microarray 0.1.0__py3-none-any.whl

microarray/tools/_toptable.py

@@ -0,0 +1,360 @@
+"""Extract and format top differentially expressed genes.
+
+This module provides functions for extracting results from empirical Bayes
+moderated differential expression analysis, with multiple testing correction
+and flexible sorting/filtering options.
+"""
+
+import warnings
+from typing import Literal
+
+import numpy as np
+import pandas as pd
+from anndata import AnnData
+from scipy import stats
+from statsmodels.stats.multitest import multipletests
+
+
+def top_table(
+    data: AnnData | dict,
+    group: str | None = None,
+    coef: int | None = None,
+    reference: int | None = None,
+    number: int = 10,
+    adjust_method: str = "fdr_bh",
+    sort_by: Literal["logfc", "avg_expr", "t", "p_value", "b", "none"] = "b",
+    p_value: float = 1.0,
+    lfc: float = 0.0,
+) -> pd.DataFrame:
+    """Extract top differentially expressed genes from a moderated fit.
+
+    Creates a summary table of the top genes ranked by evidence of differential
+    expression, with multiple testing correction and flexible filtering/sorting.
+    This is the final step in limma-style differential expression analysis.
+
+    Parameters
+    ----------
+    fit : dict
+        Linear model fit object from ebayes(), containing:
+        - coefficients: coefficient estimates
+        - t: moderated t-statistics
+        - p_value: raw p-values
+        - lods: B-statistics (log-odds)
+        - genes: gene identifiers
+        - adata: AnnData object for annotation retrieval
+    coef : int, optional
+        Which coefficient/contrast to extract. If None, uses the first coefficient
+        (index 0). Coefficients correspond to columns in the design matrix.
+    reference : int, optional
+        Reference coefficient index for a pairwise contrast. When provided,
+        statistics are computed for (coef - reference). This is useful for
+        no-intercept group-mean designs where columns represent groups.
+    number : int, default=10
+        Maximum number of genes to return.
+    adjust_method : str, default="fdr_bh"
+        Multiple testing correction method. Options:
+        - "bonferroni": Bonferroni correction
+        - "sidak": Sidak correction
+        - "holm-sidak": Holm-Sidak correction
+        - "holm": Holm correction
+        - "simes-hochberg": Simes-Hochberg correction
+        - "hommel": Hommel correction
+        - "fdr_bh": Benjamini-Hochberg FDR (default)
+        - "fdr_by": Benjamini-Yekutieli FDR
+        - "fdr_tsbh": Two-stage Benjamini-Hochberg
+        - "fdr_tsbky": Two-stage Benjamini-Krieger-Yekutieli
+    sort_by : {"logfc", "avg_expr", "t", "p_value", "b", "none"}, default="b"
+        Column to sort results by:
+        - "logfc": Log fold-change (descending absolute value)
+        - "avg_expr": Average expression level (descending)
+        - "t": Moderated t-statistic (descending absolute value)
+        - "p_value": P-value (ascending)
+        - "b": B-statistic/log-odds (descending)
+        - "none": No sorting (gene order as in original data)
+    p_value : float, default=1.0
+        Filter genes by adjusted p-value threshold. Only genes with
+        p_adj ≤ p_value are returned.
+    lfc : float, default=0.0
+        Filter genes by log-fold-change threshold. Only genes with
+        |logfc| ≥ lfc are returned.
+
+    Returns:
+    -------
+    pd.DataFrame
+        Results table with the following columns:
+        - Gene identifiers (index)
+        - logfc: Log2 fold-change for the selected coefficient
+        - avg_expr: Average expression across all samples
+        - t: Moderated t-statistic
+        - p_value: Raw p-value
+        - p_adj: Adjusted p-value (multiple testing corrected)
+        - b: B-statistic (log-odds of differential expression)
+
+        Additional columns from adata.var are also included if available.
+
+    Notes:
+    -----
+    **Multiple Testing Correction:**
+
+    When testing thousands of genes simultaneously, multiple testing correction
+    is essential to control false discovery rate (FDR) or family-wise error rate
+    (FWER). The default "fdr_bh" (Benjamini-Hochberg) controls FDR, meaning that
+    among genes called significant at level α, we expect α proportion to be false
+    positives.
