Merge pull request #178 from Proteobench/issue40_CV

added median CV, Q75 and Q90 to datapoint.result

proteobench/modules/dda_quant/datapoint.py

@@ -7,9 +7,9 @@
 import pandas as pd
 
 
-def filter_df_numquant_weighted_sum(row, min_quant=3):
+def filter_df_numquant_median_abs_epsilon(row, min_quant=3):
     if isinstance(row, dict) and min_quant in row and isinstance(row[min_quant], dict):
-        return row[min_quant].get("weighted_sum")
+        return row[min_quant].get("median_abs_epsilon")
     return None
 
 
@@ -41,8 +41,6 @@ class Datapoint:
     allowed_miscleavages: int = 0
     min_peptide_length: int = 0
     max_peptide_length: int = 0
-    weighted_sum: int = 0
-    nr_prec: int = 0
     is_temporary: bool = True
     intermediate_hash: str = ""
     results: dict = None

proteobench/modules/dda_quant/module.py

@@ -170,10 +170,28 @@ def get_metrics(df, min_nr_observed=1):
         # take abs value of df["epsilon"]
         # TODO use nr_missing to filter df before computing stats.
         df_slice = df[df["nr_observed"] >= min_nr_observed]
-        weighted_sum = round(df_slice["epsilon"].abs().mean(), ndigits=3)
         nr_prec = len(df_slice)
-
-        return {min_nr_observed: {"weighted_sum": weighted_sum, "nr_prec": nr_prec}}
+        # median abs unafected by outliers
+        median_abs_epsilon = df_slice["epsilon"].abs().mean()
+        # variance affected by outliers
+        variance_epsilon = df_slice["epsilon"].var()
+        # TODO more concise way to describe distribution of CV's
+        cv_median = (df_slice["CV_A"].median() + df_slice["CV_B"].median()) / 2
+        cv_q75 = (df_slice["CV_A"].quantile(0.75) + df_slice["CV_B"].quantile(0.75)) / 2
+        cv_q90 = (df_slice["CV_A"].quantile(0.9) + df_slice["CV_B"].quantile(0.9)) / 2
+        cv_q95 = (df_slice["CV_A"].quantile(0.95) + df_slice["CV_B"].quantile(0.95)) / 2
+
+        return {
+            min_nr_observed: {
+                "median_abs_epsilon": median_abs_epsilon,
+                "variance_epsilon": variance_epsilon,
+                "nr_prec": nr_prec,
+                "CV_median": cv_median,
+                "CV_q90": cv_q90,
+                "CV_q75": cv_q75,
+                "CV_q95": cv_q95,
+            }
+        }
 
     def generate_datapoint(
         self,

proteobench/modules/dda_quant/plot.py

@@ -92,7 +92,7 @@ def plot_metric(self, benchmark_metrics_df: pd.DataFrame) -> go.Figure:
         fig = go.Figure(
             data=[
                 go.Scatter(
-                    x=benchmark_metrics_df["weighted_sum"],
+                    x=benchmark_metrics_df["median_abs_epsilon"],
                     y=benchmark_metrics_df["nr_prec"],
                     mode="markers",
                     text=hover_texts,

proteobench/modules/template/datapoint.py

@@ -46,8 +46,6 @@ class Datapoint:
     missed_cleavages: int = 0
     min_pep_length: int = 0
     max_pep_length: int = 0
-    weighted_sum: int = 0
-    nr_prec: int = 0
 
     def calculate_benchmarking_metric_1(self, intermediate_data):
         """Calculates the first benchmarking metric based on the intermediate data.

test/test_module_dda_quant.py

@@ -128,7 +128,7 @@ def test_benchmarking(self):
             TESTDATA_FILES["MaxQuant"], "MaxQuant", user_input, None
         )
         self.assertTrue(isinstance(all_datapoints, pd.DataFrame))
-        self.assertEqual(len(all_datapoints.results[1]), 6)
+        self.assertEqual(len(all_datapoints.results[len(all_datapoints.results) - 1]), 6)
 
 
 class TestWrongFormatting(unittest.TestCase):

webinterface/pages/DDA_Quant.py

@@ -10,8 +10,8 @@
 from streamlit_extras.let_it_rain import rain
 
 from proteobench.modules.dda_quant.datapoint import (
+    filter_df_numquant_median_abs_epsilon,
     filter_df_numquant_nr_prec,
-    filter_df_numquant_weighted_sum,
 )
 from proteobench.modules.dda_quant.module import Module
 from proteobench.modules.dda_quant.parse_settings import (
@@ -198,7 +198,9 @@ def _main_page(self):
             all_datapoints = st.session_state[ALL_DATAPOINTS]
             all_datapoints = Module().obtain_all_data_point(all_datapoints)
 
-            all_datapoints["weighted_sum"] = all_datapoints.apply(filter_df_numquant_weighted_sum, min_quant=3)
+            all_datapoints["median_abs_epsilon"] = all_datapoints.apply(
+                filter_df_numquant_median_abs_epsilon, min_quant=3
+            )
             all_datapoints["nr_prec"] = all_datapoints.apply(filter_df_numquant_nr_prec, min_quant=3)
 
             fig2 = PlotDataPoint().plot_metric(all_datapoints)
@@ -246,8 +248,9 @@ def _run_proteobench(self):
 
     def slider_callback(self):
         min_quant = st.session_state[st.session_state["slider_id"]]
-        st.session_state[ALL_DATAPOINTS]["weighted_sum"] = [
-            filter_df_numquant_weighted_sum(v, min_quant=min_quant) for v in st.session_state[ALL_DATAPOINTS]["results"]
+        st.session_state[ALL_DATAPOINTS]["median_abs_epsilon"] = [
+            filter_df_numquant_median_abs_epsilon(v, min_quant=min_quant)
+            for v in st.session_state[ALL_DATAPOINTS]["results"]
         ]
         st.session_state[ALL_DATAPOINTS]["nr_prec"] = [
             filter_df_numquant_nr_prec(v, min_quant=min_quant) for v in st.session_state[ALL_DATAPOINTS]["results"]
@@ -331,7 +334,7 @@ def generate_results(
 
         if recalculate:
             all_datapoints["weighted_sum"] = [
-                filter_df_numquant_weighted_sum(v, min_quant=3) for v in all_datapoints["results"]
+                filter_df_numquant_median_abs_epsilon(v, min_quant=3) for v in all_datapoints["results"]
             ]
             all_datapoints["nr_prec"] = [filter_df_numquant_nr_prec(v, min_quant=3) for v in all_datapoints["results"]]