Skip to content

EPIraction

gentropy.datasource.intervals.epiraction.IntervalsEpiraction

Interval dataset from EPIraction.

Source code in src/gentropy/datasource/intervals/epiraction.py 
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
class IntervalsEpiraction:
    """Interval dataset from EPIraction."""

    DATASET_NAME: ClassVar[str] = "epiraction"
    PMID: ClassVar[str] = "40027634"
    VALID_INTERVAL_TYPES: ClassVar[list[str]] = ["promoter", "enhancer"]

    @staticmethod
    def read(spark: SparkSession, path: str) -> DataFrame:
        """Read EPIraction dataset (tsv with header).

        Args:
            spark (SparkSession): Spark session
            path (str): Path to the EPIraction dataset (tsv files)

        Returns:
            DataFrame: Raw EPIraction DataFrame.
        """
        return (
            spark.read.option("delimiter", "\t")
            .option("mode", "DROPMALFORMED")
            .option("header", "true")
            .csv(path)
        )

    @classmethod
    def parse(
        cls: type[IntervalsEpiraction],
        raw_epiraction_df: DataFrame,
        target_index: TargetIndex,
    ) -> Intervals:
        """Parse EPIraction Intervals.

        Args:
            raw_epiraction_df (DataFrame): Raw EPIraction DataFrame.
            target_index (TargetIndex): Target index.

        Returns:
            Intervals: Parsed Intervals dataset.
        """
        base = (
            raw_epiraction_df.filter(f.col("class").isin(cls.VALID_INTERVAL_TYPES))
            .withColumn("chromosome", f.regexp_replace(f.col("chr"), r"^chr", ""))
            .withColumnRenamed("TargetGeneEnsemblID", "geneId")
            .withColumnRenamed("CellType", "biosampleName")
            .withColumnRenamed("Score", "score")
            .withColumnRenamed("class", "intervalType")
            .withColumn(
                "resourceScore",
                f.array(
                    f.struct(
                        f.lit("H3K27ac").alias("name"),
                        f.col("H3K27ac").cast("float").alias("value"),
                    ),
                    f.struct(
                        f.lit("Open").alias("name"),
                        f.col("Open").cast("float").alias("value"),
                    ),
                    f.struct(
                        f.lit("Cofactor").alias("name"),
                        f.col("Cofactor").cast("float").alias("value"),
                    ),
                    f.struct(
                        f.lit("CTCF").alias("name"),
                        f.col("CTCF").cast("float").alias("value"),
                    ),
                    f.struct(
                        f.lit("HiC_contacts").alias("name"),
                        f.col("HiC_contacts").cast("float").alias("value"),
                    ),
                    f.struct(
                        f.lit("abc_tissue").alias("name"),
                        f.col("abc_tissue").cast("float").alias("value"),
                    ),
                ),
            )
            .withColumn("start", f.col("start").cast("long"))
            .withColumn("end", f.col("end").cast("long"))
            .withColumn("intervalType", f.lower(f.trim(f.col("intervalType"))))
        )

        # Target Index: preferred TSS (+ fallbacks)
        ti = target_index._df.select(
            f.col("id").alias("geneId"),
            f.col("tss").cast("long").alias("tss_primary"),
            f.col("canonicalTranscript.start").cast("long").alias("ct_start"),
            f.col("canonicalTranscript.end").cast("long").alias("ct_end"),
            f.col("canonicalTranscript.strand").alias("ct_strand"),
            f.col("genomicLocation.start").cast("long").alias("gl_start"),
            f.col("genomicLocation.end").cast("long").alias("gl_end"),
            f.col("genomicLocation.strand").cast("int").alias("gl_strand"),
        )

        ct_tss = f.when(f.col("ct_strand") == "+", f.col("ct_start")).when(
            f.col("ct_strand") == "-", f.col("ct_end")
        )
        gl_tss = f.when(f.col("gl_strand") == 1, f.col("gl_start")).when(
            f.col("gl_strand") == -1, f.col("gl_end")
        )

        ti_with_tss = ti.withColumn(
            "tss_from_target_index", f.coalesce(f.col("tss_primary"), ct_tss, gl_tss)
        )

        has_input_tss = "distanceToTSS" in base.columns
        base_with_fallback = (
            base.withColumn("tss_from_input", f.col("distanceToTSS").cast("long"))
            if has_input_tss
            else base.withColumn("tss_from_input", f.lit(None).cast("long"))
        )

