Plot

Functions to plot motif logo, heatmap, scatter plot, and others.

Overview

Utilities

To set seaborn resolution for notebook display:

set_sns(
    dpi=300,  # resolution for figure display
)

To save plot as SVG with editable text:

save_svg(
    path='figure.svg',  # output file path
)

To save plot as PDF with Illustrator-compatible fonts:

save_pdf(
    path='figure.pdf',  # output file path
)

To show plot or save to file:

save_show(
    path='figure.png',  # image path; if None, calls plt.show()
    show_only=False,    # if True, always show instead of save
)

---

Color & Palette

To assign colors to a list of category names:

color_dict = get_color_dict(
    categories=['A', 'B', 'C'],  # list of names to assign colors
    palette='tab20',             # seaborn palette name
)

To get a list of colors for plotting from a dict, list, or palette name:

colors = get_plt_color(
    palette=color_dict,  # dict, list, or palette name like 'tab10'
    columns=df.columns,  # column names to match colors
)

To filter hue categories by count threshold (for cleaner plots):

hue_big = get_hue_big(
    df=info,        # dataframe containing categories
    hue_col='group',  # column name for hue
    cnt_thr=10,     # minimum count to include category
)

---

Dimensionality Reduction

To reduce dimensionality using PCA, TSNE, or UMAP:

embedding_df = reduce_feature(
    df=df,           # dataframe of values
    method='pca',    # 'pca', 'tsne', or 'umap'
    complexity=20,   # perplexity for TSNE, n_neighbors for UMAP
    n=2,             # number of components
    load=None,       # path to load pre-trained model
    save=None,       # path to save trained model
    seed=123,        # random state for reproducibility
)

---

2D Scatter Plots

To plot a 2D scatter from pre-computed embeddings:

plot_2d(
    embedding_df=feat,     # dataframe with 2 columns (e.g., PCA1, PCA2)
    hue=info['group'],     # series/list for coloring points
    palette='tab20',       # color palette
    legend=True,           # show legend
    legend_title='Group',  # custom legend title
    name_list=None,        # list of names to annotate points
    s=20,                  # dot size
)

To reduce dimensions and plot in one step:

plot_cluster(
    df=df,              # dataframe of values
    method='umap',      # 'pca', 'tsne', or 'umap'
    hue=info['group'],  # series for coloring
    complexity=15,      # perplexity/n_neighbors
    palette='tab20',    # color palette
    legend=True,        # show legend
    name_list=None,     # list of names to annotate
    seed=123,           # random state
    s=50,               # dot size
)

To create an interactive Bokeh plot with search box:

plot_bokeh(
    X=embedding_df,       # 2-column dataframe from reduce_feature
    idx=info['Name'],     # identity labels for search
    hue=info['Group'],    # category for coloring
    s=7,                  # dot size
    smiles=info['SMILES'], # additional hover info (kwargs)
)

---

Rank & Distribution

To plot ranked values with top/bottom annotations:

plot_rank(
    sorted_df=sorted_df,  # pre-sorted dataframe
    x='kinase',           # column for x-axis labels
    y='values',           # column for y-axis values
    n_hi=10,              # number of top values to label
    n_lo=10,              # number of bottom values to label
    figsize=(10, 8),      # figure size
)

To plot a histogram with KDE:

plot_hist(
    df=sorted_df,      # dataframe with values
    x='values',        # column name for histogram
    figsize=(6, 2),    # figure size
)

---

Bar & Box Plots

To plot horizontal bar counts from value_counts:

plot_count(
    cnt=df['col'].value_counts(),  # value counts series
    tick_spacing=10,               # x-axis tick interval
    palette='tab20',               # color palette
)

To plot bar graph with optional strip dots:

plot_bar(
    df=info,           # unstacked dataframe
    value='MW',        # column name for values
    group='category',  # column name for grouping
    title='Title',     # plot title
    figsize=(12, 5),   # figure size
    dots=True,         # add strip dots
    rotation=90,       # x-tick rotation
    ascending=False,   # sort order
)

