aboutsummaryrefslogtreecommitdiff
path: root/gn3/computations
diff options
context:
space:
mode:
authorMuriithi Frederick Muriuki2021-07-20 15:12:38 +0300
committerMuriithi Frederick Muriuki2021-07-20 15:12:38 +0300
commit8dc60fd93b9eab9fd655a1f0c23ed038d4ac3eed (patch)
tree94305f9b648bffd99116ae28f93fadc52ed09392 /gn3/computations
parent5a8f4f3c85be4645c9a918bc25397170f4370341 (diff)
downloadgenenetwork3-8dc60fd93b9eab9fd655a1f0c23ed038d4ac3eed.tar.gz
Implement the correlation function
Issue: https://github.com/genenetwork/gn-gemtext-threads/blob/main/topics/gn1-migration-to-gn2/clustering.gmi * Implement the correlation computation function, such that it passes the tests created previously.
Diffstat (limited to 'gn3/computations')
-rw-r--r--gn3/computations/correlations2.py33
1 files changed, 32 insertions, 1 deletions
diff --git a/gn3/computations/correlations2.py b/gn3/computations/correlations2.py
index 193f646..6c456db 100644
--- a/gn3/computations/correlations2.py
+++ b/gn3/computations/correlations2.py
@@ -1,4 +1,35 @@
+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):
+ if (item1 is not None) and (item2 is not None):
+ return (item1, item2)
+ return None
+ def split_lists(accumulator, item):
+ 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):
- return tuple()
+ x, y = items_with_values(dbdata, userdata)
+ if len(x) < 6:
+ return (0.0, len(x))
+ meanx = sum(x)/len(x)
+ meany = sum(y)/len(y)
+ 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, y), [0.0, 0.0, 0.0])
+ try:
+ return ((xyd/(sqrt(sxd)*sqrt(syd))), len(x))
+ except ZeroDivisionError as zde:
+ return(0, len(x))