+
+    **Sorting and Filtering:**
+
+    The function applies filtering (p_value and lfc thresholds) before selecting
+    the top 'number' genes. Sorting determines which genes appear in the top N.
+
+    **B-statistic:**
+
+    The B-statistic (log-odds) is often the best ranking criterion because it
+    accounts for both effect size and statistical significance. Higher B values
+    indicate stronger evidence of differential expression.
+
+    **Typical Usage:**
+
+    1. Fit linear models: fit = lm_fit(adata, design)
+    2. Moderate variances: fit = ebayes(fit)
+    3. Extract top genes: results = top_table(fit, coef=1, number=100)
+
+    Examples:
+    --------
+    >>> import microarray as ma
+    >>> import numpy as np
+    >>> # After lm_fit and ebayes
+    >>> fit = ma.tl.ebayes(fit)
+    >>> # Get top 20 genes for coefficient 1 (e.g., treatment effect)
+    >>> results = ma.tl.top_table(fit, coef=1, number=20)
+    >>> print(results)
+    >>> # Filter by adjusted p-value and fold-change
+    >>> sig_genes = ma.tl.top_table(
+    ...     fit,
+    ...     coef=1,
+    ...     number=1000,
+    ...     p_value=0.05,  # FDR < 5%
+    ...     lfc=1.0,  # |log2FC| > 1 (i.e., >2-fold change)
+    ...     sort_by="logfc",
+    ... )
+    >>> print(f"Found {len(sig_genes)} significant genes")
+    >>> # Get all genes, sorted by p-value
+    >>> all_results = ma.tl.top_table(
+    ...     fit,
+    ...     coef=1,
+    ...     number=np.inf,  # Get all genes
+    ...     sort_by="p_value",
+    ... )
+    >>> # Create volcano plot
+    >>> import matplotlib.pyplot as plt
+    >>> plt.scatter(all_results["logfc"], -np.log10(all_results["p_value"]))
+    >>> plt.xlabel("Log2 Fold Change")
+    >>> plt.ylabel("-log10(P-value)")
+    >>> plt.title("Volcano Plot")
+
+    References:
+    ----------
+    Smyth, G. K. (2004). Linear models and empirical Bayes methods for assessing
+    differential expression in microarray experiments. Statistical Applications
+    in Genetics and Molecular Biology, 3(1).
+
+    Benjamini, Y., and Hochberg, Y. (1995). Controlling the false discovery rate:
+    a practical and powerful approach to multiple testing. Journal of the Royal
+    Statistical Society B, 57, 289-300.
+
+    See Also:
+    --------
+    lm_fit : Fit linear models
+    ebayes : Apply empirical Bayes moderation
+    """
+    if isinstance(data, AnnData):
+        adata = data
+        fit = adata.uns.get("lm_fit")
+        if fit is None:
+            raise ValueError("No fit object found in adata.uns['lm_fit']. Run lm_fit and ebayes first.")
+    else:
+        fit = data
+        adata = fit.get("adata") if isinstance(fit, dict) else None
+
+    # Validate that ebayes has been run
+    required_keys = ["t", "p_value", "coefficients", "genes"]
+    for key in required_keys:
+        if key not in fit:
+            raise ValueError(f"Fit object missing '{key}'. Did you run ebayes() before top_table()?")
+
+    # Select coefficient.
+    # Limma's topTable avoids using intercept-only tests by default when
+    # multiple coefficients are present.
+    n_coef = fit["coefficients"].shape[1]
+    design = fit.get("design")
+
+    if group is not None:
+        group_to_column = fit.get("group_to_column")
+        design_columns = fit.get("design_columns")
+        if group_to_column is None or design_columns is None:
+            raise ValueError("Fit object does not contain group metadata. Re-run lm_fit with groupby.")
+
+        if hasattr(group_to_column, "__len__") and len(group_to_column) == 0:
+            raise ValueError("Fit object does not contain group metadata. Re-run lm_fit with groupby.")
+
+        if hasattr(design_columns, "__len__") and len(design_columns) == 0:
+            raise ValueError("Fit object does not contain group metadata. Re-run lm_fit with groupby.")