        joined = base_with_fallback.alias("iv").join(
            ti_with_tss.alias("ti"), on="geneId", how="inner"
        )

        tss = f.coalesce(f.col("ti.tss_from_target_index"), f.col("iv.tss_from_input"))

        dist_core = f.when(
            (tss >= f.col("iv.start")) & (tss <= f.col("iv.end")), f.lit(0)
        ).otherwise(
            f.least(f.abs(tss - f.col("iv.start")), f.abs(tss - f.col("iv.end")))
        )
        distance_expr = (
            f.when(f.col("iv.intervalType") == "promoter", f.lit(0))
            .when(tss.isNull(), f.lit(None).cast("long"))
            .otherwise(dist_core)
        )

        parsed = joined.withColumn(
            "distanceToTss", distance_expr.cast("double")
        ).withColumn(
            "intervalId",
            f.sha1(
                f.concat_ws(
                    "_",
                    f.col("iv.chromosome"),
                    f.col("iv.start"),
                    f.col("iv.end"),
                    f.col("iv.geneId"),
                    f.lit(cls.DATASET_NAME),
                )
            ),
        )

        return Intervals(
            _df=(
                parsed.select(
                    f.col("iv.chromosome").alias("chromosome"),
                    f.col("iv.start").cast("string").alias("start"),
                    f.col("iv.end").cast("string").alias("end"),
                    f.col("iv.geneId").alias("geneId"),
                    f.col("iv.biosampleName").alias("biosampleName"),
                    f.col("iv.intervalType").alias("intervalType"),
                    f.col("distanceToTss").cast("integer").alias("distanceToTss"),
                    f.col("iv.score").cast("double").alias("score"),
                    f.col("iv.resourceScore").alias("resourceScore"),
                    f.lit(cls.DATASET_NAME).alias("datasourceId"),
                    f.lit(cls.PMID).alias("pmid"),
                )
            ),
            _schema=Intervals.get_schema(),
        )

parse(raw_epiraction_df: DataFrame, target_index: TargetIndex) -> Intervals classmethod

Parse EPIraction Intervals.

Parameters:

Name Type Description Default
raw_epiraction_df DataFrame

Raw EPIraction DataFrame.

 required
target_index TargetIndex

Target index.

 required

Returns:

Name Type Description
Intervals Intervals

Parsed Intervals dataset.
Source code in src/gentropy/datasource/intervals/epiraction.py 
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
@classmethod
def parse(
    cls: type[IntervalsEpiraction],
    raw_epiraction_df: DataFrame,
    target_index: TargetIndex,
) -> Intervals:
    """Parse EPIraction Intervals.

    Args:
        raw_epiraction_df (DataFrame): Raw EPIraction DataFrame.
        target_index (TargetIndex): Target index.

    Returns:
        Intervals: Parsed Intervals dataset.
    """
    base = (
        raw_epiraction_df.filter(f.col("class").isin(cls.VALID_INTERVAL_TYPES))
        .withColumn("chromosome", f.regexp_replace(f.col("chr"), r"^chr", ""))
        .withColumnRenamed("TargetGeneEnsemblID", "geneId")
        .withColumnRenamed("CellType", "biosampleName")
        .withColumnRenamed("Score", "score")
        .withColumnRenamed("class", "intervalType")
        .withColumn(
            "resourceScore",
            f.array(
                f.struct(
                    f.lit("H3K27ac").alias("name"),
                    f.col("H3K27ac").cast("float").alias("value"),
                ),
                f.struct(
                    f.lit("Open").alias("name"),
                    f.col("Open").cast("float").alias("value"),
                ),
                f.struct(
                    f.lit("Cofactor").alias("name"),
                    f.col("Cofactor").cast("float").alias("value"),
                ),
                f.struct(
                    f.lit("CTCF").alias("name"),
                    f.col("CTCF").cast("float").alias("value"),
                ),
                f.struct(
                    f.lit("HiC_contacts").alias("name"),
                    f.col("HiC_contacts").cast("float").alias("value"),
                ),
                f.struct(
                    f.lit("abc_tissue").alias("name"),
                    f.col("abc_tissue").cast("float").alias("value"),
                ),
            ),
        )
        .withColumn("start", f.col("start").cast("long"))
        .withColumn("end", f.col("end").cast("long"))
        .withColumn("intervalType", f.lower(f.trim(f.col("intervalType"))))
    )

