
for n, dr in enumerate(input_data.values):
if algorithm == 'median':
dr = ng.process.proc_bl.med(dr, mw=med_mw, sf=med_sf, sigma=med_sigma)
elif algorithm == 'cbf_pc':
dr = ng.process.proc_bl.cbf(dr, last=cbf_last_pc)
elif algorithm == 'cbf_explicit':
dr = ng.process.proc_bl.cbf_explicit(dr, calc=slice(cbf_explicit_start, cbf_explicit_end))
input_data.values[n, :] = dr
output_data = input_data
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra
View = spectra(output_data, styles=styles);
elif config['target'] == 'spectra_number':
target = spc[config['spectra_number'], :].reshape(1, -1)
else:
target = config['target']
xCS, ints, ind, target = icoshift(target, spc,
inter=intervals,
n=maximum_shift,
coshift_preprocessing=config['coshift_preprocessing'],
coshift_preprocessing_max_shift=config['coshift_preprocessing_max_shift'],
average2_multiplier=config['average2_multiplier'],
fill_with_previous=config['fill_with_previous'],
)
output_data = input_data.copy()
output_data[:] = xCS
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra, difference
regions = []
ymin, ymax = np.min(spc.flatten()), np.max(spc.flatten())
for r in config['selected_data_regions']:
x0, y0, x1, y1 = r[1:]
regions.append((x0, ymax, x1, ymin))
View = spectra(output_data, styles=styles, regions=regions)
Difference = difference(input_data, output_data)
spc = None
scores.columns = columns
weights = pd.DataFrame(pca.components_)
weights.columns = input_data.columns
dso_pc = {}
weightsi = []
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra, scatterplot
for n in range(0, pca.components_.shape[0]):
pcd = pd.DataFrame(weights.values[n:n + 1, :])
pcd.columns = input_data.columns
vars()['PC%d' % (n + 1)] = spectra(pcd, styles=styles)
weightsi.append("PC %d" % (n + 1))
weights.index = weightsi
if config['plot_sample_numbers']:
label_index = 'Sample'
else:
label_index = None
# Build scores plots for all combinations up to n
score_combinations = set([ (a,b) for a in range(0,n) for b in range(a+1, n+1)])
for sc in score_combinations:
dataset = pd.read_csv(
f,
header=None,
index_col=[0],
sep=r'\x00*\t+',
converters={n + 1: np.float64
for n in range(num_of_samples)})
#dataset.convert_objects(convert_numeric=True)
# Bottom two columns are the metabolite id info, chop off
dataset = dataset.T
dataset = dataset[:-2]
dataset = dataset.astype(float)
# We've only got sample items, need to add a class column
sample_nos = [f.replace('(expno_', '').strip(')') for f in sample_nos]
dataset.index = pd.MultiIndex.from_tuples(zip(sample_nos,
[''] * len(sample_nos)),
names=['Sample', 'Class'])
vars()[target] = dataset
del dataset
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra
Raw = spectra(raw, styles=styles)
PQN = spectra(pqn, styles=styles)
TSA = spectra(tsa, styles=styles)
if config.get('scaling_enabled'):
# Get mean reference peak size
reference_peak_mean = np.mean( [r['scale'] for r in reference_peaks if r ] )
print('Reference peak mean %s' % reference_peak_mean)
# Now scale; using the same peak regions & information (so we don't have to worry about something
# being shifted out of the target region in the first step)
for n,refp in enumerate(reference_peaks):
if refp:
# Scale the spectra
amplitude = reference_peak_mean/refp['amplitude']
input_data.iloc[n] *= amplitude
# -- optionally use the line widths and take max within each of these for each spectra (peak shiftiness)
# Filter the original data with those locations and output\
output_data = input_data
region = output_data.iloc[:,start:end:d]
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra
View = spectra(output_data, styles=styles);
Region = spectra(region, styles=styles);
data = None;
import pandas as pd real = pd.DataFrame(input_data.values.real) real.columns = input_data.columns real.index = input_data.index imag = pd.DataFrame(input_data.values.imag) imag.columns = input_data.columns imag.index = input_data.index # Generate simple result figure (using pathomx libs) from pathomx.figures import spectra Real = spectra(real) Imag = spectra(imag)
scores.columns = scoresl
weights = pd.DataFrame(plsr.x_weights_.T)
weights.columns = input_data.columns
dso_lv = {}
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra, scatterplot
weightsdc = []
for n in range(0, plsr.x_weights_.shape[1]):
lvd = pd.DataFrame(plsr.x_weights_[:, n:n + 1].T)
lvd.columns = input_data.columns
vars()['LV%d' % (n + 1)] = spectra(lvd, styles=styles)
#weightsdl.append("Weights on LV %s" % (n+1))
weightsdc.append("LV %s" % (n + 1))
weights.index = weightsdc
# Build scores plots for all combinations up to n
score_combinations = set([(a, b) for a in range(0, n)
for b in range(a + 1, n + 1)])
if config['plot_sample_numbers']:
label_index = 'Sample'
else:
label_index = None
zip(output_data.columns.values,
range(1,
len(output_data.columns.values) + 1))),
names=['Label', 'Measurement'])
# Fill in header defaults where items are missing
if any([c is None for c in output_data.index.names]):
labels = config['row_header_defaults'].split(',')
output_data.index.names = [
l if l is not None else labels[n]
for n, l in enumerate(output_data.index.names)
]
if any([c is None for c in output_data.columns.names]):
labels = config['column_header_defaults'].split(',')
output_data.columns.names = [
l if l is not None else labels[n]
for n, l in enumerate(output_data.columns.names)
]
# If we get here and don't have a sample Class entry on the index, create it
if 'Class' not in output_data.index.names:
output_data['Class'] = [''] * output_data.shape[0]
output_data.set_index(['Class'], append=True, inplace=True)
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra, heatmap
Spectra = spectra(output_data, styles=styles)
Heatmap = heatmap(output_data)
output_data = input_data.T # Generate simple result figure (using pathomx libs) from pathomx.figures import spectra View = spectra(output_data)
index=input_data.index,
columns=input_data.columns)
if config.get('scaling_enabled'):
# Get mean reference peak size
reference_peak_mean = np.mean([r['scale'] for r in reference_peaks if r])
print('Reference peak mean %s' % reference_peak_mean)
# Now scale; using the same peak regions & information (so we don't have to worry about something
# being shifted out of the target region in the first step)
for n, refp in enumerate(reference_peaks):
if refp:
# Scale the spectra
amplitude = reference_peak_mean / refp['amplitude']
input_data.iloc[n] *= amplitude
# -- optionally use the line widths and take max within each of these for each spectra (peak shiftiness)
# Filter the original data with those locations and output\
output_data = input_data
region = output_data.iloc[:, start:end:d]
# Generate simple result figure (using pathomx libs)
from pathomx.figures import spectra
View = spectra(output_data, styles=styles)
Region = spectra(region, styles=styles)
data = None
