edgepython 0.2.0__py3-none-any.whl

edgepython/__init__.py

@@ -0,0 +1,114 @@
+# This code was written by Claude (Anthropic). The project was directed by Lior Pachter.
+"""
+edgePython: Python port of the edgeR Bioconductor package.
+
+Empirical analysis of digital gene expression data in Python.
+"""
+
+__version__ = "0.1.0"
+
+# --- Classes ---
+from .classes import DGEList, DGEExact, DGEGLM, DGELRT, TopTags
+from .classes import cbind_dgelist, rbind_dgelist
+from .compressed_matrix import CompressedMatrix
+
+# --- DGEList construction & accessors ---
+from .dgelist import (
+    make_dgelist,
+    valid_dgelist,
+    get_counts,
+    get_dispersion,
+    get_dispersion_type,
+    get_offset,
+    get_norm_lib_sizes,
+)
+
+# --- Normalization ---
+from .normalization import calc_norm_factors, normalize_chip_to_input, calc_norm_offsets_for_chip
+
+# --- Expression ---
+from .expression import cpm, rpkm, tpm, ave_log_cpm, cpm_by_group, rpkm_by_group
+
+# --- Filtering ---
+from .filtering import filter_by_expr
+
+# --- Dispersion estimation ---
+from .dispersion import (
+    estimate_disp,
+    WLEB,
+    estimate_common_disp,
+    estimate_tagwise_disp,
+    estimate_trended_disp,
+    estimate_glm_common_disp,
+    estimate_glm_trended_disp,
+    estimate_glm_tagwise_disp,
+)
+
+# --- GLM fitting ---
+from .glm_fit import glm_fit, glm_ql_fit, mglm_one_group, mglm_one_way
+
+# --- GLM testing ---
+from .glm_test import glm_lrt, glm_ql_ftest, glm_treat
+
+# --- Exact test ---
+from .exact_test import (
+    exact_test,
+    exact_test_double_tail,
+    equalize_lib_sizes,
+    q2q_nbinom,
+    split_into_groups,
+)
+
+# --- Results ---
+from .results import top_tags, decide_tests
+
+# --- I/O ---
+from .io import (
+    read_data,
+    to_anndata,
+    read_dge,
+    read_10x,
+    catch_salmon,
+    catch_kallisto,
+    catch_rsem,
+    feature_counts_to_dgelist,
+    read_bismark2dge,
+    seurat_to_pb,
+)
+
+# --- Visualization ---
+from .visualization import (
+    plot_md,
+    plot_bcv,
+    plot_mds,
+    plot_smear,
+    plot_ql_disp,
+    ma_plot,
+    gof,
+)
+
+# --- Splicing ---
+from .splicing import diff_splice, diff_splice_dge, splice_variants
+
+# --- Gene sets ---
+from .gene_sets import camera, fry, roast, mroast, romer, goana, kegga
+
+# --- Utilities ---
+from .utils import (
+    model_matrix,
+    model_matrix_meth,
+    nearest_tss,
+    add_prior_count,
+    pred_fc,
+    good_turing,
+    thin_counts,
+    gini,
+    sum_tech_reps,
+    zscore_nbinom,
+)
+
+# --- Single-cell GLM ---
+from .sc_fit import glm_sc_fit, glm_sc_test, shrink_sc_disp
+
+# --- limma utilities ---
+from .limma_port import squeeze_var

edgepython/classes.py

@@ -0,0 +1,517 @@
+# This code was written by Claude (Anthropic). The project was directed by Lior Pachter.
+"""
+Core data classes for edgePython.
+
+Port of edgeR's S4 classes (DGEList, DGEExact, DGEGLM, DGELRT, TopTags)
+as Python dataclasses with dict-like access, subsetting, and display.
+"""
+
+import numpy as np
+import pandas as pd
+from copy import deepcopy
+
+
+class _EdgeRBase(dict):
+    """Base class providing dict-like access, subsetting, and display."""
