Triomics analysis

Introduction

Installation

import pandas as pd
import numpy as np 
import umap
import os
import scanpy as sc

from scipy.stats.mstats import gmean

import matplotlib.pyplot as plt
from matplotlib import style, rcParams
from matplotlib.patches import Patch
import seaborn as sns

Helper functions

def umap_feature_plot(
    data, 
    feature, 
    hide_ticks = True, 
    legend_loc = 'top right', 
    title = 'UMAP plot', 
    point_size = 20,
    random_state = 42, 
    palette = 'flare', 
    n_neighbors = 15, 
    min_dist = 0.5, 
    spread = 1.0, 
    alpha = 1.0, 
    init_pos = 'spectral',
    n_epochs = 500,
    negative_sample_rate = 5.0):
    """
    Function to display UMAP reductions using seaborn

    Parameters:

    data : pandas.DataFrame
        Data use for dim reduction
    labels : pandas.Series
        Labels for coloring the plot
    hide_ticks : boolean
        Should axis ticks be hidden? Default is True
    legend_loc : str
        Position of the legend
    """
    
    embedding = umap.UMAP(
        random_state = random_state,
        min_dist = min_dist,
        spread = spread,
        n_neighbors = n_neighbors,
        learning_rate = alpha,
        negative_sample_rate = negative_sample_rate,
        init = init_pos,
        metric = 'euclidean',
        n_epochs = n_epochs
    )
    
    embedding = pd.DataFrame(embedding.fit_transform(data), columns = ['UMAP1','UMAP2'])

    labels = data[feature]
        
    plot = sns.scatterplot(
        x = 'UMAP1',
        y = 'UMAP2',
        data = embedding,
        hue = labels.tolist(),
        alpha = .8,
        linewidth = 0.1,
        s = point_size,
        palette = palette,
        edgecolor = 'grey'    
    )

    plot.set_title(title)

    if(hide_ticks):
        plot.axes.xaxis.set_ticks([])
        plot.axes.yaxis.set_ticks([])
        plot.tick_params(left=False,bottom=False)

    plot.legend(loc = legend_loc, fontsize = 8)

    out_file = "figures/" + title + ".png"
    plt.savefig(out_file, dpi = 600, transparent = False, facecolor = plot.get_facecolor())

def single_cell_umap(data, labels, hide_ticks = True, legend_loc = 'top right', title = 'UMAP plot', random_state = 42,point_size = 10, palette = 'deep'):
    """
    Function to display UMAP reductions using seaborn

    Parameters:

    data : pandas.DataFrame
        Data use for dim reduction
    labels : pandas.Series
        Labels for coloring the plot
    hide_ticks : boolean
        Should axis ticks be hidden? Default is True
    legend_loc : str
        Position of the legend
    """
    embedding = umap.UMAP(random_state = random_state)
    embedding = pd.DataFrame(embedding.fit_transform(data), columns = ['UMAP1','UMAP2'])
    
    plot = sns.scatterplot(
        x = 'UMAP1',
        y = 'UMAP2',
        data = embedding,
        hue = labels.tolist(),
        alpha = .8,
        linewidth = 0,
        s = point_size,
        palette = palette
    )

    plot.set_title(title)

    if(hide_ticks):
        plot.axes.xaxis.set_ticks([])
        plot.axes.yaxis.set_ticks([])
        plot.tick_params(left=False,bottom=False)

    plot.legend(loc = legend_loc, fontsize = 8)

    out_file = "figures/" + title + ".png"
    plt.savefig(out_file, dpi = 600, transparent = False, facecolor = plot.get_facecolor())


def clr(data):
    """
    Compute clr normalization using the gmean function from scipy

    data : data to transform
    """
    return np.log(data+1) - np.log(gmean(data)+1)