Skip to content

ENCODE-rE2G

gentropy.datasource.intervals.e2g.IntervalsE2G

Interval dataset from E2G.

Source code in src/gentropy/datasource/intervals/e2g.py 
 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
class IntervalsE2G:
    """Interval dataset from E2G."""

    DATASET_NAME: ClassVar[str] = "E2G"
    PMID: ClassVar[str] = "38014075"  # PMID for the E2G paper
    VALID_INTERVAL_TYPES: ClassVar[list[str]] = ["promoter", "genic", "intergenic"]

    @staticmethod
    def read(spark: SparkSession, path: str) -> DataFrame:
        """Read E2G dataset.

        Args:
            spark (SparkSession): Spark session
            path (str): Path to the E2G dataset (tsv.gz files)

        Returns:
            DataFrame: Raw E2G DataFrame.
        """
        return (
            spark.read.option("delimiter", "\t")
            .option("header", "true")
            .csv(path)
            .withColumn("file_path", f.input_file_name())
        )

    @classmethod
    def parse(
        cls: type[IntervalsE2G],
        raw_e2g_df: DataFrame,
        biosample_mapping: DataFrame,
        target_index: TargetIndex,
        biosample_index: BiosampleIndex,
    ) -> Intervals:
        """Parse E2G dataset.

        Args:
            raw_e2g_df (DataFrame): Raw E2G DataFrame.
            biosample_mapping (DataFrame): Biosample mapping DataFrame.
            target_index (TargetIndex): Target index.
            biosample_index (BiosampleIndex): Biosample index.

        Returns:
            Intervals: Parsed Intervals dataset.
        """
        base = (
            raw_e2g_df.withColumn(
                "studyId", f.regexp_extract(f.col("file_path"), r"([^/]+)\.bed\.gz$", 1)
            )
            .withColumn("chromosome", f.regexp_replace(f.col("chr"), "^chr", ""))
            .withColumnRenamed("TargetGeneEnsemblID", "geneId")
            .withColumnRenamed("CellType", "biosampleName")
            .withColumnRenamed("Score", "score")
            .withColumnRenamed("class", "intervalType")
            .withColumn(
                "resourceScore",
                f.array(
                    f.struct(
                        f.lit("DNase").alias("name"),
                        f.col("`normalizedDNase_prom.Feature`")
                        .cast("float")
                        .alias("value"),
                    ),
                    f.struct(
                        f.lit("HiC_contacts").alias("name"),
                        f.col("`3DContact.Feature`").cast("float").alias("value"),
                    ),
                ),
            )
            .withColumn("start", f.col("start").cast("long"))
            .withColumn("end", f.col("end").cast("long"))
        )

        base = base.withColumn(
            "intervalType", f.lower(f.trim(f.col("intervalType")))
        ).filter(f.col("intervalType").isin(cls.VALID_INTERVAL_TYPES))

        # Target Index: preferred TSS + fallbacks (canonical transcript, genomicLocation)
        ti = target_index._df.select(
            f.col("id").alias("geneId"),
            f.col("tss").cast("long").alias("tss_primary"),
            f.col("canonicalTranscript.chromosome").alias("ct_chr"),
            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.chromosome").alias("gl_chr"),
            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_coalesced", f.coalesce(f.col("tss_primary"), ct_tss, gl_tss)
        )

        # Join & recompute distanceToTss
        joined = base.alias("iv").join(
            ti_with_tss.alias("ti"), on="geneId", how="inner"
        )
        tss = f.col("ti.tss_coalesced")

        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("_", "chromosome", "start", "end", "geneId", "studyId")
                ),
            )
            .join(
                biosample_mapping.select("biosampleName", "biosampleId"),
                on="biosampleName",
                how="left",
            )
            .join(
                biosample_index.df.select("biosampleId"), on="biosampleId", how="inner"
            )
        )

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

parse(raw_e2g_df: DataFrame, biosample_mapping: DataFrame, target_index: TargetIndex, biosample_index: BiosampleIndex) -> Intervals classmethod

Parse E2G dataset.