+
+    def __getattr__(self, name):
+        try:
+            return self[name]
+        except KeyError:
+            raise AttributeError(f"'{type(self).__name__}' has no attribute '{name}'")
+
+    def __setattr__(self, name, value):
+        self[name] = value
+
+    def __delattr__(self, name):
+        try:
+            del self[name]
+        except KeyError:
+            raise AttributeError(f"'{type(self).__name__}' has no attribute '{name}'")
+
+    @property
+    def shape(self):
+        if 'counts' in self:
+            return self['counts'].shape
+        return None
+
+    def __repr__(self):
+        cls = type(self).__name__
+        components = list(self.keys())
+        s = self.shape
+        if s is not None:
+            return f"{cls} with {s[0]} rows and {s[1]} columns\nComponents: {', '.join(components)}"
+        return f"{cls}\nComponents: {', '.join(components)}"
+
+    def _copy(self):
+        """Deep copy of the object."""
+        return deepcopy(self)
+
+    def head(self, n=5):
+        """Show first n rows."""
+        if 'table' in self:
+            return self['table'].head(n)
+        if 'counts' in self:
+            return pd.DataFrame(
+                self['counts'][:n],
+                index=_get_rownames(self)[:n] if _get_rownames(self) is not None else None,
+                columns=_get_colnames(self) if _get_colnames(self) is not None else None
+            )
+        return None
+
+    def tail(self, n=5):
+        """Show last n rows."""
+        if 'table' in self:
+            return self['table'].tail(n)
+        if 'counts' in self:
+            return pd.DataFrame(
+                self['counts'][-n:],
+                index=_get_rownames(self)[-n:] if _get_rownames(self) is not None else None,
+                columns=_get_colnames(self) if _get_colnames(self) is not None else None
+            )
+        return None
+
+
+def _get_rownames(obj):
+    """Get row names from genes or counts."""
+    if 'genes' in obj and obj['genes'] is not None:
+        return list(obj['genes'].index)
+    if 'counts' in obj and obj['counts'] is not None:
+        c = obj['counts']
+        if hasattr(c, 'index'):
+            return list(c.index)
+    return None
+
+
+def _get_colnames(obj):
+    """Get column names from samples or counts."""
+    if 'samples' in obj and obj['samples'] is not None:
+        return list(obj['samples'].index)
+    return None
+
+
+def _subset_matrix_or_df(x, i=None, j=None):
+    """Subset a matrix, DataFrame, or vector by row (i) and/or column (j)."""
+    if x is None:
+        return None
+    if isinstance(x, pd.DataFrame):
+        if i is not None and j is not None:
+            return x.iloc[i, j]
+        elif i is not None:
+            return x.iloc[i]
+        elif j is not None:
+            return x.iloc[:, j]
+        return x
+    if isinstance(x, np.ndarray):
+        if x.ndim == 2:
+            if i is not None and j is not None:
+                return x[np.ix_(np.atleast_1d(i), np.atleast_1d(j))] if not isinstance(i, slice) else x[i, :][:, j]
+            elif i is not None:
+                return x[i] if isinstance(i, slice) else x[np.atleast_1d(i)]
+            elif j is not None:
+                return x[:, j] if isinstance(j, slice) else x[:, np.atleast_1d(j)]
+        elif x.ndim == 1:
+            if i is not None:
+                return x[i] if isinstance(i, slice) else x[np.atleast_1d(i)]
+        return x
+    return x
+
+
+def _resolve_index(idx, names):
+    """Resolve index to integer array. Supports bool, int, str, slice."""
+    if idx is None:
+        return None
+    if isinstance(idx, slice):
+        return idx
+    idx = np.atleast_1d(idx)
+    if idx.dtype == bool:
+        return np.where(idx)[0]
+    if idx.dtype.kind in ('U', 'S', 'O') and names is not None:
+        names_arr = np.asarray(names)
+        result = []
+        for name in idx:
+            matches = np.where(names_arr == name)[0]
+            if len(matches) == 0:
+                raise KeyError(f"Name '{name}' not found")
+            result.append(matches[0])
+        return np.array(result)
+    return idx.astype(int)
+
+
+class DGEList(_EdgeRBase):
+    """Digital Gene Expression data list.
+
+    Attributes
+    ----------
+    counts : ndarray
+        Matrix of counts (genes x samples).
