# ==============================================
# 1. Import libraries
# ==============================================
import pandas as pd
import numpy as np
from kmodes.kmodes import KModes
from sklearn.metrics import silhouette_score
from scipy.spatial.distance import pdist, squareform
from scipy.cluster.hierarchy import linkage, fcluster
from sklearn.preprocessing import LabelEncoder
# ==============================================
# 2. Generate Dummy Genetic Mutation Data
# ==============================================
np.random.seed(42)
n_samples = 200
genes = [‘BRCA1’, ‘BRCA2’, ‘TP53’, ‘EGFR’, ‘KRAS’]
mutations = [‘missense’, ‘nonsense’, ‘frameshift’, ‘silent’]
chromosomes = [‘chr1’, ‘chr7’, ‘chr12’, ‘chr17’]
zygosity = [‘heterozygous’, ‘homozygous’]
data = {
‘Gene’: np.random.choice(genes, n_samples),
‘Mutation_Type’: np.random.choice(mutations, n_samples),
‘Chromosome’: np.random.choice(chromosomes, n_samples),
‘Zygosity’: np.random.choice(zygosity, n_samples),
‘Variant_Presence’: np.random.choice([‘present’, ‘absent’], n_samples)
}
df = pd.DataFrame(data)
print(“Sample Data:”)
print(df.head())
# ==============================================
# 3. Encode categorical data numerically
# ==============================================
df_encoded = df.apply(LabelEncoder().fit_transform)
# ==============================================
# 4. Function for silhouette using Hamming distance
# ==============================================
def categorical_silhouette_score(X, labels):
dist_matrix = pdist(X, metric=’hamming’)
return silhouette_score(squareform(dist_matrix), labels, metric=’precomputed’)
# ==============================================
# 5. K-Modes Clustering
# ==============================================
km = KModes(n_clusters=3, init=’Huang’, n_init=5, verbose=0, random_state=42)
km_labels = km.fit_predict(df)
km_silhouette = categorical_silhouette_score(df_encoded, km_labels)
# Add cluster membership to dataframe
df[‘KModes_Cluster’] = km_labels
print(f”\nK-Modes Silhouette Score: {km_silhouette:.3f}”)
# ==============================================
# 6. Hierarchical Clustering (Hamming distance)
# ==============================================
dist_matrix = pdist(df_encoded, metric=’hamming’)
linkage_matrix = linkage(dist_matrix, method=’average’)
hier_labels = fcluster(linkage_matrix, t=3, criterion=’maxclust’)
hier_silhouette = categorical_silhouette_score(df_encoded, hier_labels)
# Add cluster membership
df[‘Hierarchical_Cluster’] = hier_labels
print(f”Hierarchical Clustering Silhouette Score: {hier_silhouette:.3f}”)
# ==============================================
# 7. Latent Class Analysis (proxy via K-Modes Cao init)
# ==============================================
km_lca = KModes(n_clusters=3, init=’Cao’, n_init=10, verbose=0, random_state=42)
lca_labels = km_lca.fit_predict(df)
lca_silhouette = categorical_silhouette_score(df_encoded, lca_labels)
# Add cluster membership
df[‘LCA_Cluster’] = lca_labels
print(f”LCA (proxy via K-Modes Cao init) Silhouette Score: {lca_silhouette:.3f}”)
# ==============================================
# 8. Compare and display results
# ==============================================
results = pd.DataFrame({
‘Method’: [‘K-Modes (Huang)’, ‘Hierarchical (Hamming)’, ‘LCA (Cao init)’],
‘Silhouette_Score’: [km_silhouette, hier_silhouette, lca_silhouette]
}).sort_values(by=’Silhouette_Score’, ascending=False)
print(“\n=== Cluster Validation Results ===”)
print(results)
best_method = results.iloc[0][‘Method’]
print(f”\n🏆 Best clustering method: {best_method}”)
# ==============================================
# 9. Display dataframe with cluster memberships
# ==============================================
print(“\nSample data with cluster memberships:”)
print(df.head())