Parameters:

Name Type Description Default
raw_e2g_df DataFrame

Raw E2G DataFrame.

 required
biosample_mapping DataFrame

Biosample mapping DataFrame.

 required
target_index TargetIndex

Target index.

 required
biosample_index BiosampleIndex

Biosample index.

 required

Returns:

Name Type Description
Intervals Intervals

Parsed Intervals dataset.
Source code in src/gentropy/datasource/intervals/e2g.py 
 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
@classmethod
def parse(
    cls: type[IntervalsE2G],
    raw_e2g_df: DataFrame,
    biosample_mapping: DataFrame,
    target_index: TargetIndex,
    biosample_index: BiosampleIndex,
) -> Intervals:
    """Parse E2G dataset.

    Args:
        raw_e2g_df (DataFrame): Raw E2G DataFrame.
        biosample_mapping (DataFrame): Biosample mapping DataFrame.
        target_index (TargetIndex): Target index.
        biosample_index (BiosampleIndex): Biosample index.

    Returns:
        Intervals: Parsed Intervals dataset.
    """
    base = (
        raw_e2g_df.withColumn(
            "studyId", f.regexp_extract(f.col("file_path"), r"([^/]+)\.bed\.gz$", 1)
        )
        .withColumn("chromosome", f.regexp_replace(f.col("chr"), "^chr", ""))
        .withColumnRenamed("TargetGeneEnsemblID", "geneId")
        .withColumnRenamed("CellType", "biosampleName")
        .withColumnRenamed("Score", "score")
        .withColumnRenamed("class", "intervalType")
        .withColumn(
            "resourceScore",
            f.array(
                f.struct(
                    f.lit("DNase").alias("name"),
                    f.col("`normalizedDNase_prom.Feature`")
                    .cast("float")
                    .alias("value"),
                ),
                f.struct(
                    f.lit("HiC_contacts").alias("name"),
                    f.col("`3DContact.Feature`").cast("float").alias("value"),
                ),
            ),
        )
        .withColumn("start", f.col("start").cast("long"))
        .withColumn("end", f.col("end").cast("long"))
    )

    base = base.withColumn(
        "intervalType", f.lower(f.trim(f.col("intervalType")))
    ).filter(f.col("intervalType").isin(cls.VALID_INTERVAL_TYPES))

    # Target Index: preferred TSS + fallbacks (canonical transcript, genomicLocation)
    ti = target_index._df.select(
        f.col("id").alias("geneId"),
        f.col("tss").cast("long").alias("tss_primary"),
        f.col("canonicalTranscript.chromosome").alias("ct_chr"),
        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.chromosome").alias("gl_chr"),
        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_coalesced", f.coalesce(f.col("tss_primary"), ct_tss, gl_tss)
    )

    # Join & recompute distanceToTss
    joined = base.alias("iv").join(
        ti_with_tss.alias("ti"), on="geneId", how="inner"
    )
    tss = f.col("ti.tss_coalesced")

    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("_", "chromosome", "start", "end", "geneId", "studyId")
            ),
        )
        .join(
            biosample_mapping.select("biosampleName", "biosampleId"),
            on="biosampleName",
            how="left",
        )
        .join(
            biosample_index.df.select("biosampleId"), on="biosampleId", how="inner"
        )
    )

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

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

Read E2G dataset.

Parameters:

Name Type Description Default
spark SparkSession

Spark session

 required
path str

Path to the E2G dataset (tsv.gz files)

 required

Returns:

Name Type Description
DataFrame DataFrame

Raw E2G DataFrame.
Source code in src/gentropy/datasource/intervals/e2g.py 
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
@staticmethod
def read(spark: SparkSession, path: str) -> DataFrame:
    """Read E2G dataset.

    Args:
        spark (SparkSession): Spark session
        path (str): Path to the E2G dataset (tsv.gz files)

    Returns:
        DataFrame: Raw E2G DataFrame.
    """
    return (
        spark.read.option("delimiter", "\t")
        .option("header", "true")
        .csv(path)
        .withColumn("file_path", f.input_file_name())
    )
  • 2025-10-28
  • 2025-10-31
  • Contributors