【ADHD血液診断】レプリケートを考慮して次元削減

import time

import pickle

import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

import seaborn as sns

from scipy.stats import pearsonr

from sklearn.decomposition import PCA

from sklearn.manifold import TSNE

import umap

from scipy.sparse.csgraph import connected_components

import warnings

warnings.filterwarnings('ignore')

df_std = pd.read_csv("src/df_std.csv", index_col=0)

names = list(set([s.split("_rep")[0] for s in df_sample["Sample Name"].tolist()]))

data = []

for n in names:

    idxes = []

    for i, s in enumerate(df_sample["Sample Name"].tolist()):

        if n in s:

            idxes.append(i)

    # print(df_sample.iloc[idxes, 0])

    v = df_std.iloc[idxes, :].apply(lambda x: np.mean(x)).tolist()

    y = 0 if "CTRL" in n else 1

    v.append(y)

    data.append(v)

df_rep = pd.DataFrame(data, index=names, columns=df_std.columns.tolist()+["y"])

df_rep.to_csv("src/df_rep.csv")

df_rep.shape

(76, 16043)

x = df_rep.iloc[:, :-1]

x_scaled = (x - x.mean()) / x.std()

pca = PCA()

pca.fit(x_scaled)

score = pd.DataFrame(pca.transform(x_scaled), index=x_scaled.index)

plt.scatter(score.iloc[:, 0], score.iloc[:, 1], c=df_xy.iloc[:, -1], cmap=plt.get_cmap('jet'))

clb = plt.colorbar()

clb.set_label('ADHD Status', labelpad=-20, y=1.1, rotation=0)

plt.xlabel('PC1')

plt.ylabel('PC2')

plt.show()

# 寄与率

contribution_ratios = pd.DataFrame(pca.explained_variance_ratio_)

cumulative_contribution_ratios = contribution_ratios.cumsum()

cont_cumcont_ratios = pd.concat([contribution_ratios, cumulative_contribution_ratios], axis=1).T

cont_cumcont_ratios.index = ['contribution_ratio', 'cumulative_contribution_ratio']  # 行の名前を変更

# 寄与率を棒グラフで、累積寄与率を線で入れたプロット図を重ねて描画

x_axis = range(1, contribution_ratios.shape[0] + 1)  # 1 から成分数までの整数が x 軸の値

plt.rcParams['font.size'] = 18

plt.bar(x_axis, contribution_ratios.iloc[:, 0], align='center')  # 寄与率の棒グラフ

plt.plot(x_axis, cumulative_contribution_ratios.iloc[:, 0], 'r.-')  # 累積寄与率の線を入れたプロット図

plt.xlabel('Number of principal components')  # 横軸の名前

plt.ylabel('Contribution ratio(blue),\nCumulative contribution ratio(red)')  # 縦軸の名前。\n で改行しています

plt.show()

tsne = TSNE(n_components=2, random_state=42)

corrdinates = pd.DataFrame(tsne.fit_transform(x_scaled))

plt.scatter(corrdinates.iloc[:, 0], corrdinates.iloc[:, 1], c=df_xy.iloc[:, -1], cmap=plt.get_cmap('jet'))

clb = plt.colorbar()

clb.set_label('ADHD Status', labelpad=-20, y=1.1, rotation=0)

plt.xlabel('tSNE 1')

plt.ylabel('tSNE 2')

plt.show()

embedding = pd.DataFrame(umap.UMAP().fit_transform(x_scaled))

plt.scatter(embedding.iloc[:, 0], embedding.iloc[:, 1], c=df_xy.iloc[:, -1], cmap=plt.get_cmap('jet'))

clb = plt.colorbar()

clb.set_label('ADHD Status', labelpad=-20, y=1.1, rotation=0)

plt.xlabel('UMAP 1')

plt.ylabel('UMAP 2')

plt.show()