From b7b1868ba85c38a4accd9fae85dddfd4b59f478c Mon Sep 17 00:00:00 2001
From: John Vivian <jtvivian@gmail.com>
Date: Thu, 30 May 2019 13:48:39 -0700
Subject: [PATCH] Change ANOVA to percent of total genes (resolves #56)

---
 gene_outlier_detection/lib.py        | 13 ++++---------
 tests/test_gene_outlier_detection.py |  3 +--
 2 files changed, 5 insertions(+), 11 deletions(-)

diff --git a/gene_outlier_detection/lib.py b/gene_outlier_detection/lib.py
index a52e8a4..99c9fdb 100644
--- a/gene_outlier_detection/lib.py
+++ b/gene_outlier_detection/lib.py
@@ -94,7 +94,7 @@ def anova_distances(
     df: pd.DataFrame,
     genes: List[str],
     group: str = "tissue",
-    n_genes=2000,
+    percent_genes=0.10,
 ):
     """
     Calculates distance to each group via pairwise distance using top N ANOVA genes
@@ -104,19 +104,14 @@ def anova_distances(
         df: background dataset
         genes: genes to use for pairwise distance
         group: Column to use as class discriminator
-        n_genes: Number of ANOVA genes to use
+        percent_genes: Percent of ANOVA genes to use for pairwise distance
 
     Returns:
         DataFrame of pairwise distances
     """
     click.echo(f"Ranking background datasets by {group} via ANOVA")
-    if n_genes >= len(genes):
-        click.secho(
-            f"# of ANOVA genes {n_genes} greater than {len(genes)}", fg="yellow"
-        )
-        skb_genes = genes
-    else:
-        skb_genes = select_k_best_genes(df, genes, n=n_genes)
+    n_genes = int(percent_genes * len(genes))
+    skb_genes = select_k_best_genes(df, genes, n=n_genes)
     dist = pairwise_distances(np.array(sample[skb_genes]).reshape(1, -1), df[skb_genes])
     dist = pd.DataFrame([dist.ravel(), df["tissue"]]).T
     dist.columns = ["Distance", "Group"]
diff --git a/tests/test_gene_outlier_detection.py b/tests/test_gene_outlier_detection.py
index 8767d4c..1fe8af0 100644
--- a/tests/test_gene_outlier_detection.py
+++ b/tests/test_gene_outlier_detection.py
@@ -106,9 +106,8 @@ def test_anova_distances(load_data):
 
     sample, df, genes = load_data
     dist = anova_distances(sample, df, genes)
-    print(dist)
     assert list(dist.Group) == ["Thyroid", "Brain"]
-    assert [int(x) for x in dist.MedianDistance] == [57, 131]
+    assert [int(x) for x in dist.MedianDistance] == [63, 142]
 
 
 def test_run_model(model_output):