feat: change StudyLocusId hashing method to md5 (and change StudyLocusId to string type)

src/gentropy/assets/schemas/colocalisation.json

@@ -4,13 +4,13 @@
     {
       "name": "leftStudyLocusId",
       "nullable": false,
-      "type": "long",
+      "type": "string",
       "metadata": {}
     },
     {
       "name": "rightStudyLocusId",
       "nullable": false,
-      "type": "long",
+      "type": "string",
       "metadata": {}
     },
     {

src/gentropy/assets/schemas/l2g_feature.json

@@ -3,7 +3,7 @@
   "fields": [
     {
       "name": "studyLocusId",
-      "type": "long",
+      "type": "string",
       "nullable": false,
       "metadata": {}
     },

src/gentropy/assets/schemas/l2g_gold_standard.json

@@ -3,7 +3,7 @@
   "fields": [
     {
       "name": "studyLocusId",
-      "type": "long",
+      "type": "string",
       "nullable": false,
       "metadata": {}
     },

src/gentropy/assets/schemas/l2g_predictions.json

@@ -3,7 +3,7 @@
   "fields": [
     {
       "name": "studyLocusId",
-      "type": "long",
+      "type": "string",
       "nullable": false,
       "metadata": {}
     },

src/gentropy/assets/schemas/study_locus.json

@@ -4,7 +4,7 @@
       "metadata": {},
       "name": "studyLocusId",
       "nullable": false,
-      "type": "long"
+      "type": "string"
     },
     {
       "metadata": {},

src/gentropy/assets/schemas/study_locus_overlap.json

@@ -4,13 +4,13 @@
       "metadata": {},
       "name": "leftStudyLocusId",
       "nullable": false,
-      "type": "long"
+      "type": "string"
     },
     {
       "metadata": {},
       "name": "rightStudyLocusId",
       "nullable": false,
-      "type": "long"
+      "type": "string"
     },
     {
       "metadata": {},

src/gentropy/dataset/dataset.py

@@ -310,3 +310,18 @@ def flag_duplicates(test_column: Column) -> Column:
             )
             > 1
         )
+
+    @staticmethod
+    def generate_identifier(uniqueness_defining_columns: list[str]) -> Column:
+        """Hashes the provided columns to generate a unique identifier.
+
+        Args:
+            uniqueness_defining_columns (list[str]): list of columns defining uniqueness
+
+        Returns:
+            Column: column with a unique identifier
+        """
+        hashable_columns = [f.when(f.col(column).cast("string").isNull(), f.lit("None"))
+                                 .otherwise(f.col(column).cast("string"))
+                                 for column in uniqueness_defining_columns]
+        return f.md5(f.concat(*hashable_columns))

src/gentropy/dataset/study_locus.py

@@ -447,38 +447,28 @@ def _align_overlapping_tags(
         )
 
     @staticmethod
-    def assign_study_locus_id(
-        study_id_col: Column,
-        variant_id_col: Column,
-        finemapping_col: Column = None,
-    ) -> Column:
-        """Hashes a column with a variant ID and a study ID to extract a consistent studyLocusId.
+    def assign_study_locus_id(uniqueness_defining_columns: list[str]) -> Column:
+        """Hashes the provided columns to extract a consistent studyLocusId.
 
         Args:
-            study_id_col (Column): column name with a study ID
-            variant_id_col (Column): column name with a variant ID
-            finemapping_col (Column, optional): column with fine mapping methodology
+            uniqueness_defining_columns (list[str]): list of columns defining uniqueness
 
