AAclust: Selecting redundancy-reduced scale sets

The Amino Acid clustering (AAclust) class is k-optimized clustering wrapper for selecting redundancy-reduced sets of numerical scales, introduced in [Breimann24a].

We load an example scale dataset to showcase it:

import aaanalysis as aa
aa.options["verbose"] = False

# Create test dataset of 25 amino acid scales
df_scales = aa.load_scales()
X = df_scales.T

AAclust can utilize any clustering model that uses the n_clusters parameter:

from sklearn.cluster import KMeans

# AAclust with KMens (default)
aac = aa.AAclust(model_class=KMeans)

By fitting AAclust, its three-step algorithm is performed to select an optimized n_clusters (k). The three steps involve (1) an estimation of lower bound of k, (2) refinement of k, and (3) an optional clustering merging. Various results are saved as attributes:

# Fit clustering model (KMeans by default)
aac = aa.AAclust()
aac.fit(X)
# Get output parameters
n_clusters = aac.n_clusters
print("n_clusters: ", n_clusters)

n_clusters:  48

Instead of optimizing the number of clusters, we can pre-defined it using the n_clusters parameter:

# Fit clustering model with pre-selected k
labels = aac.fit(X, n_clusters=5).labels_

We can obtain visualize the clustering results and the obtained clustering centers using the respective plotting AAclustPlot class. All data points are visualized in the PCA plot including the cluster centers highlighted by an ‘x’:

import matplotlib.pyplot as plt

aac_plot = aa.AAclustPlot()
aa.plot_settings()
ax, df_components = aac_plot.centers(X, labels=labels)
plt.tight_layout()
plt.show()

To obtain redundancy-reduced scale sets, AAclust selects one medoid per cluster, which is the scale closest to center of the respective cluster. These can be highlighted using the AAclustPlot.medoids method

aac_plot = aa.AAclustPlot()
aa.plot_settings()
ax, df_components = aac_plot.medoids(X, labels=labels)
plt.tight_layout()
plt.show()

For further details, see our Feature Engineering API, AAontology Usage Principels, and AAclust Usage Principels.