To plot grouped bar chart for multiple value columns:

plot_group_bar(
    df=info,                    # dataframe
    value_cols=['pKa1', 'pKb2'],  # list of value columns
    group='Name',               # grouping column
    figsize=(12, 5),            # figure size
    order=None,                 # custom x-axis order
    rotation=90,                # x-tick rotation
)

To plot box plot with optional dots:

plot_box(
    df=info,           # dataframe
    value='MW',        # value column
    group='category',  # grouping column
    figsize=(6, 3),    # figure size
    dots=True,         # show individual points
    rotation=90,       # x-tick rotation
)

---

Violin Plot

To plot violin plot with optional strip dots:

ax = plot_violin(
    data=df_long,        # long-form dataframe
    value='value',       # column for values
    group='variable',    # column for grouping
    ylabel='Accuracy',   # y-axis label
    dots=True,           # add strip dots
    figsize=(5, 3),      # figure size
)

To add statistical annotations to violin/box plots:

add_stats(
    ax=ax,                 # axes from plot_violin
    data=df_long,          # same dataframe
    value='value',         # value column
    group='variable',      # group column
    pairs=None,            # comparison pairs (auto-generated if None)
    test='t-test_ind',     # statistical test
    text_format='star',    # annotation format ('star' or 'simple')
)

---

Correlation & Heatmap

To plot correlation scatter with Spearman/Pearson stats:

plot_rel(
    df=norm,                   # dataframe with columns to correlate
    x='AAK1',                  # x-axis column
    y='BIKE',                  # y-axis column
    method='spearman',         # 'spearman' or 'pearson'
    text_location=(0.8, 0.1),  # position for stats text
    index_list=None,           # indices to annotate
    hue=None,                  # optional hue column
    reg_line=True,             # show regression line
)

To calculate distance/similarity matrices:

dist_df, sim_df = get_similarity(
    df=df,              # dataframe of features
    metric='euclidean',  # distance metric
)

To plot correlation/similarity heatmap:

plot_corr(
    df_corr=corr,          # correlation matrix
    inverse_color=False,   # reverse colormap
    figsize=(15, 10),      # figure size
)

---

Evaluation Metrics

To plot AUCDF (Area Under CDF) for ranking evaluation:

aucdf = get_AUCDF(
    df=sorted_df,                   # dataframe with rank values
    col='values',                   # column name for ranks
    reverse=False,                  # reverse CDF direction
    plot=True,                      # show plot
    xlabel='Rank of reported kinase',  # x-axis label
)

To plot normalized confusion matrix:

plot_confusion_matrix(
    target=y_true,            # true labels
    pred=y_pred,              # predicted labels
    class_names=['0', '1'],   # class label names
    normalize=True,           # normalize by row
    title='Confusion matrix',  # plot title
    cmap=plt.cm.Blues,        # colormap
)

---

Pie & Composition

To plot pie chart from value counts:

plot_pie(
    value_counts=df['col'].value_counts(),  # value counts series
    hue_order=None,        # custom category order
    labeldistance=1,       # label distance from center
    fontsize=10,           # label font size
    fontcolor='black',     # label color
    palette='tab20',       # color palette
    figsize=(4, 3),        # figure size
)

To plot stacked bar showing composition percentages:

plot_composition(
    df=df,                # dataframe
    bin_col='bin',        # column for x-axis bins
    hue_col='category',   # column for stacked hue
    palette='tab20',      # color palette
    legend_title='Type',  # legend title
    rotate=45,            # x-tick rotation
    xlabel=None,          # x-axis label
    ylabel='Percentage',  # y-axis label
    figsize=(5, 3),       # figure size
)

To plot bar count with value labels on top:

plot_cnt(
    cnt=df['col'].value_counts(),  # value counts series
    xlabel=None,                   # x-axis label
    ylabel='Count',                # y-axis label
    figsize=(6, 3),                # figure size
)

Setup

Utils

---

set_sns

def set_sns(
    dpi:int=300
):

Set seaborn resolution for notebook display

---

save_svg

def save_svg(
    path
):

---

save_pdf

def save_pdf(
    path
):

---

save_show

def save_show(
    path:NoneType=None, # image path, e.g., img.svg, if not None, will save, else plt.show()
    show_only:bool=False
):