         Returns:
             Column: column with a study locus ID
 
         Examples:
             >>> df = spark.createDataFrame([("GCST000001", "1_1000_A_C", "SuSiE-inf"), ("GCST000002", "1_1000_A_C", "pics")]).toDF("studyId", "variantId", "finemappingMethod")
-            >>> df.withColumn("study_locus_id", StudyLocus.assign_study_locus_id(f.col("studyId"), f.col("variantId"), f.col("finemappingMethod"))).show()
-            +----------+----------+-----------------+-------------------+
-            |   studyId| variantId|finemappingMethod|     study_locus_id|
-            +----------+----------+-----------------+-------------------+
-            |GCST000001|1_1000_A_C|        SuSiE-inf|3801266831619496075|
-            |GCST000002|1_1000_A_C|             pics|1581844826999194430|
-            +----------+----------+-----------------+-------------------+
+            >>> df.withColumn("study_locus_id", StudyLocus.assign_study_locus_id(["studyId", "variantId", "finemappingMethod"])).show(truncate=False)
+            +----------+----------+-----------------+--------------------------------+
+            |studyId   |variantId |finemappingMethod|study_locus_id                  |
+            +----------+----------+-----------------+--------------------------------+
+            |GCST000001|1_1000_A_C|SuSiE-inf        |109804fe1e20c94231a31bafd71b566e|
+            |GCST000002|1_1000_A_C|pics             |de310be4558e0482c9cc359c97d37773|
+            +----------+----------+-----------------+--------------------------------+
             <BLANKLINE>
         """
-        if finemapping_col is None:
-            finemapping_col = f.lit(None).cast(StringType())
-        variant_id_col = f.coalesce(variant_id_col, f.rand().cast("string"))
-        return f.xxhash64(study_id_col, variant_id_col, finemapping_col).alias(
-            "studyLocusId"
-        )
+        return Dataset.generate_identifier(uniqueness_defining_columns).alias("studyLocusId")
+
 
     @classmethod
     def calculate_credible_set_log10bf(cls: type[StudyLocus], logbfs: Column) -> Column:

src/gentropy/datasource/eqtl_catalogue/finemapping.py

@@ -260,7 +260,7 @@ def from_susie_results(
                 .select(
                     *study_locus_cols,
                     StudyLocus.assign_study_locus_id(
-                        f.col("studyId"), f.col("variantId"), f.col("finemappingMethod")
+                        ["studyId", "variantId", "finemappingMethod"]
                     ),
                     StudyLocus.calculate_credible_set_log10bf(
                         f.col("locus.logBF")

src/gentropy/datasource/finngen/finemapping.py

@@ -471,7 +471,7 @@ def from_finngen_susie_finemapping(
         ).withColumn(
             "studyLocusId",
             StudyLocus.assign_study_locus_id(
-                f.col("studyId"), f.col("variantId"), f.col("finemappingMethod")
+                ["studyId", "variantId", "finemappingMethod"]
             ),
         )
 

src/gentropy/datasource/gwas_catalog/associations.py

@@ -9,7 +9,7 @@
 from typing import TYPE_CHECKING
 
 import pyspark.sql.functions as f
-from pyspark.sql.types import DoubleType, FloatType, IntegerType, LongType
+from pyspark.sql.types import DoubleType, FloatType, IntegerType, StringType
 from pyspark.sql.window import Window
 
 from gentropy.assets import data
@@ -1109,7 +1109,7 @@ def from_source(
         """
         return StudyLocusGWASCatalog(
             _df=gwas_associations.withColumn(
-                "studyLocusId", f.monotonically_increasing_id().cast(LongType())
+                "studyLocusId", f.monotonically_increasing_id().cast(StringType())
             )
             .transform(
                 # Map/harmonise variants to variant annotation dataset:
@@ -1188,7 +1188,7 @@ def update_study_id(
             .drop("subStudyDescription", "updatedStudyId")
         ).withColumn(
             "studyLocusId",
-            StudyLocus.assign_study_locus_id(f.col("studyId"), f.col("variantId")),
+            StudyLocus.assign_study_locus_id(["studyId", "variantId"]),
         )
         return self
 

src/gentropy/datasource/open_targets/l2g_gold_standard.py

@@ -52,7 +52,7 @@ def parse_positive_curation(
             )
             .withColumn(
                 "studyLocusId",
-                StudyLocus.assign_study_locus_id(f.col("studyId"), f.col("variantId")),
+                StudyLocus.assign_study_locus_id(["studyId", "variantId"]),
             )
             .groupBy("studyLocusId", "studyId", "variantId", "geneId")
             .agg(f.collect_set("source").alias("sources"))

