Merge pull request #96 from aertslab/render_plot

Functional rework small fixes

src/crested/pl/_utils.py

@@ -90,16 +90,14 @@ def render_plot(
             label.set_fontsize(y_tick_fontsize)
             label.set_rotation(y_label_rotation)
     if tight_rect:
-        plt.tight_layout(rect=tight_rect)
+        fig.tight_layout(rect=tight_rect)
     else:
-        plt.tight_layout()
+        fig.tight_layout()
     if save_path:
         plt.savefig(save_path)
 
     if show:
         plt.show()
-    else:
-        plt.close(fig)
 
     if not show and not save_path:
         return fig

src/crested/tl/_crested.py

@@ -225,6 +225,8 @@ def load_model(self, model_path: os.PathLike, compile: bool = True) -> None:
             the model weights (e.g. when finetuning a model). If False, you should
             provide a TaskConfig to the Crested object before calling fit.
         """
+        if compile and self.config is not None:
+            logger.warning("Loading a model with compile=True. The CREsted config object will be ignored.")
         self.model = keras.models.load_model(model_path, compile=compile)
 
     def fit(

src/crested/utils/_utils.py

@@ -14,116 +14,7 @@
 
 from crested._genome import Genome, _resolve_genome
 from crested._io import _extract_tracks_from_bigwig
-
-
-def get_hot_encoding_table(
-    alphabet: str = "ACGT",
-    neutral_alphabet: str = "N",
-    neutral_value: float = 0.0,
-    dtype=np.float32,
-) -> np.ndarray:
-    """Get hot encoding table to encode a DNA sequence to a numpy array with shape (len(sequence), len(alphabet)) using bytes."""
-
-    def str_to_uint8(string) -> np.ndarray:
-        """Convert string to byte representation."""
-        return np.frombuffer(string.encode("ascii"), dtype=np.uint8)
-
-    # 256 x 4
-    hot_encoding_table = np.zeros(
-        (np.iinfo(np.uint8).max + 1, len(alphabet)), dtype=dtype
-    )
-
-    # For each ASCII value of the nucleotides used in the alphabet
-    # (upper and lower case), set 1 in the correct column.
-    hot_encoding_table[str_to_uint8(alphabet.upper())] = np.eye(
-        len(alphabet), dtype=dtype
-    )
-    hot_encoding_table[str_to_uint8(alphabet.lower())] = np.eye(
-        len(alphabet), dtype=dtype
-    )
-
-    # For each ASCII value of the nucleotides used in the neutral alphabet
-    # (upper and lower case), set neutral_value in the correct column.
-    hot_encoding_table[str_to_uint8(neutral_alphabet.upper())] = neutral_value
-    hot_encoding_table[str_to_uint8(neutral_alphabet.lower())] = neutral_value
-
-    return hot_encoding_table
-
-
-HOT_ENCODING_TABLE = get_hot_encoding_table()
-
-
-def one_hot_encode_sequence(sequence: str, expand_dim: bool = True) -> np.ndarray:
-    """
-    One hot encode a DNA sequence.
-
-    Will return a numpy array with shape (1, len(sequence), 4) if expand_dim is True, otherwise (len(sequence),4).
-    Alphabet is ACGT.
-
-    Parameters
-    ----------
-    sequence
-        The DNA sequence to one hot encode.
-    expand_dim
-        Whether to expand the dimensions of the output array.
-
-    Returns
-    -------
-    The one hot encoded DNA sequence.
-    """
-    if expand_dim:
-        return np.expand_dims(
-            HOT_ENCODING_TABLE[np.frombuffer(sequence.encode("ascii"), dtype=np.uint8)],
-            axis=0,
-        )
-    else:
-        return HOT_ENCODING_TABLE[
-            np.frombuffer(sequence.encode("ascii"), dtype=np.uint8)
-        ]
-
-
-def generate_mutagenesis(x, include_original=True, flanks=(0, 0)):
-    """Generate all possible single point mutations in a sequence."""
-    _, L, A = x.shape
-    start, end = 0, L
-    x_mut = []
-    start = flanks[0]
-    end = L - flanks[1]
-    for length in range(start, end):
-        for a in range(A):
-            if not include_original:
-                if x[0, length, a] == 1:
-                    continue
-            x_new = np.copy(x)
-            x_new[0, length, :] = 0
-            x_new[0, length, a] = 1
-            x_mut.append(x_new)
-    return np.concatenate(x_mut, axis=0)
-
-
-def generate_motif_insertions(x, motif, flanks=(0, 0), masked_locations=None):
-    """Generate motif insertions in a sequence."""
-    _, L, A = x.shape
-    start, end = 0, L
-    x_mut = []
-    motif_length = motif.shape[1]
-    start = flanks[0]
-    end = L - flanks[1] - motif_length + 1
-    insertion_locations = []
-
-    for motif_start in range(start, end):
-        motif_end = motif_start + motif_length
-        if masked_locations is not None:
-            if np.any(
-                (motif_start <= masked_locations) & (masked_locations < motif_end)
-            ):
-                continue
-        x_new = np.copy(x)
-        x_new[0, motif_start:motif_end, :] = motif
-        x_mut.append(x_new)
-        insertion_locations.append(motif_start)
-
-    return np.concatenate(x_mut, axis=0), insertion_locations
+from crested.utils._seq_utils import one_hot_encode_sequence
 
 
 class EnhancerOptimizer: