def post_collection(self, name, analysis_ids):
self.load_dataframes()
parent_collections = [
get_collection(current_user, collection_id) for collection_id in self.loaded_collection_ids
]
label_df = self._label_df
# merge collection attributes
if len(self.loaded_collection_ids) > 1:
collection_lengths = [
len(self._label_df[self._label_df.original_collection_id == collection_id])
for collection_id in self.loaded_collection_ids
]
collection_attrs = [collection.get_file_attributes() for collection in parent_collections]
all_attr_keys = set.union(*[set(collection_attr.keys()) for collection_attr in collection_attrs])
attr_values = {
attr_key: [
collection_attr[attr_key] if attr_key in collection_attr else None
for collection_attr in collection_attrs
]
for attr_key in all_attr_keys
}
attrs = {}
for attr_key, attr_values_ in attr_values.items():
valid_values = [v for v in attr_values_ if v is not None]
if len(valid_values) == 1:
attrs[attr_key] = valid_values[0]
else:
vals = [[v for _ in range(length)] for v, length in zip(valid_values, collection_lengths)]
label_df[attr_key] = [item for sublist in vals for item in sublist]
else:
attrs = get_collection(current_user, self.loaded_collection_ids[0]).get_file_attributes()
filename = os.path.join(self.root_dir, 'processed.h5')
attrs['processing_log'] = self.processing_log
description = '\n\n'.join([collection.description for collection in parent_collections] + [self.processing_log])
new_data = {
'name': name,
'description': description,
'analysis_ids': analysis_ids,
'parent_collection_id': self.loaded_collection_ids[0],
'group_can_read': all([collection.group_can_read for collection in parent_collections]),
'all_can_read': all([collection.all_can_read for collection in parent_collections]),
'group_can_write': all([collection.group_can_write for collection in parent_collections]),
'all_can_write': all([collection.all_can_write for collection in parent_collections])
}
self.write_collection(self._numeric_df, label_df, attrs, filename)
new_collection = upload_collection(current_user, filename, new_data)
return new_collection
def results_filename(self):
if self.results_collection_id is not None:
try:
return get_collection(current_user,
self.results_collection_id).filename
except:
return None
return None
def get_collections(self, collection_ids):
super().get_collections(collection_ids)
try:
self.processing_log = get_collection(current_user, collection_ids[0]).get_attr('processing_log')
except Exception as e:
print(e)
self.processing_log = ''
x = [float(i) for i in self._numeric_df.columns]
self.x_axis_range = [max(x), min(x)]
y_max = np.max(self._numeric_df.values)
self.y_axis_range = [-0.05*y_max, 1.05 * y_max]
def render_collection(collection_id=None):
try:
current_user = get_current_user()
collection = get_collection(current_user, collection_id)
if request.method == 'DELETE':
delete_collection(current_user, collection_id)
return redirect(url_for('collections.render_collection_list'))
if request.method == 'POST':
dataset_name = request.form.get('dataset_name')
dataset_type = request.form.get('dtype')
dt.collections.create_new_label_dataset(current_user, collection,
dataset_name, dataset_type)
return render_template('pages/collection_entry.html',
page_data=CollectionPageData(
current_user, collection))
except Exception as e:
return handle_exception_browser(e)
def _save_loadings(group_key, name):
with h5py.File(self.results_filename, 'r') as results_file:
group = results_file[group_key]
values = [
np.array(group['pls']['x_loadings']),
np.array(group['feature_p_values']),
np.array(group['opls']['P_ortho'])
]
columns = ['x_loadings', 'p_value'] + [
f'P_ortho[{i}]' for i in range(values[2].shape[1])
]
input_collection_id = results_file.attrs[
'input_collection_id'] if 'input_collection_id' in results_file.attrs else None
if input_collection_id is not None:
try:
input_collection = get_collection(current_user,
input_collection_id)
x = input_collection.get_dataset('x')
try:
x_min = input_collection.get_dataset('x_min')
x_max = input_collection.get_dataset('x_max')
except:
x_min = x_max = None
except:
x_min = x_max = x = None
else:
x_min = x_max = x = None
df = pd.DataFrame(np.column_stack(values), columns=columns)
if x is not None and len(x) == len(df):
df.index = x
df.index.name = 'x'
if x_min is not None and len(x_min) == len(df):
df['x_min'] = x_min
df['x_max'] = x_max
df = df[['x_min', 'x_max'] + columns]
if file_format == 'csv':
df.to_csv(name)
def get_plot(self, queries, group_by, labels, theme, bin_collection_id, legend_style, background_color):
print(background_color)
labels = labels or []
self.load_dataframes()
if bin_collection_id is not None:
print(bin_collection_id)
bin_collection = get_collection(current_user, bin_collection_id)
x_mins = bin_collection.get_dataset('x_min').ravel().tolist()
x_maxes = bin_collection.get_dataset('x_max').ravel().tolist()
colors = [DEFAULT_PLOTLY_COLORS[i % 2] for i in range(len(x_mins))]
shapes = [
go.layout.Shape(
type='rect',
xref='x',
yref='paper',
x0=x_min,
y0=0,
x1=x_max,
y1=1,
fillcolor=color,
opacity=0.2,
layer='below',
line_width=0
)
for x_min, x_max, color in zip(x_mins, x_maxes, colors)
]
else:
shapes = []
axis_line_style = {
'zerolinecolor': '#375A7F', # darkly primary
'gridcolor': '#444444' # darkly secondary
} if theme == 'plotly_dark' and background_color != 'rgba(255,255,255,1)' else {
'zerolinecolor': '#2C3E50', # flatly primary
'gridcolor': '#95A5A6' # flatly secondary
}
if legend_style in ('full', 'groups'):
layout = go.Layout(
height=700,
font={'size': 16},
margin={'t': 25, 'l': 25, 'b': 25, 'r': 25},
template=theme,
plot_bgcolor=background_color,
paper_bgcolor=background_color,
xaxis={
'title': 'Chemical Shift (ppm)',
'autorange': 'reversed',
**axis_line_style
},
yaxis={
'title': 'Intensity',
**axis_line_style
},
shapes=shapes
)
else: # if legend_style == 'none'
layout = go.Layout(
height=700,
font={'size': 16},
margin={'t': 25, 'l': 25, 'b': 25, 'r': 25},
template=theme,
plot_bgcolor=background_color,
paper_bgcolor=background_color,
xaxis={
'title': 'Chemical Shift (ppm)',
'autorange': 'reversed',
**axis_line_style
},
yaxis={
'title': 'Intensity',
**axis_line_style
},
shapes=shapes,
showlegend=False
)
color_indices = [self._label_df.query(query).index for query in queries]
if len(color_indices) > len(DEFAULT_PLOTLY_COLORS): # repeat default color list
colors = []
while len(colors) < len(color_indices):
colors += DEFAULT_PLOTLY_COLORS
else:
colors = DEFAULT_PLOTLY_COLORS
colors = colors[:len(color_indices)]
x = self._numeric_df.columns.values.astype(float)
figure = go.Figure(layout=layout)
if legend_style == 'full' or legend_style == 'groups':
figure.add_trace(
go.Scatter( # dummy series to use as stand-in for legend title
x=[0],
y=[0],
name=','.join(group_by),
mode='markers',
marker={
'opacity': 0,
'size': 0,
'color': 'rgba(0,0,0,0)'
}
)
)
for query, color in zip(queries, colors):
# split query
figure.add_trace(
go.Scatter( # dummy series to label colors
x=[0],
y=[0],
name=','.join(re.findall(r'["](\w+)["]', query)), # pretty kludgy
mode='lines',
marker={'color': color},
legendgroup=query
)
)
figure.add_trace(
go.Scatter( # dummy series to provide space between color key and "heading"
x=[0],
y=[0],
name='',
mode='markers',
marker={
'opacity': 0,
'size': 0,
'color': 'rgba(0,0,0,0)'
}
)
)
if legend_style == 'full':
figure.add_trace(
go.Scatter( # dummy series to use as stand-in for legend title
x=[0],
y=[0],
name=f"({', '.join(labels)})" if len(labels) else 'Spectrum #',
mode='markers',
marker={
'opacity': 0,
'size': 0,
'color': 'rgba(0,0,0,0)'
}
)
)
for query, color in zip(queries, colors):
y_values = self._numeric_df.loc[self._label_df.query(query).index]
for i, row in y_values.iterrows():
text = '<br>'.join([f'{label}=={self._label_df.loc[i][label]}' for label in self._label_df.columns])
if len(labels):
name = f"({', '.join([f'{self._label_df.loc[i][label]}' for label in labels])})"
else:
name = f'({i})'
if legend_style == 'groups':
figure.add_trace(
go.Scatter(
x=x,
y=row,
text=text,
name=','.join(re.findall(r'["](\w+)["]', query)), # pretty kludgy
mode='lines',
marker={'color': color, 'size': 1},
legendgroup=query,
showlegend=False
)
)
else:
figure.add_trace(
go.Scatter(
x=x,
y=row,
text=text,
name=name,
mode='lines',
marker={'color': color, 'size': 2},
showlegend=(legend_style == 'full')
)
)
return figure
def get_loading_significance_table(self, group_key, theme=None, wrap=True):
description = h5py.File(
self.results_filename)[group_key].attrs['description']
if self.results_file_ready:
theme, style_header, style_cell = self._get_table_styles(theme)
with h5py.File(self.results_filename, 'r') as file:
feature_labels = np.array(file[group_key]['feature_labels'])
loadings = np.array(
file[group_key]['pls']['x_loadings']).ravel()
p_values = np.array(file[group_key]['feature_p_values'])
alpha = file[group_key].attrs['outer_alpha']
base_collection_id = file.attrs[
'input_collection_id'] if 'input_collection_id' in file.attrs else None
x_min = x_max = None
if base_collection_id is not None:
try:
base_collection = get_collection(current_user,
int(base_collection_id))
x = base_collection.get_dataset('/x').ravel()
x_min = base_collection.get_dataset('/x_min').ravel()
x_max = base_collection.get_dataset('/x_max').ravel()
x_min = x_min[np.in1d(x, feature_labels)]
x_max = x_max[np.in1d(x, feature_labels)]
except Exception as e:
x_min = x_max = None
valid_bin_boundaries = (x_min is not None and x_max is not None
and x_max.shape[0] == x_min.shape[0] ==
feature_labels.shape[0])
is_significant = p_values < alpha
df = pd.DataFrame()
df['Bin'] = feature_labels
df['Loading'] = loadings
if valid_bin_boundaries:
df['Bin Max'] = x_max
df['Bin Min'] = x_min
df['p Value'] = p_values
df['Significant'] = ['*' if s else '' for s in is_significant]
df = df.sort_values(['Significant', 'Bin'],
ascending=[False, True])
# format table for better display in browser
df['p Value'] = df['p Value'].round(7).apply(
lambda val: f'{val:.7f}')
df['Bin'] = df['Bin'].round(4).apply(lambda val: f'{val:.4f}')
df['Loading'] = df['Loading'].round(5).apply(
lambda val: f'{val:.5f}')
if valid_bin_boundaries:
df['Bin Max'] = df['Bin Max'].round(4).apply(
lambda val: f'{val:.4f}')
df['Bin Min'] = df['Bin Min'].round(4).apply(
lambda val: f'{val:.4f}')
del df['Significant']
df['Index'] = [str(i) for i in df.index]
if valid_bin_boundaries:
df = df[[
'Index', 'Bin Max', 'Bin', 'Bin Min', 'Loading', 'p Value'
]]
else:
df = df[['Index', 'Bin', 'Loading', 'p Value']]
style_data_conditional = [{
'if': {
'filter_query': f'{{p Value}} < {alpha}'
},
'backgroundColor': '#D2F9F1'
}, {
'if': {
'filter_query': f'{{p Value}} > {alpha}'
},
'backgroundColor': '#F9D9D2'
}]
table = dash_table.DataTable(
id=f'feature-table',
columns=[{
'name': i,
'id': i
} for i in df.columns],
data=df.to_dict('records'),
style_table={
'height': '500px',
'overflowY': 'scroll'
},
fixed_rows={
'headers': True,
'data': 0
},
is_focused=True,
style_data_conditional=style_data_conditional,
style_header=style_header,
style_cell=style_cell,
style_cell_conditional=[
{
'if': {
'column_id': 'Index'
},
'width': f'{max(df.Index.str.len().max(), 5) + 2}ch'
},
{
'if': {
'column_id': 'Bin'
},
'width': f'{df.Bin.str.len().max() + 2}ch'
},
{
'if': {
'column_id': 'Bin Max'
},
'width': f'{df.Bin.str.len().max() + 2}ch'
},
{
'if': {
'column_id': 'Bin Min'
},
'width': f'{df.Bin.str.len().max() + 2}ch'
},
{
'if': {
'column_id': 'p Value'
},
'width': '12ch'
},
])
if wrap:
return html.Div(
[dbc.Row(html.H4(description)),
dbc.Row(table)])
else:
return table
return [dash_table.DataTable(id='feature-table')]