    # Target Index: preferred TSS (+ fallbacks)
    ti = target_index._df.select(
        f.col("id").alias("geneId"),
        f.col("tss").cast("long").alias("tss_primary"),
        f.col("canonicalTranscript.start").cast("long").alias("ct_start"),
        f.col("canonicalTranscript.end").cast("long").alias("ct_end"),
        f.col("canonicalTranscript.strand").alias("ct_strand"),
        f.col("genomicLocation.start").cast("long").alias("gl_start"),
        f.col("genomicLocation.end").cast("long").alias("gl_end"),
        f.col("genomicLocation.strand").cast("int").alias("gl_strand"),
    )

    ct_tss = f.when(f.col("ct_strand") == "+", f.col("ct_start")).when(
        f.col("ct_strand") == "-", f.col("ct_end")
    )
    gl_tss = f.when(f.col("gl_strand") == 1, f.col("gl_start")).when(
        f.col("gl_strand") == -1, f.col("gl_end")
    )

    ti_with_tss = ti.withColumn(
        "tss_from_target_index", f.coalesce(f.col("tss_primary"), ct_tss, gl_tss)
    )

    has_input_tss = "distanceToTSS" in base.columns
    base_with_fallback = (
        base.withColumn("tss_from_input", f.col("distanceToTSS").cast("long"))
        if has_input_tss
        else base.withColumn("tss_from_input", f.lit(None).cast("long"))
    )

    joined = base_with_fallback.alias("iv").join(
        ti_with_tss.alias("ti"), on="geneId", how="inner"
    )

    tss = f.coalesce(f.col("ti.tss_from_target_index"), f.col("iv.tss_from_input"))

    dist_core = f.when(
        (tss >= f.col("iv.start")) & (tss <= f.col("iv.end")), f.lit(0)
    ).otherwise(
        f.least(f.abs(tss - f.col("iv.start")), f.abs(tss - f.col("iv.end")))
    )
    distance_expr = (
        f.when(f.col("iv.intervalType") == "promoter", f.lit(0))
        .when(tss.isNull(), f.lit(None).cast("long"))
        .otherwise(dist_core)
    )

    parsed = joined.withColumn(
        "distanceToTss", distance_expr.cast("double")
    ).withColumn(
        "intervalId",
        f.sha1(
            f.concat_ws(
                "_",
                f.col("iv.chromosome"),
                f.col("iv.start"),
                f.col("iv.end"),
                f.col("iv.geneId"),
                f.lit(cls.DATASET_NAME),
            )
        ),
    )

    return Intervals(
        _df=(
            parsed.select(
                f.col("iv.chromosome").alias("chromosome"),
                f.col("iv.start").cast("string").alias("start"),
                f.col("iv.end").cast("string").alias("end"),
                f.col("iv.geneId").alias("geneId"),
                f.col("iv.biosampleName").alias("biosampleName"),
                f.col("iv.intervalType").alias("intervalType"),
                f.col("distanceToTss").cast("integer").alias("distanceToTss"),
                f.col("iv.score").cast("double").alias("score"),
                f.col("iv.resourceScore").alias("resourceScore"),
                f.lit(cls.DATASET_NAME).alias("datasourceId"),
                f.lit(cls.PMID).alias("pmid"),
            )
        ),
        _schema=Intervals.get_schema(),
    )

read(spark: SparkSession, path: str) -> DataFrame staticmethod

Read EPIraction dataset (tsv with header).

Parameters:

Name Type Description Default
spark SparkSession

Spark session

 required
path str

Path to the EPIraction dataset (tsv files)

 required

Returns:

Name Type Description
DataFrame DataFrame

Raw EPIraction DataFrame.
Source code in src/gentropy/datasource/intervals/epiraction.py 
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
@staticmethod
def read(spark: SparkSession, path: str) -> DataFrame:
    """Read EPIraction dataset (tsv with header).

    Args:
        spark (SparkSession): Spark session
        path (str): Path to the EPIraction dataset (tsv files)

    Returns:
        DataFrame: Raw EPIraction DataFrame.
    """
    return (
        spark.read.option("delimiter", "\t")
        .option("mode", "DROPMALFORMED")
        .option("header", "true")
        .csv(path)
    )
  • 2025-10-28
  • 2025-10-31
  • Contributors