"""
Reproducibility plots for deviations from sample-group means.
Generates per-replicate histograms of deviations from group mean and
lays them out in a grid for quick visual inspection.
"""
import pandas as pd
from dash.dcc import Graph
from plotly import express as px
from plotly import graph_objects as go
from plotly.subplots import make_subplots
[docs]
def get_reproducibility_dataframe(data_table: pd.DataFrame, sample_groups: dict) -> pd.DataFrame:
"""Compute per-replicate deviations from the group mean.
:param data_table: DataFrame with sample columns.
:param sample_groups: Mapping group -> list of column names.
:returns: Nested dict ``{group: {replicate: [deviations...]}}`` suitable for plotting.
"""
repro_data: dict = {}
for sample_group, sample_columns in sample_groups.items():
sgroup_data_table: pd.DataFrame = data_table[sample_columns]
mean: pd.Series = sgroup_data_table.mean(axis=1)
for col in sgroup_data_table.columns:
if sample_group not in repro_data:
repro_data[sample_group] = {}
coldata: pd.Series = sgroup_data_table[col].dropna()
coldata = coldata-mean.loc[coldata.index]
repro_data[sample_group][col] = list(coldata.values)
return repro_data
[docs]
def get_max(plot_data: dict, minval=2) -> int:
"""Return the maximum absolute deviation across all replicates.
:param plot_data: Output from ``get_reproducibility_dataframe``.
:param minval: Minimum value to enforce.
:returns: Integer maximum used for x-axis range sizing.
"""
maxval: int = minval
for _, sgroup_data in plot_data.items():
for __, replicate_values in sgroup_data.items():
maxval = max([maxval, max(replicate_values)])
return int(maxval)
[docs]
def make_graph(graph_id: str, defaults: dict, plot_data: dict, title: str, table_type: str, dlname:str, num_per_row: int = 2) -> Graph:
"""Create a grid of histograms of deviations from group mean.
:param graph_id: Component ID for the ``Graph``.
:param defaults: Dict with ``height``, ``width``, ``config``.
:param plot_data: Nested dict from ``get_reproducibility_dataframe``.
:param title: Plot title.
:param table_type: Label for x-axis description.
:param dlname: Name for the downloaded figure file.
:param num_per_row: Number of subplots per row.
:returns: Dash ``Graph`` with the subplot figure.
"""
sample_groups: list = sorted(list(plot_data.keys()))
num_plots: int = len(sample_groups)
rows: int = int(num_plots/num_per_row)
if num_plots % num_per_row != 0:
rows += 1
fig: go.Figure = make_subplots(
rows=rows,
cols=num_per_row,
subplot_titles=sample_groups,
x_title=f'Value deviation from mean ({table_type})',
# y_title='Percent of values',
y_title='Count',
# 0.05 # 3 riviä = 0.05, 2 = 0.1 5 = 0.025
vertical_spacing=0.1/max(1, (rows-1))
)
xmax: int = get_max(plot_data)
xsize: int = int((xmax*2)/50)
xbins: dict = {'start': -xmax, 'end': xmax, 'size': xsize}
current_row: int = 1
current_col: int = 1
for sgroup in sample_groups:
sgroup_data: dict = plot_data[sgroup]
for replicate, rep_list in sgroup_data.items():
fig.add_trace(
go.Histogram(
x=rep_list,
name=replicate,
showlegend=False,
xbins=xbins,
# histnorm='percent'
),
row=current_row,
col=current_col
)
current_col += 1
if current_col > num_per_row:
current_col = 1
current_row += 1
fig.update_layout(
barmode='overlay',
title=title,
height=defaults['height'] * (rows) * 0.5,
width=defaults['width'],
hovermode='x unified'
)
config = defaults['config'].copy()
config['toImageButtonOptions'] = config['toImageButtonOptions'].copy()
config['toImageButtonOptions']['filename'] = dlname
# xmax = int(xmax*0.25)
tick_distance: int = max(1, round(xmax/4))
fig.update_xaxes(range=[-xmax, xmax], dtick=tick_distance)
fig.update_traces(opacity=0.5)
return Graph(id=graph_id, figure=fig, config=config)