Show plot or save path

Color/palette

---

get_color_dict

def get_color_dict(
    categories, # list of names to assign color
    palette:str='tab20', # choose from sns.color_palette
):

Assign colors to a list of names (allow duplicates), returns a dictionary of unique name with corresponding color

info = Data.get_kinase_info()

info.modi_group.value_counts().index

Index(['TK', 'CAMK', 'Other', 'CMGC', 'AGC', 'STE', 'TKL', 'Atypical', 'CK1',
       'NEK', 'RGC'],
      dtype='object', name='modi_group')

info.pspa_category_big.str.split('_').str[0].value_counts()[:20].index

Index(['Basophilic', 'Pro-directed', 'Acidophilic', 'Map3k', 'Map4k',
       'Alpha/mlk', 'Fgf and vegf receptors', 'Assorted', 'Ripk/wnk', 'Pkc',
       'Ephrin receptors', 'Eif2ak/tlk', 'Nek/ask', 'Pdgf receptors', 'Src',
       'Jak', 'Tam receptors', 'Cmgc', 'Tec', 'Ulk/ttbk'],
      dtype='object', name='pspa_category_big')

---

get_subfamily_color

def get_subfamily_color(
    
):

# get_subfamily_color()

Sometimes plot functions do not take dict of colors but list of colors:

---

get_plt_color

def get_plt_color(
    palette, # dict, list, or set name (tab10)
    columns, # columns in the df for plot
):

Given a dict, list or set name, return the list of names; if dict, need to provide column names of the df.

get_plt_color('Set2',['a','b'])

---

get_hue_big

def get_hue_big(
    df, hue_col, # column of hue
    cnt_thr:int=10, # higher or equal to this threshold will be considered
):

Get part of hue according to its value counts; applied when the groups are too many.

When the categories are too many to show on the plot, we can filter and only show those big categories:

hue_big=get_hue_big(info,'pspa_category_big',10)
hue_big

5      Acidophilic
6      Acidophilic
11      Basophilic
12      Basophilic
13      Basophilic
          ...     
516    Acidophilic
517    Acidophilic
519          Map4k
520          Map3k
521          Map3k
Name: pspa_category_big, Length: 294, dtype: object

# plot_2d(embed,hue=hue_pspa,hue_title='PSPA category',palette='tab20')

Dimensionality reduction

---

reduce_feature

def reduce_feature(
    df:DataFrame, method:str='pca', # dimensionality reduction method, accept both capital and lower case
    complexity:int=20, # None for PCA; perfplexity for TSNE, recommend: 30; n_neigbors for UMAP, recommend: 15
    n:int=2, # n_components
    load:str=None, # load a previous model, e.g. model.pkl
    save:str=None, # pkl file to be saved, e.g. pca_model.pkl
    seed:int=123, # seed for random_state
    kwargs:VAR_KEYWORD
):

Reduce the dimensionality given a dataframe of values

A very common way to reduce feature number is to use dimensionality reduction method. reduce_feature is a dimensionality reduction function that can apply three dimensionality reduction methods: PCA, UMAP, TSNE. The later two is non-linear transformation, and PCA is linear transformation. Therefore, for plotting purpose, it is good to use UMAP/TSNE, by setting n (n_components) to 2 for 2d plot; for featuring purpose, it is good to use PCA, and set n to values to a rational values, like 64, 128 etc.

# Load data
df=Data.get_aa_morgan()

# Use PCA to reduce dimension; reduce the number of features to 20
feat = reduce_feature(df,'pca',n=2)
feat.head()

	PCA1	PCA2
aa
A	-0.954819	-0.953761
C	-0.041108	0.205091
D	-0.087707	0.245634
E	-0.271097	1.023801
F	2.253264	-0.251185

2D scatter plot

---

plot_2d

def plot_2d(
    embedding_df:DataFrame, # a dataframe of values that is waited for dimensionality reduction
    hue:str=None, # colname of color
    palette:str='tab20', # color scheme, could be tab10 if less categories
    legend:bool=False, # whether or not add the legend on the side
    name_list:NoneType=None, # a list of names to annotate each dot in the plot
    s:int=20, # size of the dot
    legend_title:str=None, # new argument to override legend title
    kwargs:VAR_KEYWORD
):

