Changeset 1074

Timestamp:

May 15, 2009, 2:38:58 PM (14 years ago)

Author:

Jari Häkkinen

Message:

Addresses #118. Added documentation for qQuantiler.

Location:

plugins/base2/net.sf.basedb.normalizers/trunk

Files:

• README (modified) (2 diffs)
• src/net/sf/basedb/plugins/qQuantileNormalization.java (modified) (1 diff)

plugins/base2/net.sf.basedb.normalizers/trunk/README

@@ -6 +6 @@
 plug-in set is a compilation of normalisers for expression data. See
 ``Documentation`` below for further information about the different
-plug-ins in this package. Common to the plug-ins provided with this
-package is that they work on bioassay set with either 1-channel or
-2-channel. The algorithms are working on expression values, that is
-for 2-channel data, ratios ch1/ch2 are used.
+plug-ins in this package. Common to most of the plug-ins provided with
+this package is that they work on bioassay sets with either 1-channel
+and 2-channel data. The algorithms are working on expression values,
+that is for 2-channel data, ratio ch1/ch2 are used.
 
 `Normalization package for BASE` is free software. See the file
@@ -50 +50 @@
 == qQuantile normalization ==
 
-Write me.
+The current implementation of qQuantile normalization supports only
+1-channel arrays.
 
 The qQuantile normalization is inspired by the 'Cubic Spline'
 normalization in Illumina Beadstudio and the work by Workman et al.,
 http://www.pubmedcentral.nih.gov/articlerender.fcgi?tool=pubmed&pubmedid=12225587
+
+In qQuantile normalization, all assays (including the target) are
+sorted in increasing intensity. The sorted list of probe intensities
+are partitioned into q groups, and each of theses q groups are
+adjusted (normalized) with the corresponding target group. After
+normalization the intensity distribution of each assay will be
+approximately the same as the target distribution. q is calculated as
+q=max(10,min(100,target_size/10)). The program will stop if the number
+of well defined expression values in the target or any of the assays
+in the set is smaller than q.
+
+The target is defined by selecting a subset of the assays in the
+bioassay set, and the target expression values are the medians of
+probe intensities over the bioassay set. Probes with no well defined
+measurements in the bioassay set are simply ignored in target
+calculation.
+
+Since the normalization calculations are based on geometric means and
+performed in log space the intensities must be positive and larger
+than 0. Rather than expecting the user of qQuantile normalization to
+remove such intensity the underlying algorithm silently ignores zero
+and negative intensities.
+
+Background subtraction and proper filtration should be done on the
+bioassay set before running this plug-in.
 
 

plugins/base2/net.sf.basedb.normalizers/trunk/src/net/sf/basedb/plugins/qQuantileNormalization.java

@@ -78 +78 @@
       (
         "qQuantile normalization",
-        "In q-quantile normalization ... Assays are not mixed.\n\n" +
-        "Background subtraction and proper filtration have to be done on the " +
+        "The qQuantile normalization is inspired by the 'Cubic Spline' " +
+        "normalization in Illumina Beadstudio and the work by Workman et al., " +
+        "http://www.pubmedcentral.nih.gov/articlerender.fcgi?tool=pubmed&pubmedid=12225587\n\n" +
+        "In qQuantile normalization, all assays (including the target) are " +
+        "sorted in increasing intensity. The sorted list of probe intensities " +
+        "are partitioned into q groups, and each of theses q groups are " +
+        "adjusted (normalized) with the corresponding target group. After " +
+        "normalization the intensity distribution of each assay will be " +
+        "approximately the same as the target distribution. q is calculated " +
+        "as q=max(10,min(100,target_size/10)). The program will stop if the " +
+        "number of well defined expression values in the target or any of the " +
+        "assays in the set is smaller than q.\n\n" +
+        "The target is defined by selecting a subset of the assays in the " +
+        "bioassay set, and the target expression values are the medians of " +
+        "probe intensities over the bioassay set. Probes with no well defined " +
+        "measurements in the bioassay set are simply ignored in target " +
+        "calculation.\n\n" +
+        "Since the normalization calculations are based on geometric means " +
+        "and performed in log space the intensities must be positive and " +
+        "larger than 0. Rather than expecting the user of qQuantile " +
+        "normalization to remove such intensity the underlying algorithm " +
+        "silently ignores zero and negative intensities.\n\n" +
+        "Background subtraction and proper filtration should be done on the " +
         "bioassay set before running this plug-in.\n\n" +
         "Only 1-channel data is supported\n\n" +