+
+        if not isinstance(group_to_column, dict):
+            raise ValueError("Fit object contains invalid 'group_to_column' metadata. Re-run lm_fit with groupby.")
+
+        if not isinstance(design_columns, list):
+            design_columns = list(design_columns)
+
+        if group not in group_to_column:
+            valid_groups = ", ".join(str(x) for x in group_to_column.keys())
+            raise ValueError(f"Unknown group '{group}'. Valid groups: {valid_groups}")
+        coef = int(design_columns.index(group_to_column[group]))
+    elif coef is not None:
+        warnings.warn("'coef' is deprecated. Use 'group' instead.", DeprecationWarning, stacklevel=2)
+
+    is_binary_one_hot = False
+    if design is not None and design.ndim == 2 and design.shape[1] == n_coef:
+        is_binary = np.all(np.isclose(design, 0.0) | np.isclose(design, 1.0))
+        row_sums = np.sum(design, axis=1)
+        one_hot_rows = np.all(np.isclose(row_sums, 1.0))
+        nonempty_groups = np.all(np.sum(design, axis=0) > 0)
+        is_binary_one_hot = bool(is_binary and one_hot_rows and nonempty_groups)
+
+    if coef is None:
+        if is_binary_one_hot and n_coef == 2:
+            # Common limma design (~0 + group): default to group2 - group1.
+            coef = 1
+            reference = 0
+        elif n_coef == 1:
+            coef = 0
+        else:
+            coef = 0
+            if design is not None and design.ndim == 2 and design.shape[1] == n_coef:
+                intercept_cols = np.where(np.all(np.isclose(design, 1.0), axis=0))[0]
+                candidates = [i for i in range(n_coef) if i not in set(intercept_cols)]
+                if candidates:
+                    coef = candidates[0]
+
+    if coef < 0 or coef >= n_coef:
+        raise ValueError(f"Invalid coefficient index {coef}. Valid range: 0 to {fit['coefficients'].shape[1] - 1}")
+
+    # For two-group one-hot designs (no intercept), treat coef as group-vs-group
+    # by default, matching expected DE behavior without an explicit contrasts step.
+    if reference is None and is_binary_one_hot and n_coef == 2:
+        reference = 1 - coef
+
+    if reference is not None:
+        if reference < 0 or reference >= n_coef:
+            raise ValueError(
+                f"Invalid reference index {reference}. Valid range: 0 to {fit['coefficients'].shape[1] - 1}"
+            )
+        if reference == coef:
+            raise ValueError("reference must be different from coef")
+
+    # Extract statistics for selected coefficient
+    # n_genes = len(fit["genes"])
+    if reference is None:
+        logFC = fit["coefficients"][:, coef]
+        t_stat = fit["t"][:, coef]
+        p_val = fit["p_value"][:, coef]
+        B_stat = fit["lods"][:, coef]
+    else:
+        cvec = np.zeros(n_coef, dtype=float)
+        cvec[coef] = 1.0
+        cvec[reference] = -1.0
+
+        logFC = fit["coefficients"] @ cvec
+
+        cov_coef = fit.get("cov_coefficients")
+        if cov_coef is None:
+            raise ValueError("Fit object missing 'cov_coefficients'.")
+        c_unscaled = float(np.sqrt(cvec @ cov_coef @ cvec))
+
+        se = c_unscaled * np.sqrt(fit["s2_post"])
+        with np.errstate(divide="ignore", invalid="ignore"):
+            t_stat = logFC / se
+        p_val = 2 * stats.t.sf(np.abs(t_stat), fit["df_total"])
+
+        var_prior = fit.get("var_prior")
+        if var_prior is None:
+            B_stat = np.full_like(logFC, np.nan, dtype=float)
+        else:
+            var_prior_contrast = float(np.sum((cvec**2) * np.asarray(var_prior)))
+            with np.errstate(divide="ignore", invalid="ignore"):
+                r = (c_unscaled**2 + var_prior_contrast) / (c_unscaled**2)
+                t2 = t_stat**2
+                df_prior = fit.get("df_prior", np.inf)
+                if np.isfinite(df_prior) and df_prior <= 1e6:
+                    kernel = (1.0 + fit["df_total"]) / 2.0 * np.log((t2 + fit["df_total"]) / (t2 / r + fit["df_total"]))
+                else:
+                    kernel = t2 * (1.0 - 1.0 / r) / 2.0
+                proportion = fit.get("proportion", 0.01)
+                B_stat = np.log(proportion / (1.0 - proportion)) - np.log(r) / 2.0 + kernel
+
+    # Compute average expression across samples
+    # adata.X is samples × genes, so mean over axis 0
+    if adata is None:
+        raise ValueError("AnnData object is required to compute AveExpr. Pass AnnData input to top_table.")
