DeConveil 0.1.1__py3-none-any.whl → 0.1.3__py3-none-any.whl

deconveil/__version__.py

@@ -0,0 +1 @@
+__version__ = "0.1.3"

deconveil/dds.py

@@ -17,10 +17,10 @@ from deconveil.utils_fit import fit_rough_dispersions
 from deconveil.utils_fit import fit_moments_dispersions2
 from deconveil.utils_fit import grid_fit_beta
 from deconveil.utils_fit import irls_glm
+from deconveil.utils_fit import build_design_matrix
 
 from pydeseq2.preprocessing import deseq2_norm_fit
 from pydeseq2.preprocessing import deseq2_norm_transform
-from pydeseq2.utils import build_design_matrix
 from pydeseq2.utils import dispersion_trend
 from pydeseq2.utils import mean_absolute_deviation
 from pydeseq2.utils import n_or_more_replicates

deconveil/utils_fit.py

@@ -523,3 +523,151 @@ def nbinomFn(
     ).sum(0)
 
     return prior - nll
+
+
+def build_design_matrix(
+    metadata: pd.DataFrame,
+    design_factors: Union[str, List[str]] = "condition",
+    ref_level: Optional[List[str]] = None,
+    continuous_factors: Optional[List[str]] = None,
+    expanded: bool = False,
+    intercept: bool = True,
+) -> pd.DataFrame:
+    """Build design_matrix matrix for DEA.
+
+    Unless specified, the reference factor is chosen alphabetically.
+
+    Parameters
+    ----------
+    metadata : pandas.DataFrame
+        DataFrame containing metadata information.
+        Must be indexed by sample barcodes.
+
+    design_factors : str or list
+        Name of the columns of metadata to be used as design_matrix variables.
+        (default: ``"condition"``).
+
+    ref_level : dict or None
+        An optional list of two strings of the form ``["factor", "ref_level"]``
+        specifying the factor of interest and the desired reference level, e.g.
+        ``["condition", "A"]``. (default: ``None``).
+
+    continuous_factors : list or None
+        An optional list of continuous (as opposed to categorical) factors. Any factor
+        not in ``continuous_factors`` will be considered categorical (default: ``None``).
+
+    expanded : bool
+        If true, use one column per category. Else, use n-1 columns, for each n-level
+        categorical factor.
+        (default: ``False``).
+
+    intercept : bool
+        If true, add an intercept (a column containing only ones). (default: ``True``).
+
+    Returns
+    -------
+    pandas.DataFrame
+        A DataFrame with experiment design information (to split cohorts).
+        Indexed by sample barcodes.
+    """
+    if isinstance(
+        design_factors, str
+    ):  # if there is a single factor, convert to singleton list
+        design_factors = [design_factors]
+
+    for factor in design_factors:
+        # Check that each factor has at least 2 levels
+        if len(np.unique(metadata[factor])) < 2:
+            raise ValueError(
+                f"Factors should take at least two values, but {factor} "
+                f"takes the single value '{np.unique(metadata[factor])}'."
+            )
+
+    # Check that level factors in the design don't contain underscores. If so, convert
+    # them to hyphens
+    warning_issued = False
+    for factor in design_factors:
+        if np.any(["_" in value for value in metadata[factor]]):
+            if not warning_issued:
+                warnings.warn(
+                    """Some factor levels in the design contain underscores ('_').
+                    They will be converted to hyphens ('-').""",
+                    UserWarning,
+                    stacklevel=2,
+                )
+                warning_issued = True
+            metadata[factor] = metadata[factor].apply(lambda x: x.replace("_", "-"))
+
+    if continuous_factors is not None:
+        categorical_factors = [
+            factor for factor in design_factors if factor not in continuous_factors
+        ]
+    else:
+        categorical_factors = design_factors
+
+    # Check that there is at least one categorical factor
+    if len(categorical_factors) > 0:
+        design_matrix = pd.get_dummies(
+            metadata[categorical_factors], drop_first=not expanded
+        )
+
+        if ref_level is not None:
+            if len(ref_level) != 2:
+                raise KeyError("The reference level should contain 2 strings.")
+            if ref_level[1] not in metadata[ref_level[0]].values:
+                raise KeyError(
+                    f"The metadata data should contain a '{ref_level[0]}' column"
+                    f" with a '{ref_level[1]}' level."
+                )
+
+            # Check that the reference level is not in the matrix (if unexpanded design)
+            ref_level_name = "_".join(ref_level)
+            if (not expanded) and ref_level_name in design_matrix.columns:
+                # Remove the reference level and add one
+                factor_cols = [
+                    col for col in design_matrix.columns if col.startswith(ref_level[0])
+                ]
+                missing_level = next(
+                    level
+                    for level in np.unique(metadata[ref_level[0]])
+                    if f"{ref_level[0]}_{level}" not in design_matrix.columns
+                )
+                design_matrix[f"{ref_level[0]}_{missing_level}"] = 1 - design_matrix[
+                    factor_cols
+                ].sum(1)
+                design_matrix.drop(ref_level_name, axis="columns", inplace=True)
+
+        if not expanded:
+            # Add reference level as column name suffix
+            for factor in design_factors:
+                if ref_level is None or factor != ref_level[0]:
+                    # The reference is the unique level that is no longer there
+                    ref = next(
+                        level
+                        for level in np.unique(metadata[factor])
+                        if f"{factor}_{level}" not in design_matrix.columns
+                    )
+                else:
+                    # The reference level is given as an argument
+                    ref = ref_level[1]
+                design_matrix.columns = [
+                    f"{col}_vs_{ref}" if col.startswith(factor) else col
+                    for col in design_matrix.columns
+                ]
+    else:
+        # There is no categorical factor in the design
+        design_matrix = pd.DataFrame(index=metadata.index)
+
+    if intercept:
+        design_matrix.insert(0, "intercept", 1)
+
+    # Convert categorical factors one-hot encodings to int
+    design_matrix = design_matrix.astype("int")
+
+    # Add continuous factors
+    if continuous_factors is not None:
+        for factor in continuous_factors:
+            # This factor should be numeric
+            design_matrix[factor] = pd.to_numeric(metadata[factor])
+    return design_matrix
+