src/gentropy/l2g.py

@@ -207,17 +207,22 @@ def _generate_feature_matrix(self, write_feature_matrix: bool) -> L2GFeatureMatr
             study_locus_overlap = StudyLocus(
                 _df=self.credible_set.df.join(
                     f.broadcast(
-                        self.gs_curation.select(
-                            StudyLocus.assign_study_locus_id(
-                                f.col("association_info.otg_id"),  # studyId
-                                f.concat_ws(  # variantId
+                        self.gs_curation
+                        .withColumn(
+                            "variantId",
+                            f.concat_ws(
                                     "_",
                                     f.col("sentinel_variant.locus_GRCh38.chromosome"),
                                     f.col("sentinel_variant.locus_GRCh38.position"),
                                     f.col("sentinel_variant.alleles.reference"),
                                     f.col("sentinel_variant.alleles.alternative"),
-                                ),
-                            ).alias("studyLocusId"),
+                            )
+                        )
+                        .select(
+                            StudyLocus.assign_study_locus_id(
+                                ["association_info.otg_id",  # studyId
+                                "variantId"]
+                            ),
                         )
                     ),
                     "studyLocusId",

src/gentropy/method/locus_breaker_clumping.py

@@ -112,8 +112,8 @@ def locus_breaker(
                     .cast(t.ArrayType(t.StringType()))
                     .alias("qualityControls"),
                     StudyLocus.assign_study_locus_id(
-                        f.col("studyId"), f.col("variantId")
-                    ).alias("studyLocusId"),
+                        ["studyId", "variantId"]
+                    ),
                 )
             ),
             _schema=StudyLocus.get_schema(),

src/gentropy/method/pics.py

@@ -257,7 +257,7 @@ def finemap(
                 .withColumn(
                     "studyLocusId",
                     StudyLocus.assign_study_locus_id(
-                        "studyId", "variantId", "finemappingMethod"
+                        ["studyId", "variantId", "finemappingMethod"]
                     ),
                 )
                 .drop("neglog_pvalue")

src/gentropy/method/window_based_clumping.py

@@ -247,7 +247,7 @@ def clump(
                 .withColumn(
                     "studyLocusId",
                     StudyLocus.assign_study_locus_id(
-                        f.col("studyId"), f.col("variantId")
+                        ["studyId", "variantId"]
                     ),
                 )
                 # Initialize QC column as array of strings:

src/gentropy/susie_finemapper.py

@@ -95,7 +95,7 @@ def __init__(
             .df.withColumn(
                 "studyLocusId",
                 StudyLocus.assign_study_locus_id(
-                    "studyId", "variantId", "finemappingMethod"
+                    ["studyId", "variantId", "finemappingMethod"]
                 ),
             )
             .collect()[0]
@@ -247,7 +247,7 @@ def susie_inf_to_studylocus(
                 .withColumn(
                     "studyLocusId",
                     StudyLocus.assign_study_locus_id(
-                        f.col("studyId"), f.col("variantId"), f.col("finemappingMethod")
+                        ["studyId", "variantId", "finemappingMethod"]
                     ),
                 )
                 .select(

tests/gentropy/conftest.py

@@ -617,10 +617,10 @@ def mock_l2g_feature_matrix(spark: SparkSession) -> L2GFeatureMatrix:
     return L2GFeatureMatrix(
         _df=spark.createDataFrame(
             [
-                (1, "gene1", 100.0, None),
-                (2, "gene2", 1000.0, 0.0),
+                ("1", "gene1", 100.0, None),
+                ("2", "gene2", 1000.0, 0.0),
             ],
-            "studyLocusId LONG, geneId STRING, distanceTssMean FLOAT, distanceTssMinimum FLOAT",
+            "studyLocusId STRING, geneId STRING, distanceTssMean FLOAT, distanceTssMinimum FLOAT",
         ),
         with_gold_standard=False,
     )

tests/gentropy/dataset/test_colocalisation.py

@@ -72,12 +72,12 @@ def _setup(self: TestAppendStudyMetadata, spark: SparkSession) -> None:
             _df=spark.createDataFrame(
                 [
                     (
-                        1,
+                        "1",
                         "var1",
                         "gwas1",
                     ),
                     (
-                        2,
+                        "2",
                         "var2",
                         "eqtl1",
                     ),
@@ -100,7 +100,7 @@ def _setup(self: TestAppendStudyMetadata, spark: SparkSession) -> None:
         )
         self.sample_colocalisation = Colocalisation(
             _df=spark.createDataFrame(
-                [(1, 2, "eqtl", "X", "COLOC", 1, 0.9)],
+                [("1", "2", "eqtl", "X", "COLOC", 1, 0.9)],
                 [
                     "leftStudyLocusId",
                     "rightStudyLocusId",