about summary refs log tree commit diff
path: root/gn3/computations/correlations2.py
diff options
context:
space:
mode:
Diffstat (limited to 'gn3/computations/correlations2.py')
-rw-r--r--gn3/computations/correlations2.py50
1 files changed, 50 insertions, 0 deletions
diff --git a/gn3/computations/correlations2.py b/gn3/computations/correlations2.py
new file mode 100644
index 0000000..93db3fa
--- /dev/null
+++ b/gn3/computations/correlations2.py
@@ -0,0 +1,50 @@
+"""
+DESCRIPTION:
+    TODO: Add a description for the module
+
+FUNCTIONS:
+compute_correlation:
+    TODO: Describe what the function does..."""
+
+from math import sqrt
+from functools import reduce
+## From GN1: mostly for clustering and heatmap generation
+
+def __items_with_values(dbdata, userdata):
+    """Retains only corresponding items in the data items that are not `None` values.
+    This should probably be renamed to something sensible"""
+    def both_not_none(item1, item2):
+        """Check that both items are not the value `None`."""
+        if (item1 is not None) and (item2 is not None):
+            return (item1, item2)
+        return None
+    def split_lists(accumulator, item):
+        """Separate the 'x' and 'y' items."""
+        return [accumulator[0] + [item[0]], accumulator[1] + [item[1]]]
+    return reduce(
+        split_lists,
+        filter(lambda x: x is not None, map(both_not_none, dbdata, userdata)),
+        [[], []])
+
+def compute_correlation(dbdata, userdata):
+    """Compute some form of correlation.
+
+    This is extracted from
+    https://github.com/genenetwork/genenetwork1/blob/master/web/webqtl/utility/webqtlUtil.py#L622-L647
+    """
+    x_items, y_items = __items_with_values(dbdata, userdata)
+    if len(x_items) < 6:
+        return (0.0, len(x_items))
+    meanx = sum(x_items)/len(x_items)
+    meany = sum(y_items)/len(y_items)
+    def cal_corr_vals(acc, item):
+        xitem, yitem = item
+        return [
+            acc[0] + ((xitem - meanx) * (yitem - meany)),
+            acc[1] + ((xitem - meanx) * (xitem - meanx)),
+            acc[2] + ((yitem - meany) * (yitem - meany))]
+    xyd, sxd, syd = reduce(cal_corr_vals, zip(x_items, y_items), [0.0, 0.0, 0.0])
+    try:
+        return ((xyd/(sqrt(sxd)*sqrt(syd))), len(x_items))
+    except ZeroDivisionError:
+        return(0, len(x_items))