figure s7

ddmc/figures/common.py

@@ -204,6 +204,8 @@ def get_pvals_across_clusters(
     label: pd.Series | np.ndarray[bool], centers: pd.DataFrame | np.ndarray
 ) -> np.ndarray[float]:
     pvals = []
+    if isinstance(centers, pd.DataFrame):
+        centers = centers.values
     centers_pos = centers[label]
     centers_neg = centers[~label]
     for i in range(centers.shape[1]):
@@ -212,14 +214,15 @@ def get_pvals_across_clusters(
 
 
 def plot_p_signal_across_clusters_and_binary_feature(
-    label: pd.Series | np.ndarray[bool],
+    feature: pd.Series | np.ndarray[bool],
     centers: pd.DataFrame | np.ndarray,
     label_name: str,
     ax,
 ) -> None:
     centers = centers.copy()
-    centers[label_name] = label
-    df_violin = centers.reset_index().melt(
+    centers_labeled = centers.copy()
+    centers_labeled[label_name] = feature
+    df_violin = centers_labeled.reset_index().melt(
         id_vars=label_name,
         value_vars=centers.columns,
         value_name="p-signal",
@@ -236,7 +239,7 @@ def plot_p_signal_across_clusters_and_binary_feature(
     )
     ax.legend(prop={"size": 8})
     annotation_height = df_violin["p-signal"].max() + 0.02
-    for i, pval in enumerate(get_pvals_across_clusters(label, centers)):
+    for i, pval in enumerate(get_pvals_across_clusters(feature, centers)):
         if pval < 0.05:
             annotation = "*"
         elif pval < 0.01:

ddmc/figures/figureMS7.py

@@ -0,0 +1,58 @@
+import numpy as np
+from sklearn.linear_model import LogisticRegressionCV
+from sklearn.preprocessing import StandardScaler
+
+from ddmc.clustering import DDMC
+from ddmc.datasets import CPTAC, select_peptide_subset
+from ddmc.figures.common import (
+    plot_cluster_kinase_distances,
+    plot_p_signal_across_clusters_and_binary_feature,
+    getSetup,
+)
+from ddmc.logistic_regression import plot_roc, plot_cluster_regression_coefficients
+
+
+def makeFigure():
+    axes, f = getSetup((11, 7), (2, 3), multz={0: 1})
+    cptac = CPTAC()
+    p_signal = select_peptide_subset(cptac.get_p_signal(), keep_num=50)
+    stk11m = cptac.get_mutations(["STK11.mutation.status"])["STK11.mutation.status"]
+
+    model = DDMC(n_components=30, seq_weight=100).fit(p_signal)
+    centers = model.transform(as_df=True)
+    centers = centers.loc[stk11m.index]
+    plot_p_signal_across_clusters_and_binary_feature(
+        stk11m.values, centers, "STK11M", axes[0]
+    )
+
+    centers.iloc[:, :] = StandardScaler(with_std=False).fit_transform(
+        centers.iloc[:, :]
+    )
+
+    # Logistic Regression
+    lr = LogisticRegressionCV(
+        cv=5,
+        solver="saga",
+        max_iter=10000,
+        n_jobs=-1,
+        penalty="l1",
+        class_weight="balanced",
+        random_state=10,
+    )
+
+    plot_roc(
+        lr, centers.values, stk11m.values, cv_folds=4, title="ROC STK11", ax=axes[1]
+    )
+
+    plot_cluster_regression_coefficients(
+        axes[2],
+        lr,
+    )
+
+    top_clusters = np.argsort(np.abs(lr.coef_.squeeze()))[-3:]
+    plot_cluster_kinase_distances(
+        model.predict_upstream_kinases()[top_clusters],
+        model.get_pssms(PsP_background=True, clusters=top_clusters)[0],
+        ax=axes[3],
+    )
+    return f

ddmc/logistic_regression.py

@@ -11,7 +11,7 @@
 from sklearn.metrics import RocCurveDisplay
 from sklearn.model_selection import StratifiedKFold, RepeatedKFold
 
-def plot_cluster_regression_coefficients(ax: Axes, lr, hue=None, xlabels=False, title=False):
+def plot_cluster_regression_coefficients(ax: Axes, lr, hue=None, title=False):
     """Plot LR coeficients of clusters."""
     coefs_ = pd.DataFrame(lr.coef_.T, columns=["LR Coefficient"])
     if hue:
@@ -20,7 +20,7 @@ def plot_cluster_regression_coefficients(ax: Axes, lr, hue=None, xlabels=False,
         hue = "Sample"
     else:
         coefs_["Cluster"] = np.arange(coefs_.shape[0])
-    if xlabels:
+    if xlabels is not None:
         coefs_["Cluster"] = xlabels
     p = sns.barplot(
         ax=ax,