added solution to additoinal exercise ch01

python_notebooks/solutions/solution_01_sc.py

@@ -0,0 +1,77 @@
+df = pd.read_csv('../data/Patel.csv',sep=',', header=0)
+
+df = df.set_index(df.iloc[:,0], inplace=False)
+df = df.iloc[:,1:]
+
+X = df.T
+
+print(X.shape)
+
+patients = list(map(lambda s: s[0:5],df.columns)) #take first 5 letter for patient id
+
+color_dict={'MGH26':'blue', 'MGH28':'orange', 'MGH29': 'red', 'MGH30': 'green', 'MGH31': 'pink'}
+colors=[color_dict[p] for p in patients]
+
+
+plt.figure(figsize=(10,10))
+ax = sns.heatmap(X, yticklabels=False)
+plt.ylabel("Cell type")
+plt.xlabel("Gene")
+plt.show(block=False)
+
+pca = PCA() #create a PCA object
+
+pca.fit(X)
+x_pca = pca.transform(X)
+
+var_explained=pca.explained_variance_ratio_
+
+plt.figure(figsize=(10,10))
+plt.scatter(x_pca[:,0],x_pca[:,1],c=colors)
+plt.xlabel('First Principal Component ({0:.2f}%)'.format(var_explained[0]*100))
+plt.ylabel('Second Principal Component ({0:.2f}%)'.format(var_explained[1]*100))
+
+tsne=TSNE(n_components=2,perplexity=5).fit(X)#create the T-SNE object and fit the data
+X_embedded = tsne.embedding_#project the data to the new manifold using the fitted function found before
+
+plt.figure(figsize=(10,10))
+plt.scatter(X_embedded[:, 0], X_embedded[:, 1], c=colors, s=60, lw=0)
+plt.title('KL divergence {0:.2f}\n perplexity= {1}'.format(tsne.kl_divergence_,5),fontsize=12)
+plt.xlabel('First Dimension')
+plt.ylabel('Second Dimension')
+
+dbscan = cluster.DBSCAN(eps=20, min_samples=5)
+dbscan.fit_predict(X_embedded[:, 0:2])
+
+plt.figure(figsize=(10,10))
+idx = np.where(dbscan.labels_>=0)[0]
+plt.scatter(X_embedded[idx, 0], X_embedded[idx, 1], c=[colors[i] for i in idx], s=60, lw=0)
+plt.title('KL divergence {0:.2f}\n perplexity= {1}'.format(tsne.kl_divergence_,5),fontsize=12)
+plt.xlabel('First Dimension')
+plt.ylabel('Second Dimension')
+
+
+X = X.T
+
+print(X.shape)
+
+kmeans = cluster.KMeans(5)
+kmeans.fit(X)
+cl_labels = kmeans.labels_
+
+print(Counter(cl_labels))
+
+tsne=TSNE(n_components=2,perplexity=15).fit(X)#create the T-SNE object and fit the data
+X_embedded = tsne.embedding_#project the data to the new manifold using the fitted function found before
+
+dbscan = cluster.DBSCAN(eps=2, min_samples=5)
+dbscan.fit_predict(X_embedded[:, 0:2])
+
+plt.figure(figsize=(10,10))
+idx = np.where(dbscan.labels_>=0)[0]
+plt.scatter(X_embedded[idx, 0], X_embedded[idx, 1], c=[cl_labels[i]  for i in idx], s=60, lw=0, cmap='plasma')
+plt.xlabel('First Dimension')
+plt.ylabel('Second Dimension')
+
+plt.show()
+