causarray 0.0.5__tar.gz → 0.0.6__tar.gz

{causarray-0.0.5 → causarray-0.0.6}/.gitignore

@@ -1,3 +1,6 @@
+# Planning documents
+plan/
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]
@@ -160,5 +163,20 @@ cython_debug/
 #.idea/
 
 
-test*
+# Generic module tests under tests/ ARE tracked (regression coverage for
+# the causarray API).  Exclude only benchmark / scratch files that are
+# not part of the unit-test suite — add new exclusions here when
+# benchmark or tutorial-coupled test files appear.
+tests/benchmark_adamson.py
+
 causarray/___*.py
+/docs/source/tutorial/adamson
+
+# Replogle tutorial: large data files (regenerate with prep_tutorial_data.py or
+# download from the project data repository; replogle_subset.h5ad is ~2 GB)
+docs/source/tutorial/replogle/replogle_subset.h5ad
+docs/source/tutorial/replogle/replogle_normed.h5ad
+docs/source/tutorial/replogle/replogle_results.h5
+
+# Unrelated tutorial directory (not part of causarray package docs)
+docs/source/tutorial/scp/

{causarray-0.0.5 → causarray-0.0.6}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.4
 Name: causarray
-Version: 0.0.5
+Version: 0.0.6
 Summary: causarray is a Python module for simultaneous causal inference with an array of outcomes.
-Author-email: Jin-Hong Du <jinhongd@andrew.cmu.com>, Maya Shen <myshen@andrew.cmu.edu>, Hansruedi Mathys <mathysh@pitt.edu>, Kathryn Roeder <jinhongd@andrew.cmu.com>
-Maintainer-email: Jin-Hong Du <jinhongd@andrew.cmu.com>
+Author-email: Jin-Hong Du <jinhongd@hku.hk>, Maya Shen <myshen@andrew.cmu.edu>, Hansruedi Mathys <mathysh@pitt.edu>, Kathryn Roeder <jinhongd@hku.hk>
+Maintainer-email: Jin-Hong Du <jinhongd@hku.hk>
 License: MIT License
 License-File: LICENSE
 Classifier: Intended Audience :: Developers
@@ -49,17 +49,44 @@ The module can be installed via PyPI:
 pip install causarray
 ```
 
+For optimal parallel performance, we recommend installing `llvm-openmp` if using conda:
+```cmd
+conda install -c conda-forge llvm-openmp
+```
+
 For `R` users, `reticulate` can be used to call `causarray` from `R`.
 The documentation and tutorials using both `Python` and `R` are available at [causarray.readthedocs.io](https://causarray.readthedocs.io/en/latest/).
 
 
 
-## Logs
+## Batch fitting for large-scale screens
+
+For screens with hundreds to thousands of perturbations, use the batch API
+so that peak memory is bounded by one batch at a time:
+
+```python
+from causarray import gcate_lfc_batch
+
+df_res = gcate_lfc_batch(
+    Y, X, A, r,
+    batch_size=10,    # perturbations per batch (or use n_batches= for a fixed count)
+    max_cells=2000,   # max pert cells per batch (ctrl added on top)
+    n_ctrl=2000,      # fixed ctrl subsample shared across batches
+    cache_path='results.h5',   # resume if interrupted
+    verbose=True,
+)
+```
+
+See the [Replogle-E-K562 tutorial](https://causarray.readthedocs.io/en/latest/)
+for a demonstration on 200 perturbations from a genome-wide CRISPRi screen.
+
+## Changelog
 
 - [x] (2025-01-30) Python package released on PyPI
-- [x] (2025-02-01) code for reproducing figures in paper
+- [x] (2025-02-01) Code for reproducing figures in paper
 - [x] (2025-02-02) Tutorial for Python and R
-- [ ] Documentation
+- [x] (2026-05-31) Batch fitting API (`gcate_lfc_batch`) for large-scale screens
+- [x] (2026-05-31) Documentation at [causarray.readthedocs.io](https://causarray.readthedocs.io/en/latest/)
 
 
 <!-- 

{causarray-0.0.5 → causarray-0.0.6}/README.md

@@ -24,17 +24,44 @@ The module can be installed via PyPI:
 pip install causarray
 ```
 
+For optimal parallel performance, we recommend installing `llvm-openmp` if using conda:
+```cmd
+conda install -c conda-forge llvm-openmp
+```
+
 For `R` users, `reticulate` can be used to call `causarray` from `R`.
 The documentation and tutorials using both `Python` and `R` are available at [causarray.readthedocs.io](https://causarray.readthedocs.io/en/latest/).
 
 
 
-## Logs
+## Batch fitting for large-scale screens
+
+For screens with hundreds to thousands of perturbations, use the batch API
+so that peak memory is bounded by one batch at a time:
+
+```python
+from causarray import gcate_lfc_batch
+
+df_res = gcate_lfc_batch(
+    Y, X, A, r,
+    batch_size=10,    # perturbations per batch (or use n_batches= for a fixed count)
+    max_cells=2000,   # max pert cells per batch (ctrl added on top)
+    n_ctrl=2000,      # fixed ctrl subsample shared across batches
+    cache_path='results.h5',   # resume if interrupted
+    verbose=True,
+)
+```
+
+See the [Replogle-E-K562 tutorial](https://causarray.readthedocs.io/en/latest/)
+for a demonstration on 200 perturbations from a genome-wide CRISPRi screen.
+
+## Changelog
 
 - [x] (2025-01-30) Python package released on PyPI
-- [x] (2025-02-01) code for reproducing figures in paper
+- [x] (2025-02-01) Code for reproducing figures in paper
 - [x] (2025-02-02) Tutorial for Python and R
-- [ ] Documentation
+- [x] (2026-05-31) Batch fitting API (`gcate_lfc_batch`) for large-scale screens
+- [x] (2026-05-31) Documentation at [causarray.readthedocs.io](https://causarray.readthedocs.io/en/latest/)
 
 
 <!-- 

{causarray-0.0.5 → causarray-0.0.6}/causarray/DR_estimation.py

@@ -3,6 +3,7 @@ from sklearn.linear_model import LogisticRegression
 from sklearn.tree import DecisionTreeClassifier, DecisionTreeRegressor
 from sklearn_ensemble_cv import reset_random_seeds, Ensemble, ECV
 from causarray.gcate_glm import fit_glm
+import causarray.gcate_glm as _gcate_glm  # module-qualified so _USE_FAST_BACKEND changes take effect at call time
 from causarray.utils import *
 from causarray.utils import _filter_params
 from joblib import Parallel, delayed
@@ -101,7 +102,19 @@ def cross_fitting(
     fit_pi = True if pi_hat is None else False
     pi_hat = np.zeros_like(A, dtype=float) if fit_pi else pi_hat
     fit_Y = True if Y_hat is None else False
-    Y_hat = np.zeros((Y.shape[0],Y.shape[1],A.shape[1],2), dtype=float) if fit_Y else Y_hat
+    if fit_Y:
+        _yhat_gb = Y.shape[0] * Y.shape[1] * A.shape[1] * 2 * 8 / 1e9
+        _mem_limit_gb = kwargs.get('mem_limit_gb', None)
+        if _mem_limit_gb is not None and _yhat_gb > _mem_limit_gb:
+            import warnings
+            warnings.warn(
+                f"Y_hat allocation ({_yhat_gb:.1f} GB as float64) exceeds "
+                f"mem_limit_gb={_mem_limit_gb} GB; using float32 to halve peak memory.",
+                ResourceWarning, stacklevel=3,
+            )
+            Y_hat = np.zeros((Y.shape[0], Y.shape[1], A.shape[1], 2), dtype=np.float32)
+        else:
+            Y_hat = np.zeros((Y.shape[0], Y.shape[1], A.shape[1], 2), dtype=float)
 
     # perform ECV at once
     if fit_pi and ps_model == 'random_forest_cv':
@@ -145,9 +158,19 @@ def cross_fitting(
 
         if fit_Y:
             if verbose: pprint.pprint('Fit outcome models...')
+            # Subset offset to training fold (for fitting) and test fold (for
+            # imputation) when it is a pre-computed array, so that
+            # ``fit_glm_auto`` receives arrays with matching leading
+            # dimensions in both stages.
+            params_glm_fold = params_glm
+            offset_test_arr = None
+            if 'offset' in params_glm and isinstance(params_glm['offset'], np.ndarray):
+                params_glm_fold = dict(params_glm)
+                params_glm_fold['offset'] = params_glm['offset'][train_index]
+                offset_test_arr = params_glm['offset'][test_index]
             # Fit GLM on training data and predict on test data
-            res = fit_glm(Y_train, X_train, A_train, family=family, alpha=glm_alpha,
-                impute=X_test, **params_glm)
+            res = _gcate_glm.fit_glm_auto(Y_train, X_train, A_train, family=family, alpha=glm_alpha,
+                impute=X_test, offset_test=offset_test_arr, **params_glm_fold)
             Y_hat[test_index,:,:,0] = res[1][0]
             Y_hat[test_index,:,:,1] = res[1][1]
 

{causarray-0.0.5 → causarray-0.0.6}/causarray/DR_inference.py

@@ -150,7 +150,7 @@ def fdx_control(
     return V
 
 
-def bh_correction(tvalues_init):
+def bh_correction(tvalues_init, df=None):
     '''
     Perform BH correction.
 
@@ -179,7 +179,10 @@ def bh_correction(tvalues_init):
     qvals_adj = np.full(tvalues_init.shape, np.nan)
 
     if np.sum(idx) > 0:
-        pvals[idx] = sp.stats.norm.sf(np.abs(tvalues_init[idx]))*2
+        if df is not None:
+            pvals[idx] = sp.stats.t.sf(np.abs(tvalues_init[idx]), df=df[idx]) * 2
+        else:
+            pvals[idx] = sp.stats.norm.sf(np.abs(tvalues_init[idx])) * 2
         qvals[idx] = multipletests(pvals[idx], alpha=0.05, method='fdr_bh')[1]
 
         # BH correction with empirical null adjustment
@@ -189,7 +192,10 @@ def bh_correction(tvalues_init):
         if mad>0:
             tvalues_init_adj = (tvalues_init - med) / mad
 
-            pvals_adj[idx] = sp.stats.norm.sf(np.abs(tvalues_init_adj[idx]))*2
+            if df is not None:
+                pvals_adj[idx] = sp.stats.t.sf(np.abs(tvalues_init_adj[idx]), df=df[idx]) * 2
+            else:
+                pvals_adj[idx] = sp.stats.norm.sf(np.abs(tvalues_init_adj[idx])) * 2
             qvals_adj[idx] = multipletests(pvals_adj[idx], alpha=0.05, method='fdr_bh')[1]
     
     return pvals, qvals, pvals_adj, qvals_adj