+    ave_expr = np.mean(adata.X, axis=0)
+    if ave_expr.ndim > 1:
+        ave_expr = ave_expr.squeeze()
+
+    # Apply multiple testing correction
+    # Remove NaN p-values
+    valid_pvals = ~np.isnan(p_val)
+    adj_p_val = np.full_like(p_val, np.nan)
+
+    if np.sum(valid_pvals) > 0:
+        _, adj_p_val[valid_pvals], _, _ = multipletests(
+            p_val[valid_pvals],
+            method=adjust_method,
+            is_sorted=False,
+            returnsorted=False,
+        )
+
+    # Create results DataFrame
+    results = pd.DataFrame(
+        {
+            "logfc": logFC,
+            "avg_expr": ave_expr,
+            "t": t_stat,
+            "p_value": p_val,
+            "p_adj": adj_p_val,
+            "b": B_stat,
+        },
+        index=fit["genes"],
+    )
+
+    # Add gene annotations from adata.var if available
+    if adata.var.shape[1] > 0:
+        # Join with var DataFrame (excluding index which is already used)
+        for col in adata.var.columns:
+            if col not in results.columns:
+                results[col] = adata.var[col].values
+
+    # Apply filters
+    mask = (results["p_adj"] <= p_value) & (np.abs(results["logfc"]) >= lfc)
+    results = results[mask]
+
+    # Sort results
+    if sort_by == "logfc":
+        results = results.iloc[np.argsort(-np.abs(results["logfc"].values))]
+    elif sort_by == "avg_expr":
+        results = results.sort_values("avg_expr", ascending=False)
+    elif sort_by == "t":
+        results = results.iloc[np.argsort(-np.abs(results["t"].values))]
+    elif sort_by == "p_value":
+        results = results.sort_values("p_value", ascending=True)
+    elif sort_by == "b":
+        results = results.sort_values("b", ascending=False)
+    elif sort_by == "none":
+        pass  # No sorting
+    else:
+        raise ValueError(f"Invalid sort_by: {sort_by}. Use 'logfc', 'avg_expr', 't', 'p_value', 'b', or 'none'.")
+
+    # Limit to top N genes
+    if number < len(results):
+        results = results.iloc[:number]
+
+    return results

microarray-0.1.0.dist-info/METADATA

@@ -0,0 +1,75 @@
+Metadata-Version: 2.4
+Name: microarray
+Version: 0.1.0
+Summary: Microarray analysis tools
+Author: harryhaller001
+Author-email: harryhaller001 <harryhaller001@gmail.com>
+License-Expression: MIT
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: Natural Language :: English
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3 :: Only
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
+Classifier: Typing :: Typed
+Requires-Dist: adjusttext>=1.3
+Requires-Dist: anndata
+Requires-Dist: click
+Requires-Dist: matplotlib
+Requires-Dist: requests
+Requires-Dist: scikit-learn
+Requires-Dist: scipy
+Requires-Dist: statsmodels
+Requires-Dist: ipython ; extra == 'docs'
+Requires-Dist: myst-parser ; extra == 'docs'
+Requires-Dist: nbsphinx ; extra == 'docs'
+Requires-Dist: sphinx ; extra == 'docs'
+Requires-Dist: sphinx-autoapi ; extra == 'docs'
+Requires-Dist: sphinx-autodoc-typehints ; extra == 'docs'
+Requires-Dist: sphinx-book-theme ; extra == 'docs'
+Requires-Dist: decoupler>=2.1.4 ; extra == 'test'
+Requires-Dist: ipykernel ; extra == 'test'
+Requires-Dist: ipython ; extra == 'test'
+Requires-Dist: ipywidgets ; extra == 'test'
+Requires-Dist: pre-commit ; extra == 'test'
+Requires-Dist: pytest ; extra == 'test'
+Requires-Dist: pytest-cov ; extra == 'test'
+Requires-Dist: responses ; extra == 'test'
+Requires-Dist: ruff ; extra == 'test'
+Requires-Dist: scanpy>=1.11.5 ; extra == 'test'
+Requires-Dist: tqdm>=4.67.3 ; extra == 'test'
+Requires-Dist: twine ; extra == 'test'
+Requires-Dist: ty>=0.0.16 ; extra == 'test'
+Requires-Dist: types-requests ; extra == 'test'
+Maintainer: harryhaller001
+Maintainer-email: harryhaller001 <harryhaller001@gmail.com>
+Requires-Python: >=3.11
+Project-URL: Source, https://github.com/harryhaller001/microarray
+Provides-Extra: docs
+Provides-Extra: test
+Description-Content-Type: text/markdown
+
+# microarray
+Processing microarray data in Python
+
+
+## Installation
+
+Use `pip` to install the microarray package:
+
+`pip install microarray`
+
+## Usage
+
+```python
+import microarray as ma
+```
+
+See tutorial notebook for detailed workflow examples.