deconveil/utils_processing.py

@@ -5,11 +5,93 @@ from pathlib import Path
 
 import numpy as np
 import pandas as pd
+import deconveil
 
 from typing import List, Literal, Optional, Dict, Any, cast
 
 
 
+def load_test_data(
+    modality: Literal["rna", "cnv", "metadata", "cnv_tumor"] = "rna",
+    dataset: Literal["tcga_brca"] = "tcga_brca",
+    debug: bool = False,
+    debug_seed: int = 42,
+) -> pd.DataFrame:
+    """Load TCGA-BRCA example data from the DeConveil package.
+
+    Parameters
+    ----------
+    modality : {"rna", "cnv", "metadata", "cnv_tumor"}
+        Type of data to load.
+
+    dataset : {"tcga_brca"}
+        Dataset name. Only "tcga_brca" is currently supported.
+
+    debug : bool, optional
+        If True, randomly subsample 10 samples and 100 features (if applicable).
+        Default is False.
+
+    debug_seed : int, optional
+        Random seed for reproducibility of debug subsampling. Default is 42.
+
+    Returns
+    -------
+    pandas.DataFrame
+        The requested data modality as a DataFrame.
+    """
+    assert modality in ["rna", "cnv", "metadata", "cnv_tumor"], (
+        "modality must be one of: 'rna', 'cnv', 'metadata', 'cnv_tumor'"
+    )
+    assert dataset in ["tcga_brca"], (
+        "dataset must be one of: 'tcga_brca'"
+    )
+
+    # Locate data within the package
+    datasets_path = Path(__file__).resolve().parent.parent / "datasets" / dataset
+
+    # Construct file paths
+    file_map = {
+        "rna": datasets_path / "rna.csv",
+        "cnv": datasets_path / "cnv.csv",
+        "metadata": datasets_path / "metadata.csv",
+        "cnv_tumor": datasets_path / "cnv_tumor.csv",
+    }
+
+    data_path = file_map[modality]
+    if not data_path.exists():
+        raise FileNotFoundError(f"Data file not found: {data_path}")
+
+    # Load the CSV
+    df = pd.read_csv(data_path, index_col=0)
+
+    # Apply debug mode subsampling
+    if debug:
+        df = df.sample(n=min(10, df.shape[0]), random_state=debug_seed)
+        if modality in ["rna", "cnv"]:
+            df = df.sample(n=min(100, df.shape[1]), axis=1, random_state=debug_seed)
+
+    return df
+
+
+def replace_underscores(factors: List[str]):
+    """Replace all underscores from strings in a list by hyphens.
+
+    To be used on design factors to avoid bugs due to the reliance on
+    ``str.split("_")`` in parts of the code.
+
+    Parameters
+    ----------
+    factors : list
+        A list of strings which may contain underscores.
+
+    Returns
+    -------
+    list
+        A list of strings in which underscores were replaced by hyphens.
+    """
+    return [factor.replace("_", "-") for factor in factors]
+
+
 def filter_low_count_genes(
     df: pd.DataFrame,
     other_dfs: Optional[List[pd.DataFrame]] = None,

