Generate Upsetplots#

In this notebook we will be generating two upsetplots: the first will plot counts of cells based on if they have an entry for the associated annotations, and the second will plot counts of cells based on their values for the associated annotations.

import logging
import pandas as pd
import numpy as np
from itertools import chain, combinations
from upsetplot import plot
from matplotlib import pyplot as plt
from pkg.platy import _get_folder, load_connectome_normal_lcc_annotations

We import the folder to save our upsetplots and the table of annotations which will help us make our upsetplots

folder = _get_folder()
annotations = load_connectome_normal_lcc_annotations()
annotations.fillna("N/A", inplace=True)
annotations

	skids	side	class	segment	type	group
0	2015233	l	s	head	100.0	N/A
1	1548290	l	N/A	1	N/A	N/A
2	1318919	l	s	head	88.0	15.0
3	2015241	l	s	head	100.0	N/A
4	1646603	r	N/A	3	N/A	N/A
...	...	...	...	...	...	...
2696	1302513	l	s	head	N/A	N/A
2697	1630186	l	N/A	2	N/A	N/A
2698	1441779	r	N/A	head	N/A	N/A
2699	1671147	r	m	1	165.0	N/A
2700	1048573	l	i	3	N/A	N/A

2701 rows × 6 columns

The following function returns a series where the indices are all the skids in the annotations table and the values are if they have a value for the category passed in

#returns boolean pandas series: False if element is "N/A", else True
def has_category(annotations, category="side"):
    has_list = []
    series_name = "has_{}".format(category)
    for val in annotations[category]:
        if val != "N/A":
            has_list.append(True)
        else:
            has_list.append(False)
    return pd.Series(index=annotations["skids"], data=has_list, name=series_name)

#generate boolean pandas series for the different categories
has_categs = {"has_side": None, "has_class": None, "has_segment": None, "has_type": None, "has_group": None}
for key in has_categs:
    has_categs[key] = has_category(annotations, category=key.split("_")[1])

With series of booleans for all our categories, we can concatenate our series to make a dataframe, and call groupby which will group the dataframe based on the number of cells for each subset of category values (ex: how many cells that have [“has_side”=False, “has_class”=False, “has_segment”=False, “has_type”=False, “has_group”=False] and so on).

bool_ids = list(has_categs.values())

#boolean dataframe from concatenating the boolean series
annotations_bool = pd.concat(bool_ids, axis=1, ignore_index=False, names="ID").fillna(
    "N/A"
)
#group the annotations dataframe by the categories and get the number of cells for each subset of category values
bool_counts = annotations_bool.groupby(bool_ids).size()
print(annotations_bool)
print(bool_counts)

         has_side  has_class  has_segment  has_type  has_group
skids                                                         
2015233      True       True         True      True      False
1548290      True      False         True     False      False
1318919      True       True         True      True       True
2015241      True       True         True      True      False
1646603      True      False         True     False      False
...           ...        ...          ...       ...        ...
1302513      True       True         True     False      False
1630186      True      False         True     False      False
1441779      True      False         True     False      False
1671147      True       True         True      True      False
1048573      True       True         True     False      False

[2701 rows x 5 columns]
has_side  has_class  has_segment  has_type  has_group
False     False      False        False     False          2
                     True         False     False          1
          True       True         False     False          1
                                  True      False          1
True      False      False        False     False          3
                     True         False     False        908
                                            True          18
                                  True      False          6
                                            True           2
          True       True         False     False        533
                                            True         437
                                  True      False        595
                                            True         194
dtype: int64

#generate the upsetplot of the grouped dataframe and save it
plot(bool_counts)
plt.savefig(folder / "broad_categs_upsetplot.png")

We start generating the second upsetplot by using the following function, which determines if all the skids have the particular name within the particular category

#returns boolean pandas series: True if element=name, else False
def has_element(annotations, category="side", name="l"):
    has_element = []
    series_name = "{}_{}".format(category, name)
    for val in annotations[category]:
        if val == name:
            has_element.append(True)
        else:
            has_element.append(False)
    return pd.Series(index=annotations["skids"], data=has_element, name=series_name)

For this upsetplot, we just consider the values that can be passed by the categories side, class and segment, and splice the keys based on the category so we can retrieve the values we want to pass in

#generate boolean pandas series for the different names within their respective category

has_sides = {"has_left": None, "has_right": None, "has_center": None}
for key in has_sides:
    has_sides[key] = has_element(annotations, category="side", name=key.split("_")[1][0])

has_classes = {"has_sensory": None, "has_inter": None, "has_motor": None}
for key in has_classes:
    has_classes[key] = has_element(annotations, category="class", name=key.split("_")[1][0])

has_segments = {"has_seg0": None, "has_seg1": None, "has_seg2": None, "has_seg3": None}
for key in has_segments:
    has_segments[key] = has_element(annotations, category="segment", name=key[-1])

has_hp = {"has_head": None, "has_pygidium": None}
for key in has_hp:
    has_hp[key]= has_element(annotations, category="segment", name=key.split("_")[1])
has_all = has_sides | has_classes | has_segments | has_hp

bool_ids = list(has_all.values())

#boolean dataframe from concatenating the boolean series
annotations_bool = pd.concat(bool_ids, axis=1, ignore_index=False, names="ID").fillna(
    "N/A"
)

#group the annotations dataframe by the names and get the number of cells for each subset of names
bool_counts = annotations_bool.groupby(bool_ids).size()

#generate the upsetplot of the grouped dataframe
plot(bool_counts)
plt.savefig(folder / "sub_categs_upsetplot.png")

Platynereis Data

Generate Upsetplots

Generate Upsetplots#