Sampling adjustment (#84)

* Entropy Debiasing Adjustment * fixed deps (#85) (#86) * fixed deps * no more lower versions of Python * no more lower versions of Python * trying stable rust * fixing old deps * fixing old deps * fixing imports * clang * fixing xlsx * fixing more xlsx crate issues * fixing more xlsx crate issues * fixing more xlsx crate issues * removing xlxs support * removing xlxs support * ready to release * updated lock * trying to support 3.8 * trying to support more python versions * updated lock * no 3.12 cuz maturin issues * updated lock --------- Co-authored-by: Shan Tharanga <[email protected]> --------- Co-authored-by: Eyyub Unlu <[email protected]> Co-authored-by: Shan Tharanga <[email protected]>
BVU-BILSAB · Sep 3, 2024 · eb96f38 · eb96f38
1 parent 3cc713b
commit eb96f38
Showing 1 changed file with 19 additions and 7 deletions.
diff --git a/src/lib.rs b/src/lib.rs
@@ -594,8 +594,16 @@ fn calculate_entropy(kmers: &[Box<str>], support_threshold: &usize) -> f64 {
     // and we do not need to skew the regression with unreliable data
     // We stop at 99% because at 100% we do not do resampling and all kmers are used which we
     // calculated above
-    let mut entropy_values: Vec<(f64, f64)> = (percentage_cutoff..99)
-        .into_iter()
+    let starting_point = if percentage_cutoff < 50 {
+        50
+    } else {
+        // Nearest multiple of 5 if between 30 and 50 (inclusive)
+        (percentage_cutoff + 4) / 5 * 5
+    };
+
+
+    let mut entropy_values: Vec<(f64, f64)> = (starting_point..100)
+        .step_by(5)
         .map(|percentage| {
             // Figure out the number of samples that this % represents
             let samples = (percentage * kmer_count) / 100;
@@ -613,13 +621,17 @@ fn calculate_entropy(kmers: &[Box<str>], support_threshold: &usize) -> f64 {
     // One of the data points has to be 100% of the k-mers WITHOUT random sampling
     entropy_values.push((1.0 / kmer_count as f64, all_kmers_entropy));
 
-    let (_, y) = linear_regression_of(&entropy_values).unwrap();
+    if entropy_values.len() >= 5 {
+        let (_, y) = linear_regression_of(&entropy_values).unwrap();
 
-    if y < 0_f64 {
-        return all_kmers_entropy;
-    }
+        if y < 0_f64 {
+            return all_kmers_entropy;
+        }
 
     y
+    } else {
+    return all_kmers_entropy;
+ }
 }
 
 fn get_kmers_and_headers(
@@ -882,4 +894,4 @@ fn dima(_py: Python, m: &PyModule) -> PyResult<()> {
     m.add_class::<Variant>()?;
     m.add_class::<Results>()?;
     Ok(())
-}
+}