+    samples : DataFrame
+        Sample information with columns group, lib.size, norm.factors.
+    genes : DataFrame or None
+        Gene annotation.
+    common.dispersion : float or None
+    trended.dispersion : ndarray or None
+    tagwise.dispersion : ndarray or None
+    AveLogCPM : ndarray or None
+    offset : ndarray or None
+    weights : ndarray or None
+    """
+
+    _IJ = {'counts', 'pseudo.counts', 'offset', 'weights'}
+    _IX = {'genes'}
+    _JX = {'samples'}
+    _I = {'AveLogCPM', 'trended.dispersion', 'tagwise.dispersion', 'prior.n', 'prior.df'}
+
+    def __getitem__(self, key):
+        if isinstance(key, str):
+            return super().__getitem__(key)
+        if isinstance(key, tuple):
+            if len(key) == 2:
+                i, j = key
+            else:
+                raise IndexError("Two subscripts required")
+        else:
+            raise IndexError("Two subscripts required")
+
+        rownames = _get_rownames(self)
+        colnames = _get_colnames(self)
+        i_idx = _resolve_index(i, rownames)
+        j_idx = _resolve_index(j, colnames)
+
+        out = self._copy()
+
+        for k in self._IJ:
+            if k in out and out[k] is not None:
+                out[k] = _subset_matrix_or_df(out[k], i_idx, j_idx)
+        for k in self._IX:
+            if k in out and out[k] is not None:
+                out[k] = _subset_matrix_or_df(out[k], i_idx)
+        for k in self._JX:
+            if k in out and out[k] is not None:
+                out[k] = _subset_matrix_or_df(out[k], j=j_idx)
+        for k in self._I:
+            if k in out and out[k] is not None:
+                out[k] = _subset_matrix_or_df(out[k], i_idx)
+
+        # Drop empty group levels after column subsetting
+        if j_idx is not None and 'samples' in out and 'group' in out['samples'].columns:
+            out['samples']['group'] = out['samples']['group'].cat.remove_unused_categories() if hasattr(out['samples']['group'], 'cat') else out['samples']['group']
+
+        return out
+
+    def __setitem__(self, key, value):
+        super().__setitem__(key, value)
+
+    @property
+    def nrow(self):
+        if 'counts' in self:
+            return self['counts'].shape[0]
+        return 0
+
+    @property
+    def ncol(self):
+        if 'counts' in self:
+            return self['counts'].shape[1]
+        return 0
+
+    def __len__(self):
+        return self.nrow
+
+    def dim(self):
+        return self.shape
+
+    def dimnames(self):
+        return (_get_rownames(self), _get_colnames(self))
+
+    def to_dataframe(self):
+        """Convert counts to DataFrame."""
+        return pd.DataFrame(
+            self['counts'],
+            index=_get_rownames(self),
+            columns=_get_colnames(self)
+        )
+
+
+class DGEExact(_EdgeRBase):
+    """Results of exact test for differential expression.
+
+    Attributes
+    ----------
+    table : DataFrame
+        With columns logFC, logCPM, PValue.
+    comparison : list
+        Two group names being compared.
+    genes : DataFrame or None
+    """
+
+    def __getitem__(self, key):
+        if isinstance(key, str):
+            return super().__getitem__(key)
+        if isinstance(key, tuple):
+            if len(key) == 2:
+                i, j = key
+            else:
+                raise IndexError("Two subscripts required")
+        else:
+            raise IndexError("Two subscripts required (rows, columns)")
+
+        if j is not None:
+            raise IndexError("Subsetting columns not allowed for DGEExact objects.")
+
+        out = self._copy()
+        rownames = list(out['table'].index) if out.get('table') is not None else None
+        i_idx = _resolve_index(i, rownames)
+
+        if 'table' in out and out['table'] is not None:
+            out['table'] = _subset_matrix_or_df(out['table'], i_idx)
+        if 'genes' in out and out['genes'] is not None:
+            out['genes'] = _subset_matrix_or_df(out['genes'], i_idx)
+        return out
+
+    def __repr__(self):
+        out = ""
+        if 'comparison' in self and self['comparison'] is not None:
+            out += f"Comparison of groups: {self['comparison'][1]}-{self['comparison'][0]}\n"
+        if 'table' in self and self['table'] is not None:
+            out += str(self['table'])
+        return out
+
+    @property
+    def shape(self):
+        if 'table' in self:
+            return self['table'].shape
+        return None
+
+
+class DGEGLM(_EdgeRBase):
+    """Fitted GLM object for DGE data.