Given a dataframe of values, plot it in 2D. The method could be ‘pca’, ‘tsne’, or ‘umap’.

plot_2d(feat)

/tmp/ipykernel_674/2710618909.py:17: UserWarning: Ignoring `palette` because no `hue` variable has been assigned.
  g = sns.relplot(

Dimensionality reduction + 2d plot

---

plot_cluster

def plot_cluster(
    df:DataFrame, # a dataframe of values that is waited for dimensionality reduction
    method:str='pca', # dimensionality reduction method, choose from pca, umap, and tsne
    hue:str=None, # colname of color
    complexity:int=30, # this argument does not affect pca but others; recommend 30 for tsne, 15 for umap
    palette:str='tab20', # color scheme, could be tab10 if less categories
    legend:bool=False, # whether or not add the legend on the side
    name_list:NoneType=None, # a list of names to annotate each dot in the plot
    seed:int=123, # seed for dimensionality reduction
    s:int=50, # size of the dot
    legend_title:str=None, # new argument to override legend title
    kwargs:VAR_KEYWORD
):

Given a dataframe of values, plot it in 2D. The method could be ‘pca’, ‘tsne’, or ‘umap’.

plot_cluster(df,method='pca',legend=True)

/tmp/ipykernel_674/39475780.py:22: UserWarning: Ignoring `palette` because no `hue` variable has been assigned.
  g = sns.relplot(

Bokeh interactive plot

---

plot_bokeh

def plot_bokeh(
    X:DataFrame, # a dataframe of two columns from dimensionality reduction
    idx, # pd.Series or list that indicates identities for searching box
    hue:None, # pd.Series or list that indicates category for each sample
    s:int=3, # dot size
    kwargs:VAR_KEYWORD
):

Make interactive 2D plot with a searching box and window of dot information when pointing

# PCA reduce dimension to 2
X = reduce_feature(df)

# get info
info=Data.get_aa_info()

#plot
plot_bokeh(X, 
           idx = info.Name,
           hue = info.Name,
           s=7, 
           smiles = info.SMILES)

Loading BokehJS ...

Rank

---

plot_rank

def plot_rank(
    sorted_df:DataFrame, # a sorted dataframe
    x:str, # column name for x axis
    y:str, # column name for y aixs
    n_hi:int=10, # if not None, show the head n names
    n_lo:int=10, # if not None, show the tail n names
    figsize:tuple=(10, 8), # figure size
    data:NoneType=None, hue:NoneType=None, size:NoneType=None, style:NoneType=None, palette:NoneType=None,
    hue_order:NoneType=None, hue_norm:NoneType=None, sizes:NoneType=None, size_order:NoneType=None,
    size_norm:NoneType=None, markers:bool=True, style_order:NoneType=None, legend:str='auto', ax:NoneType=None
):

Plot rank from a sorted dataframe

# load data
# df = Data.get_pspa_raw().set_index('kinase')
df = pd.read_csv('https://github.com/sky1ove/katlas_raw/raw/refs/heads/main/nbs/raw/pspa_st_raw.csv').set_index('kinase')


# get sorted dataframe
sorted_df = df.max(1).reset_index(name='values').sort_values('values')
sorted_df.head()

	kinase	values
68	CK1G2	189898.392
294	VRK2	4191709.640
8	ALPHAK3	4573611.730
249	PRPK	8495330.790
38	CAMLCK	9413689.600

plot_rank(sorted_df,x='kinase',y='values')
plt.xlabel('kinase');