+
+## License
+
+MIT

microarray-0.1.0.dist-info/RECORD

@@ -0,0 +1,44 @@
+microarray/__init__.py,sha256=VNXeP5rIxEHgkCbR69_Yvu-LTMqT-ipV7LXLXrStFF8,301
+microarray/_version.py,sha256=RlPHvA08BJFT7InYKmrqUbpk4No6mfqPzrxOXfUS1bA,81
+microarray/datasets/__init__.py,sha256=KECQdeG_6jM1BLHnolCvC5Bl26odJIMER93F634ifT4,118
+microarray/datasets/_arrayexpress.py,sha256=kcbe8qUInt3rhZ2Ioc1WheeSYvPzLWPVm2lD8fkEOas,134
+microarray/datasets/_cdf_files.py,sha256=E6F2KTA2okDaGe5YdW27JqcYLScJ7gksppMY63iT61Q,1160
+microarray/datasets/_geo.py,sha256=OlUEZfH_w6eDnnENF__3utpQGxc2OztraE3OBAUxptM,111
+microarray/datasets/_utils.py,sha256=G-BlIufEm-KUTlz_mAx8pOC5-6nyyIsZ5gU3rLuHw-M,4310
+microarray/io/__init__.py,sha256=rnBtYn59d0JQt6U3klQYgeaMV7mqJy_iP37zTfMPRiY,419
+microarray/io/_anndata_converter.py,sha256=p3_uJsx___dVVRSink6RcGlyQpyiZK6paxb-1XhVRvw,7579
+microarray/io/_cdf.py,sha256=8qHOYVOxlCx_Do0mwIvU5ahIma67LMinLsa86-8Xcqw,19825
+microarray/io/_cel.py,sha256=JCqYvgWvuTqjrcY9y129YYmwMltDYTkeS6774E98rFw,23440
+microarray/io/_read.py,sha256=uVvNsZUtoBngML-AifRACugofzQi9MfWJTNneXxNjR8,3745
+microarray/plotting/__init__.py,sha256=Fjy-omJEHAIKUD8E1EyOJSJ5hQzlPpNTpRF_EbMQwNM,817
+microarray/plotting/_base.py,sha256=0G1kI6SiF2pQt9MXdIVBrUlfII54ehPne0V4hPTEfjw,6508
+microarray/plotting/_cel.py,sha256=_6TK9n6b2bjPYsq01C69a6g1JLsKUaq6kWmnEWcGP68,1962
+microarray/plotting/_de_plots.py,sha256=IrLhM4XSjLTVn52KbVyzz-MGj2KjSNDxRSxQ6BCGH4E,8961
+microarray/plotting/_diagnostic_plots.py,sha256=cSW89oBfnx7YKOHBiH9-bGgraXde6LGCVHQH5jPW84o,9283
+microarray/plotting/_heatmap.py,sha256=cZLVJ3y9lzlrLmfNZiac0ZTpimfVlyb6sjLY5SF9s68,10694
+microarray/plotting/_ma_plots.py,sha256=Gn5SG_kLRQSUpRb73uOw98rj5MkPDU83uPh97o1NSSs,4615
+microarray/plotting/_pca.py,sha256=nWhdjt9_UUNyzpetGovVEmd1LV18PQXN07UUY-7Ozc8,10565
+microarray/plotting/_qc_plots.py,sha256=ITSVy_Lof8Ek81cK1VBz5vP5BZeW5dTrI747uKefFpk,10963
+microarray/plotting/_score.py,sha256=sR_xJPRmg_aAmQ8WNYHhj1VmqC8Q8d-DNZmwZ91ugeE,1474
