Changes for v1.2.0 (#9)

Fix standard error calculation for combined replicates.
Update manuscript link in documentation.
Update version string.

docs/conf.py

@@ -51,9 +51,9 @@
 # built documents.
 #
 # The short X.Y version.
-version = '1.1'
+version = '1.2'
 # The full version, including alpha/beta/rc tags.
-release = '1.1.1'
+release = '1.2.0'
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.
@@ -272,7 +272,7 @@
 
 .. _HDF5: http://pandas.pydata.org/pandas-docs/stable/io.html#hdf5-pytables
 
-.. _Enrich2 manuscript: http://biorxiv.org/content/early/2016/09/25/075150
+.. _Enrich2 manuscript: https://genomebiology.biomedcentral.com/articles/10.1186/s13059-017-1272-5
 
 .. Replacement aliases for intersphinx library documentation
 

docs/index.rst

@@ -1,4 +1,3 @@
-
 Enrich2: deep mutational scanning data analysis
 ============================================================================
 
@@ -12,6 +11,12 @@ To cite Enrich2, please reference `A statistical framework for analyzing deep mu
 
 Enrich2 was written by `Alan F Rubin <mailto:[email protected]>`_ |ORCID_icon| http://orcid.org/0000-0003-1474-605X
 
+.. error:: Important notice for users of Enrich2 v1.0 or v1.1
+
+Enrich2 v1.2.0 corrects an error in the software that, for most datasets, resulted in the standard errors for combined scores being over-estimated. The counts, scores, and replicate-wise standard errors are unaffected.
+
+If you have analyzed datasets that contain replicates with a previous version of Enrich2, the easiest way to get the correct standard error values is to delete the experiment HDF5_ file (the file name ends with ``'_exp.h5'``) and re-run the program. This will recalculate combined scores and standard errors without redoing other parts of the analysis.
+
 .. |ORCID_icon| image:: _static/iD_icon.png
     :target: http://orcid.org
 

enrich2/experiment.py

@@ -394,9 +394,9 @@ def calc_scores(self, label):
                     data.loc[:, idx[cnd, 'epsilon']] = 0.
                 # multiple replicates
                 else:
-                    betaML, sigma2ML, eps = rml_estimator(y, sigma2i)
+                    betaML, var_betaML, eps = rml_estimator(y, sigma2i)
                     data.loc[:, idx[cnd, 'score']] = betaML
-                    data.loc[:, idx[cnd, 'SE']] = np.sqrt(sigma2ML)
+                    data.loc[:, idx[cnd, 'SE']] = np.sqrt(var_betaML)
                     data.loc[:, idx[cnd, 'epsilon']] = eps
 
                 # special case for normalized wild type variant

enrich2/main.py

@@ -49,7 +49,7 @@
 __author__ = "Alan F Rubin"
 __copyright__ = "Copyright 2016-2017, Alan F Rubin"
 __license__ = "GPLv3"
-__version__ = "1.1.1"
+__version__ = "1.2.0"
 __maintainer__ = "Alan F Rubin"
 __email__ = "[email protected]"
 

enrich2/random_effects.py

@@ -45,4 +45,5 @@ def rml_estimator(y, sigma2i, iterations=50):
                                          axis=0) / (sw - (sw2 / sw))
         eps = np.abs(sigma2ML - sigma2ML_new)
         sigma2ML = sigma2ML_new
-    return betaML, sigma2ML, eps
+    var_betaML = 1 / np.sum(1 / (sigma2i + sigma2ML), axis=0)
+    return betaML, var_betaML, eps

setup.py

@@ -36,7 +36,7 @@
 
 setup(
     name="Enrich2",
-    version="1.1.1",
+    version="1.2.0",
 
     packages=find_packages(),
     package_data={