{deconveil-0.1.1.dist-info → deconveil-0.1.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: DeConveil
-Version: 0.1.1
+Version: 0.1.3
 Summary: An extension of PyDESeq2/DESeq2 designed to account for genome aneuploidy
 Home-page: https://github.com/caravagnalab/DeConveil
 Author: Katsiaryna Davydzenka

{deconveil-0.1.1.dist-info → deconveil-0.1.3.dist-info}/RECORD

@@ -1,15 +1,16 @@
 deconveil/__init__.py,sha256=_6FL_AYiycv9nP3mKJiQ4zl4aU83YSWnV2YoIZr9Mv0,188
-deconveil/dds.py,sha256=ccYi6o6c6yw_5AXWdpbs0MIALp1xLsUvtp-QhEzH3cQ,49034
+deconveil/__version__.py,sha256=XEqb2aiIn8fzGE68Mph4ck1FtQqsR_am0wRWvrYPffQ,22
+deconveil/dds.py,sha256=0MNwtDzCjqjoJR-rrCmVu3JOaDd3gXuToOzTBXJMxak,49039
 deconveil/default_inference.py,sha256=J40O0-qZChLnLrLGmhwxjaTVsV7REWAUQOTf8qSwWk0,9466
 deconveil/ds.py,sha256=Vb9p152U1KXltrXFpMoBxY6YRW25dP4CO26_osbz6Aw,29476
 deconveil/grid_search.py,sha256=iOHR8ur10MyrrfEZHr409lGulGxODufsjG6j7lQ7tWs,5181
 deconveil/inference.py,sha256=B3zf3q_mbCTX3gHJwuXnTuy9uyXOxEjuWyaSR6VtVEo,10429
 deconveil/utils_clustering.py,sha256=twspPvXQ6pvw_NaY1ebyvswuH3ZvVBGn7DeOpZ1XatI,5939
-deconveil/utils_fit.py,sha256=ODtIwFKKKchQBiwdNhPSOCt5wsPcpLHCKRENf6JmF18,15785
+deconveil/utils_fit.py,sha256=SdGcBQjN3cyzbSFessufYOOOJAQCOjNcy3etbwmodsM,21583
 deconveil/utils_plot.py,sha256=1JQthYXaEUKUWa0fy8owkyJ1CTkQxlrSRAqPkXMk7Us,9857
-deconveil/utils_processing.py,sha256=CB99CwQst7eUiIgE58yl7_3E6uD9CgQoU_Qmprjyt-s,4141
-deconveil-0.1.1.dist-info/licenses/LICENSE,sha256=BJ0f3JRteiF7tjiARi8syxiu4yKmckc0nWlHCKXttKQ,1078
-deconveil-0.1.1.dist-info/METADATA,sha256=zuxmrMBmMhAz8bcoDdMZhZcs38oomC-mNhue4hsiVY4,1097
-deconveil-0.1.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-deconveil-0.1.1.dist-info/top_level.txt,sha256=yAWZbw0eg8XpbMsswoq-VzBGfQHrfWOqNHnu2qQ2xO4,10
-deconveil-0.1.1.dist-info/RECORD,,
+deconveil/utils_processing.py,sha256=9j35FAfQ7oNjdH1FWHP90DBTyL5RwlgdVbbW9de10VI,6560
+deconveil-0.1.3.dist-info/licenses/LICENSE,sha256=BJ0f3JRteiF7tjiARi8syxiu4yKmckc0nWlHCKXttKQ,1078
+deconveil-0.1.3.dist-info/METADATA,sha256=yL6AwQ5ziGhrI5lE4FDCHOadT81W6yEIDWGsHni6Q5w,1097
+deconveil-0.1.3.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+deconveil-0.1.3.dist-info/top_level.txt,sha256=yAWZbw0eg8XpbMsswoq-VzBGfQHrfWOqNHnu2qQ2xO4,10
+deconveil-0.1.3.dist-info/RECORD,,