Distribution histogram

---

plot_hist

def plot_hist(
    df:DataFrame, # a dataframe that contain values for plot
    x:str, # column name of values
    figsize:tuple=(6, 2), data:NoneType=None, y:NoneType=None, hue:NoneType=None,
    weights:NoneType=None, # Vector variables
    stat:str='count', bins:str='auto', binwidth:NoneType=None,
    binrange:NoneType=None, # Histogram computation parameters
    discrete:NoneType=None, cumulative:bool=False, common_bins:bool=True, common_norm:bool=True,
    multiple:str='layer', element:str='bars', fill:bool=True, shrink:int=1, # Histogram appearance parameters
    kde:bool=False, kde_kws:NoneType=None,
    line_kws:NoneType=None, # Histogram smoothing with a kernel density estimate
    thresh:int=0, pthresh:NoneType=None, pmax:NoneType=None, cbar:bool=False, cbar_ax:NoneType=None,
    cbar_kws:NoneType=None, # Bivariate histogram parameters
    palette:NoneType=None, hue_order:NoneType=None, hue_norm:NoneType=None,
    color:NoneType=None, # Hue mapping parameters
    log_scale:NoneType=None, legend:bool=True, ax:NoneType=None, # Axes information
):

# we can use the same df
sorted_df.head()

	kinase	values
68	CK1G2	189898.392
294	VRK2	4191709.640
8	ALPHAK3	4573611.730
249	PRPK	8495330.790
38	CAMLCK	9413689.600

plot_hist(sorted_df,'values')

Bar graph

---

plot_count

def plot_count(
    cnt, # from df['x'].value_counts()
    tick_spacing:float=None, # tick spacing for x axis
    palette:str='tab20'
):

Make bar plot from df[‘x’].value_counts()

cnt = info.Name.value_counts()

plot_count(cnt)

---

plot_bar

def plot_bar(
    df, value, # colname of value
    group, # colname of group
    title:NoneType=None, figsize:tuple=(12, 5), fontsize:int=14,
    dots:bool=True, # whether or not add dots in the graph
    rotation:int=90, ascending:bool=False, ymin:NoneType=None, data:NoneType=None, x:NoneType=None, y:NoneType=None,
    hue:NoneType=None, order:NoneType=None, hue_order:NoneType=None, estimator:str='mean', errorbar:tuple=('ci', 95),
    n_boot:int=1000, seed:NoneType=None, units:NoneType=None, weights:NoneType=None, orient:NoneType=None,
    color:NoneType=None, palette:NoneType=None, saturation:float=0.75, fill:bool=True, hue_norm:NoneType=None,
    width:float=0.8, dodge:str='auto', gap:int=0, log_scale:NoneType=None, native_scale:bool=False,
    formatter:NoneType=None, legend:str='auto', capsize:int=0, err_kws:NoneType=None, ci:Deprecated=<deprecated>,
    errcolor:Deprecated=<deprecated>, errwidth:Deprecated=<deprecated>, ax:NoneType=None
):

Plot bar graph from unstacked dataframe; need to indicate columns of values and categories

info['cat'] = (info.pKa1>2).astype(int)

# get an unstacked dataframe
plot_bar(info,value='MW',group='cat',palette='tab20')

---

plot_group_bar

def plot_group_bar(
    df, value_cols, # list of column names for values, the order depends on the first item
    group, # column name of group (e.g., 'kinase')
    figsize:tuple=(12, 5), order:NoneType=None, title:NoneType=None, fontsize:int=14, rotation:int=90,
    data:NoneType=None, x:NoneType=None, y:NoneType=None, hue:NoneType=None, hue_order:NoneType=None,
    estimator:str='mean', errorbar:tuple=('ci', 95), n_boot:int=1000, seed:NoneType=None, units:NoneType=None,
    weights:NoneType=None, orient:NoneType=None, color:NoneType=None, palette:NoneType=None, saturation:float=0.75,
    fill:bool=True, hue_norm:NoneType=None, width:float=0.8, dodge:str='auto', gap:int=0, log_scale:NoneType=None,
    native_scale:bool=False, formatter:NoneType=None, legend:str='auto', capsize:int=0, err_kws:NoneType=None,
    ci:Deprecated=<deprecated>, errcolor:Deprecated=<deprecated>, errwidth:Deprecated=<deprecated>, ax:NoneType=None
):

Plot grouped bar graph from dataframe.

plot_group_bar(info,['pKa1','pKb2'],'Name')

Stacked bar graph

---

plot_stacked

def plot_stacked(
    df, column, hue, figsize:tuple=(5, 4), xlabel:NoneType=None, ylabel:NoneType=None, add_value:bool=True,
    kwargs:VAR_KEYWORD
):

Violin plot

---

plot_violin

def plot_violin(
    data, value:str='value', group:str='variable', ylabel:NoneType=None, dots:bool=True, figsize:tuple=(5, 3),
    kwargs:VAR_KEYWORD
):

Plot violin plots (with optional strip dots) for long-form data.

# plot_violin(data,ylabel='Accuracy',figsize=(4,3))

---

add_stats

def add_stats(
    ax, data, value:str='value', group:str='variable', pairs:NoneType=None, test:str='t-test_ind', loc:str='inside',
    text_format:str='star', kwargs:VAR_KEYWORD
):

Box plot

---

plot_box

def plot_box(
    df, value, # colname of value
    group, # colname of group
    title:NoneType=None, figsize:tuple=(6, 3), fontsize:int=14, dots:bool=True, rotation:int=90, data:NoneType=None,
    x:NoneType=None, y:NoneType=None, hue:NoneType=None, order:NoneType=None, hue_order:NoneType=None,
    orient:NoneType=None, color:NoneType=None, palette:NoneType=None, saturation:float=0.75, fill:bool=True,
    dodge:str='auto', width:float=0.8, gap:int=0, whis:float=1.5, linecolor:str='auto', linewidth:NoneType=None,
    fliersize:NoneType=None, hue_norm:NoneType=None, native_scale:bool=False, log_scale:NoneType=None,
    formatter:NoneType=None, legend:str='auto', ax:NoneType=None
):

Plot box plot.

plot_box(info,value='MW',group='cat',palette='tab20')

Correlation

---

plot_rel

def plot_rel(
    df, # dataframe that contains data
    x, # x axis values, or colname of x axis
    y, # y axis values, or colname of y axis
    text_location:tuple=(0.8, 0.1), # relative coords in Axes (0–1)
    method:str='spearman', # correlation method: 'pearson' or 'spearman'
    index_list:NoneType=None, # list of indices to annotate
    hue:NoneType=None, reg_line:bool=True, data:NoneType=None, x_estimator:NoneType=None, x_bins:NoneType=None,
    x_ci:str='ci', scatter:bool=True, fit_reg:bool=True, ci:int=95, n_boot:int=1000, units:NoneType=None,
    seed:NoneType=None, order:int=1, logistic:bool=False, lowess:bool=False, robust:bool=False, logx:bool=False,
    x_partial:NoneType=None, y_partial:NoneType=None, truncate:bool=True, dropna:bool=True, x_jitter:NoneType=None,
    y_jitter:NoneType=None, label:NoneType=None, color:NoneType=None, marker:str='o', scatter_kws:NoneType=None,
    line_kws:NoneType=None, ax:NoneType=None
):

Given a dataframe and the name of two columns, plot the two columns’ correlation with either Pearson or Spearman. Annotate points if their index is in index_list.

norm = Data.get_pspa().iloc[:,:-6].T

norm.head()

kinase	AAK1	ACVR2A	ACVR2B	AKT1	AKT2	AKT3	ALK2	ALK4	ALPHAK3	AMPKA1	...	NTRK3	TXK	TYK2	TYRO3	FLT1	KDR	FLT4	WEE1_TYR	YES1	ZAP70
-5P	0.0720	0.0415	0.0533	0.0603	0.0602	0.0705	0.0536	0.0552	0.0571	0.0555	...	0.0656	0.0784	0.0606	0.0661	0.0642	0.0634	0.0457	0.0531	0.0677	0.0602
-5G	0.0245	0.0481	0.0517	0.0594	0.0617	0.0624	0.0659	0.0574	0.0478	0.0504	...	0.0751	0.0795	0.0674	0.0666	0.0688	0.0672	0.0531	0.0640	0.0571	0.0880
-5A	0.0284	0.0584	0.0566	0.0552	0.0643	0.0745	0.0662	0.0605	0.0253	0.0534	...	0.0594	0.0669	0.0561	0.0548	0.0597	0.0556	0.0488	0.0559	0.0537	0.0623
-5C	0.0456	0.0489	0.0772	0.0605	0.0582	0.0628	0.0762	0.0483	0.0384	0.0588	...	0.0517	0.0662	0.0521	0.0533	0.0618	0.0517	0.0553	0.0560	0.0530	0.0496
-5S	0.0425	0.0578	0.0533	0.0516	0.0534	0.0442	0.0567	0.0574	0.0571	0.0504	...	0.0452	0.0561	0.0506	0.0543	0.0614	0.0541	0.0512	0.0433	0.0527	0.0471