+
+    Attributes
+    ----------
+    coefficients : ndarray
+        Matrix of coefficients.
+    fitted.values : ndarray
+    deviance : ndarray
+    counts : ndarray
+    offset : ndarray or CompressedMatrix
+    weights : ndarray or None
+    design : ndarray
+    dispersion : float or ndarray
+    df.residual : ndarray
+    samples : DataFrame
+    genes : DataFrame or None
+    AveLogCPM : ndarray or None
+    """
+
+    _IX = {'counts', 'offset', 'weights', 'genes', 'coefficients', 'fitted.values',
+           'unshrunk.coefficients', 'leverage', 'unit.deviance.adj', 'unit.df.adj'}
+    _I = {'AveLogCPM', 'dispersion', 'prior.n', 'prior.df', 's2.post', 's2.prior',
+          'df.prior', 'df.residual', 'df.residual.zeros', 'df.residual.adj',
+          'deviance', 'deviance.adj', 'iter', 'failed'}
+
+    def __getitem__(self, key):
+        if isinstance(key, str):
+            return super().__getitem__(key)
+        if isinstance(key, tuple):
+            if len(key) == 2:
+                i, j = key
+            else:
+                raise IndexError("Two subscripts required")
+        else:
+            raise IndexError("Two subscripts required")
+
+        if j is not None:
+            raise IndexError("Subsetting columns not allowed for DGEGLM object.")
+
+        out = self._copy()
+        rownames = _get_rownames(self)
+        i_idx = _resolve_index(i, rownames)
+
+        for k in self._IX:
+            if k in out and out[k] is not None:
+                out[k] = _subset_matrix_or_df(out[k], i_idx)
+        for k in self._I:
+            if k in out and out[k] is not None:
+                out[k] = _subset_matrix_or_df(out[k], i_idx)
+        return out
+
+    @property
+    def nrow(self):
+        if 'coefficients' in self:
+            return self['coefficients'].shape[0]
+        if 'counts' in self:
+            return self['counts'].shape[0]
+        return 0
+
+    @property
+    def ncol(self):
+        if 'counts' in self:
+            return self['counts'].shape[1]
+        return 0
+
+
+class DGELRT(_EdgeRBase):
+    """Likelihood ratio test results for DGE GLMs.
+
+    Attributes
+    ----------
+    table : DataFrame
+        With columns logFC, logCPM, LR (or F), PValue.
+    comparison : str
+        Name of coefficient tested.
+    coefficients, fitted.values, etc. : inherited from DGEGLM fit.
+    """
+
+    _IX = {'counts', 'offset', 'weights', 'genes', 'coefficients', 'fitted.values',
+           'table', 'unshrunk.coefficients', 'leverage', 'unit.deviance.adj', 'unit.df.adj'}
+    _I = {'AveLogCPM', 'dispersion', 'prior.n', 'prior.df', 's2.post', 's2.prior',
+          'df.prior', 'df.residual', 'df.residual.zeros', 'df.residual.adj',
+          'deviance', 'deviance.adj', 'iter', 'failed', 'df.test', 'df.total'}
+
+    def __getitem__(self, key):
+        if isinstance(key, str):
+            return super().__getitem__(key)
+        if isinstance(key, tuple):
+            if len(key) == 2:
+                i, j = key
+            else:
+                raise IndexError("Two subscripts required")
+        else:
+            raise IndexError("Two subscripts required")
+
+        if j is not None:
+            raise IndexError("Subsetting columns not allowed for DGELRT object.")
