Add cleveland plot

chemap/fingerprint_computation.py

@@ -339,7 +339,12 @@ def _rdkit_unfolded(
 
     if cfg.count:
         out: UnfoldedCount = []
-        for s, mol in zip(smiles, mols):
+        for s, mol in tqdm(
+                zip(smiles, mols),
+                disable=not show_progress,
+                desc="Compute fingerprints",
+                total=len(mols)
+                ):
             if mol is None:
                 _handle_invalid(cfg.invalid_policy, s)
                 if cfg.invalid_policy == "keep":
@@ -356,7 +361,12 @@ def _rdkit_unfolded(
         return out
 
     out: UnfoldedBinary = []
-    for s, mol in zip(smiles, mols):
+    for s, mol in tqdm(
+            zip(smiles, mols),
+            disable=not show_progress,
+            desc="Compute fingerprints",
+            total=len(mols)
+            ):
         if mol is None:
             _handle_invalid(cfg.invalid_policy, s)
             if cfg.invalid_policy == "keep":
@@ -386,7 +396,12 @@ def _rdkit_folded_dense(
     n_features: Optional[int] = None
     pending_invalid: List[int] = []  # indices in `rows` that need backfill after we learn D
 
-    for s, mol in zip(smiles, mols):
+    for s, mol in tqdm(
+            zip(smiles, mols),
+            disable=not show_progress,
+            desc="Compute fingerprints",
+            total=len(mols)
+            ):
         if mol is None:
             _handle_invalid(cfg.invalid_policy, s)
             if cfg.invalid_policy == "keep":
@@ -450,7 +465,12 @@ def _rdkit_folded_csr(
     if cfg.folded_weights is not None:
         w = np.asarray(cfg.folded_weights, dtype=np.float32).ravel()
 
-    for s, mol in zip(smiles, mols):
+    for s, mol in tqdm(
+            zip(smiles, mols),
+            disable=not show_progress,
+            desc="Compute fingerprints",
+            total=len(mols)
+            ):
         if mol is None:
             _handle_invalid(cfg.invalid_policy, s)
 

chemap/metrics.py

@@ -219,55 +219,55 @@ def tanimoto_similarity_matrix_dense(references: np.ndarray, queries: np.ndarray
 # This is O(R*Q*avg_nnz_merge) and can be expensive for large R,Q.
 # For huge datasets prefer ANN (PyNNDescent/UMAP) with `tanimoto_distance_sparse`.
 @numba.njit(parallel=True, fastmath=True, cache=True)
-def tanimoto_similarity_matrix_sparse_binary(references, queries) -> np.ndarray:
+def tanimoto_similarity_matrix_sparse_binary(fingerprints_1, fingerprints_2) -> np.ndarray:
     """
     Pairwise Tanimoto similarity between two sets of unfolded or sparse binary fingerprints.
 
     Parameters
     ----------
-    references
+    fingerprints_1
         List of 1D numpy arrays of sorted bit indices (unique).
-    queries
+    fingerprints_2
         List of 1D numpy arrays of sorted bit indices (unique).
     """
-    R = len(references)
-    Q = len(queries)
+    R = len(fingerprints_1)
+    Q = len(fingerprints_2)
     out = np.empty((R, Q), dtype=np.float32)
     for i in numba.prange(R):
         for j in range(Q):
-            out[i, j] = tanimoto_similarity_sparse_binary(references[i], queries[j])
+            out[i, j] = tanimoto_similarity_sparse_binary(fingerprints_1[i], fingerprints_2[j])
     return out
 
 
 @numba.njit(parallel=True, fastmath=True, cache=True)
 def tanimoto_similarity_matrix_sparse(
-        references_bits,
-        references_vals,
-        queries_bits,
-        queries_vals
+        fingerprints_1_bits,
+        fingerprints_1_vals,
+        fingerprints_2_bits,
+        fingerprints_2_vals
         ) -> np.ndarray:
     """
     Pairwise generalized Tanimoto similarity between two sets of unfolded count/weight fingerprints.
 
     Parameters
     ----------
-    references_bits
+    fingerprints_1_bits
         List of 1D numpy arrays of sorted bit indices (unique) for reference fingerprints.
-    references_vals
+    fingerprints_1_vals
         List of 1D numpy arrays of counts/weights for reference fingerprints.
-    queries_bits
+    fingerprints_2_bits
         List of 1D numpy arrays of sorted bit indices (unique) for query fingerprints.
-    queries_vals
+    fingerprints_2_vals
         List of 1D numpy arrays of counts/weights for query fingerprints.
     """
-    R = len(references_bits)
-    Q = len(queries_bits)
+    R = len(fingerprints_1_bits)
+    Q = len(fingerprints_2_bits)
     out = np.empty((R, Q), dtype=np.float32)
     for i in numba.prange(R):
         for j in range(Q):
             out[i, j] = tanimoto_similarity_sparse(
-                references_bits[i], references_vals[i],
-                queries_bits[j], queries_vals[j],
+                fingerprints_1_bits[i], fingerprints_1_vals[i],
+                fingerprints_2_bits[j], fingerprints_2_vals[j],
             )
     return out
 

chemap/plotting/__init__.py

@@ -1,4 +1,5 @@
 from .chem_space_umap import create_chem_space_umap, create_chem_space_umap_gpu
+from .cleveland import ClevelandStyle, cleveland_dotplot
 from .colormap_handling import (
     LabelMapConfig,
     PaletteConfig,
@@ -19,12 +20,14 @@
 
 
 __all__ = [
+    "ClevelandStyle",
     "LabelMapConfig",
     "PaletteConfig",
     "PresentPairsConfig",
     "build_hier_label_map",
     "build_selected_label_column",
     "build_selected_palette",
+    "cleveland_dotplot",
     "create_chem_space_umap",
     "create_chem_space_umap_gpu",
     "make_hier_palette",

chemap/plotting/chem_space_umap.py

@@ -163,9 +163,9 @@ def create_chem_space_umap_gpu(
     fpgen: Optional[Any] = None,
     fingerprint_config: Optional[FingerprintConfig] = None,
     show_progress: bool = True,
-    log_count: bool = True,
+    log_count: bool = False,
     # UMAP (GPU / cuML)
-    n_neighbors: int = 15,
+    n_neighbors: int = 100,
     min_dist: float = 0.25,
 ) -> pd.DataFrame:
     """Compute fingerprints and create 2D UMAP coordinates using cuML (GPU).
@@ -221,9 +221,6 @@ def create_chem_space_umap_gpu(
         show_progress=show_progress,
     )
 
-    # Convert to sparse array
-    # fps_csr = fingerprints_to_csr(fingerprints).X
-
     # Reduce memory footprint (works well for count fingerprints)
     if not log_count:
         # stays integer-like