+microarray/plotting/_top_table_heatmap.py,sha256=1T0J28X2xU_3T2cz1p-cD3EqUDnoqtBXfko2UhwRoOw,3693
+microarray/plotting/_utils.py,sha256=MVJjkbnFrXzJzJkm5LxcNwzIQZ9PoFi6DfsonH6RIXw,8708
+microarray/preprocessing/__init__.py,sha256=UsLBTcg_ajG4GzqHYJGSImVDU9HYzz2oYrKoRef_vcU,1086
+microarray/preprocessing/_background.py,sha256=tyGScDhgwvUvJmGvzK6w1Ch3i9vlR0gA3Y9rGykMwdg,29002
+microarray/preprocessing/_log2.py,sha256=ih57Z3kcHrF7qGI_f-Z5a4Vcl42kJVUhW1BI898OTn4,2317
+microarray/preprocessing/_normalize.py,sha256=KhV7-1kmLOouDhhEnsFwD6vJm60C7LuuVz2pPfl79W8,45811
+microarray/preprocessing/_rma.py,sha256=oZJ8xj6CDWMlJPRiWTxTHnTZ645l9JPLC_LM2m8SjS4,8674
+microarray/preprocessing/_robust.py,sha256=RsgC6lgCca5Cg1a49WIUgnhdZ7eqDztGMnunWMS2_DM,5529
+microarray/preprocessing/_summarize.py,sha256=ctv-NkZDKMScEyqKDoSKZy3szg_YoVhUyHdOj9wzV8U,10844
+microarray/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+microarray/tools/__init__.py,sha256=m2WBfDG1-rmgt3fEfEMUp-Z_051fl8tRibDN10UpcVM,677
+microarray/tools/_biomart.py,sha256=4aWBtUwxELLR4uBJHHASEdpIsR1GCbG6kkvV4QJIg8w,15201
+microarray/tools/_empirical_bayes.py,sha256=cQgV7Gx4ii-0v2TCrQpNeIMjDlUOup4kmrKzAhoVvqw,14224
+microarray/tools/_fdist.py,sha256=pA3nM72ywkLlMMDk-YrG9aISdx5ShgcPHGhgaF4vgPU,6065
+microarray/tools/_linear_models.py,sha256=xrkujMD7LUYsJpa_quDRkBGGU4A1MIr-QORzWoTMf48,14265
+microarray/tools/_mds.py,sha256=FDZJajyT81YhE8CL7vLtSoftYR2HpyObvWH1CqRqrK8,3436
+microarray/tools/_pca.py,sha256=hVg172p-jB0vAogPGnfloCljq6r4JphlB2OZu8sIYkM,2837
+microarray/tools/_score.py,sha256=fJQEB2W7tuSoJN9aQ6h-uG7bPvz12qNW-OLfqb0D_ps,2979
+microarray/tools/_toptable.py,sha256=Qy8heiPZsK7YVX_AcwxUw3c4x2kdMl4Vcx_5dWb3S1Q,14278
+microarray-0.1.0.dist-info/WHEEL,sha256=iHtWm8nRfs0VRdCYVXocAWFW8ppjHL-uTJkAdZJKOBM,80
+microarray-0.1.0.dist-info/METADATA,sha256=N8pCMHUqOaZ2xs0Sxu8Xys_1Fw7U8Ee_JxzYR7NuYt4,2320
+microarray-0.1.0.dist-info/RECORD,,

microarray-0.1.0.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: uv 0.9.30
+Root-Is-Purelib: true
+Tag: py3-none-any