+
+        out = self._copy()
+        rownames = _get_rownames(self)
+        if rownames is None and 'table' in self:
+            rownames = list(self['table'].index)
+        i_idx = _resolve_index(i, rownames)
+
+        for k in self._IX:
+            if k in out and out[k] is not None:
+                out[k] = _subset_matrix_or_df(out[k], i_idx)
+        for k in self._I:
+            if k in out and out[k] is not None:
+                out[k] = _subset_matrix_or_df(out[k], i_idx)
+        return out
+
+    def __repr__(self):
+        out = ""
+        if 'comparison' in self:
+            out += f"Coefficient: {self['comparison']}\n"
+        if 'table' in self:
+            out += str(self['table'])
+        return out
+
+    @property
+    def shape(self):
+        if 'table' in self:
+            return self['table'].shape
+        if 'coefficients' in self:
+            return self['coefficients'].shape
+        return None
+
+
+class TopTags(_EdgeRBase):
+    """Top differentially expressed genes.
+
+    Attributes
+    ----------
+    table : DataFrame
+        Sorted table of top genes.
+    adjust_method : str
+    comparison : str or list
+    test : str
+        Either 'exact' or 'glm'.
+    """
+
+    def __getitem__(self, key):
+        if isinstance(key, str):
+            return super().__getitem__(key)
+        if isinstance(key, tuple):
+            if len(key) == 2:
+                i, j = key
+            else:
+                raise IndexError("Two subscripts required")
+        else:
+            raise IndexError("Two subscripts required")
+
+        out = self._copy()
+        if 'table' in out:
+            if i is not None or j is not None:
+                out['table'] = _subset_matrix_or_df(out['table'], i, j)
+        return out
+
+    def __repr__(self):
+        out = ""
+        if self.get('test') == 'exact':
+            comp = self.get('comparison', [])
+            if len(comp) >= 2:
+                out += f"Comparison of groups: {comp[1]}-{comp[0]}\n"
+        else:
+            out += f"Coefficient: {self.get('comparison', '')}\n"
+        if 'table' in self:
+            out += str(self['table'])
+        return out
+
+    @property
+    def shape(self):
+        if 'table' in self:
+            return self['table'].shape
+        return None
+
+
+def cbind_dgelist(*objects):
+    """Column-bind (combine samples) DGEList objects.
+
+    Port of edgeR's cbind.DGEList.
+    """
+    if len(objects) == 1:
+        return objects[0]
+
+    out = objects[0]._copy()
+    for obj in objects[1:]:
+        # Check gene compatibility
+        if 'genes' in out and out['genes'] is not None:
+            if not out['genes'].equals(obj.get('genes')):
+                raise ValueError("DGEList objects have different genes")
+
+        out['counts'] = np.hstack([out['counts'], obj['counts']])
+        out['samples'] = pd.concat([out['samples'], obj['samples']], ignore_index=False)
+
+        for key in ['offset', 'weights', 'pseudo.counts']:
+            if key in out and out[key] is not None and key in obj and obj[key] is not None:
+                out[key] = np.hstack([out[key], obj[key]])
+
+    # Clear dispersions
+    for key in ['common.dispersion', 'trended.dispersion', 'tagwise.dispersion']:
+        if key in out:
+            del out[key]
+
+    return out
+
+
+def rbind_dgelist(*objects):
+    """Row-bind (combine genes) DGEList objects.
+
+    Port of edgeR's rbind.DGEList.
+    """
+    if len(objects) == 1:
+        return objects[0]
+
+    out = objects[0]._copy()
+    for obj in objects[1:]:
+        out['counts'] = np.vstack([out['counts'], obj['counts']])
+        if 'genes' in out and out['genes'] is not None and 'genes' in obj and obj['genes'] is not None:
+            out['genes'] = pd.concat([out['genes'], obj['genes']], ignore_index=False)
+        for key in ['offset', 'weights', 'pseudo.counts']:
+            if key in out and out[key] is not None and key in obj and obj[key] is not None:
+                out[key] = np.vstack([out[key], obj[key]])
+
+    # Clear dispersions
+    for key in ['common.dispersion', 'trended.dispersion', 'tagwise.dispersion', 'AveLogCPM']:
+        if key in out:
+            del out[key]
+
+    return out