5 rows × 396 columns

plot_rel(df=norm, x='AAK1', y='BIKE')

Matrix heatmap

---

get_similarity

def get_similarity(
    df, metric:str='euclidean'
):

Calculate distance matrix of a df; also return inverse df (similarity df)

---

plot_corr

def plot_corr(
    df_corr, mask_method:str='upper', # "upper", "lower", or None
    inverse_color:bool=False, figsize:tuple=(15, 10), annot:bool=True, linewidths:float=0.1, kwargs:VAR_KEYWORD
):

Plot distance/similarity matrix

corr = info.head(8).corr(numeric_only=True)

plot_corr(corr)

AUCDF

We compute the area under the empirical cumulative distribution function (CDF) as a function of kinase rank using the trapezoidal rule.
Let $ r_{(1)} < r_{(2)} < < r_{(n)} $ be the sorted rank values (e.g., \(1,2,\dots,n\)), and define the empirical CDF values as:

\[ F(r_{(i)}) = \frac{i}{n} \]

The normalized area under this CDF-vs-rank curve (AUCDF) is then computed via the trapezoidal rule:

\[ \text{AUC}_{\text{CDF}} = \frac{1}{r_{\max} - r_{\min}} \sum_{i=1}^{n-1} \frac{F(r_{(i)}) + F(r_{(i+1)})}{2} \cdot (r_{(i+1)} - r_{(i)}) \]

where $ r_{} = r_{(1)} $, typically 1; $ r_{} = r_{(n)} $, typically \(n\).

This measures how quickly the cumulative mass increases across the ranked kinases. If better kinases (lower rank) tend to appear earlier in the CDF, the AUCDF will be higher.

---

get_AUCDF

def get_AUCDF(
    df, col, reverse:bool=False, plot:bool=True, xlabel:str='Rank of reported kinase'
):

Plot CDF curve and get relative area under the curve

get_AUCDF(sorted_df,'values')

/tmp/ipykernel_674/2030375725.py:15: DeprecationWarning: `trapz` is deprecated. Use `trapezoid` instead, or one of the numerical integration functions in `scipy.integrate`.
  area_under_curve = np.trapz(y_values, x_values)

np.float64(0.872608302329996)

Confusion matrix

---

plot_confusion_matrix

def plot_confusion_matrix(
    target, # pd.Series
    pred, # pd.Series
    class_names:list=['0', '1'], normalize:bool=False, title:str='Confusion matrix',
    cmap:LinearSegmentedColormap=<matplotlib.colors.LinearSegmentedColormap object at 0x7fdce51fb380>
):

Plot the confusion matrix.

target = info.MW<160
pred = info.pKa1>2.1

plot_confusion_matrix(target,pred,normalize=True)

Normalized confusion matrix

Pie plot

---

plot_pie

def plot_pie(
    value_counts, # value counts
    hue_order:NoneType=None, # list of strings
    labeldistance:float=0.8, fontsize:int=12, fontcolor:str='black', palette:str='tab20', figsize:tuple=(4, 3)
):

info=Data.get_kinase_info()

plot_pie(info.modi_group.value_counts(),
         fontsize=10,
         labeldistance=1)

Bar composition

---

calculate_pct

def calculate_pct(
    df, bin_col, hue_col
):

Get percentage for hue in each bin; with hue adding up to 1 in each bin.

---

plot_composition

def plot_composition(
    df, bin_col, hue_col, palette:str='tab20', legend_title:NoneType=None, rotate:int=45, xlabel:NoneType=None,
    ylabel:str='Percentage', figsize:tuple=(5, 3)
):

Plot bar count with value text

---

plot_cnt

def plot_cnt(
    cnt, xlabel:NoneType=None, ylabel:str='Count', figsize:tuple=(6, 3)
):