Fix figure build failures (M2, M3, M4, MS7)

ddmc/__init__.py

@@ -1,3 +1,20 @@
 """This is the __init__.py file."""
 
+import sklearn.utils
+
 __version__ = "0.0.1"
+
+# fancyimpute (unmaintained since 2020) calls check_array with the
+# `force_all_finite` kwarg, which scikit-learn renamed to `ensure_all_finite`
+# and later removed. Patch it here, at package-import time, so it is in place
+# before any submodule (however it orders its own imports) pulls in fancyimpute.
+_sklearn_check_array = sklearn.utils.check_array
+
+
+def _check_array_compat(X, **kwargs):
+    if "force_all_finite" in kwargs:
+        kwargs["ensure_all_finite"] = kwargs.pop("force_all_finite")
+    return _sklearn_check_array(X, **kwargs)
+
+
+sklearn.utils.check_array = _check_array_compat  # ty: ignore[invalid-assignment]

ddmc/clustering.py

@@ -6,31 +6,14 @@
 
 import numpy as np
 import pandas as pd
-import sklearn.utils
+from fancyimpute import SoftImpute
 from sklearn.mixture import GaussianMixture
 from sklearn.utils.validation import check_is_fitted
 
 from .binomial import AAlist, BackgroundSeqs, Binomial, frequencies
 from .motifs import compute_control_pssm, get_pspls
 from .pam250 import PAM250
 
-# fancyimpute (unmaintained since 2020) calls check_array with the
-# `force_all_finite` kwarg, which scikit-learn renamed to `ensure_all_finite`
-# and later removed. Patch it before fancyimpute's submodules import
-# check_array into their own namespaces, rather than forking fancyimpute.
-_sklearn_check_array = sklearn.utils.check_array
-
-
-def _check_array_compat(X, **kwargs):
-    if "force_all_finite" in kwargs:
-        kwargs["ensure_all_finite"] = kwargs.pop("force_all_finite")
-    return _sklearn_check_array(X, **kwargs)
-
-
-sklearn.utils.check_array = _check_array_compat  # ty: ignore[invalid-assignment]
-
-from fancyimpute import SoftImpute  # noqa: E402
-
 
 class DDMC(GaussianMixture):
     """Cluster peptides by both sequence similarity and condition-wise phosphorylation following an

ddmc/figures/common.py

@@ -191,7 +191,7 @@ def plot_cluster_kinase_distances(
             KinToPhosphotypeDict[kin] for kin in distances_pssm["Kinase"]
         ]
         try:
-            most_frequent_phosphoacceptor = AAlist[pssms[i, 5].argmax()]
+            most_frequent_phosphoacceptor = AAlist[int(np.argmax(pssms[i, :, 5]))]
         except Exception:
             most_frequent_phosphoacceptor = "S/T"
         if most_frequent_phosphoacceptor == "S" or most_frequent_phosphoacceptor == "T":

ddmc/figures/figureM3.py

@@ -84,7 +84,7 @@ def plot_fig_3abd(ax_a, ax_b, ax_d):
     # Plot kinase predictions for cluster 16
     plot_cluster_kinase_distances(
         model.predict_upstream_kinases()[[16]],
-        model.get_pssms(PsP_background=True, clusters=[16])[0],
+        model.get_pssms(PsP_background=True, clusters=[16]),
         ax=ax_d,
     )
 

ddmc/figures/figureMS7.py

@@ -54,7 +54,7 @@ def makeFigure():
 
     plot_cluster_kinase_distances(
         model.predict_upstream_kinases()[top_clusters],
-        model.get_pssms(PsP_background=True, clusters=top_clusters)[0],
+        model.get_pssms(PsP_background=True, clusters=top_clusters),
         ax=axes[3],
     )
     return f

ddmc/logistic_regression.py

@@ -62,6 +62,8 @@ def plot_roc(
     ax: Axes | None = None,
 ):
     """Plot Receiver Operating Characteristc with cross-validation folds of a given classifier model."""
+    X = np.asarray(X)
+    y = np.asarray(y)
     if kfold == "Stratified":
         cv = StratifiedKFold(n_splits=cv_folds)
     elif kfold == "Repeated":