diff --git a/src/download_encode_eclip/create_gene_bed.py b/src/download_encode_eclip/create_gene_bed.py
new file mode 100644
index 0000000000000000000000000000000000000000..f695072dde4694a272714f73ab8c4237d7d85ace
--- /dev/null
+++ b/src/download_encode_eclip/create_gene_bed.py
@@ -0,0 +1,201 @@
+#!/usr/bin/env python3
+
+# -*- coding: UTF-8 -*-
+
+"""
+Description: The goal of this script is to create a bed gene file \
+without the first and the last exons.
+"""
+
+
+from .config import Config, ConfigEncodeClip
+import pandas as pd
+from pathlib import Path
+import doctest
+
+
+class BadIDError(Exception):
+ """
+ Class defining a bad exon id error.
+ """
+ pass
+
+
+def check_id(mserie: pd.Series) -> None:
+ """
+ Check the id column of a row in a dataframe containing exons.
+
+ :param mserie: A row of a dataframe containing exon in a bed format
+
+ >>> check_id(pd.Series({"chr": 1, "start": 0, "stop": 50, "id": "1_10",
+ ... "score": 0, "strand": '-'}))
+ >>> check_id(pd.Series({"chr": 1, "start": 0, "stop": 50, "id": "cd_10",
+ ... "score": 0, "strand": '-'}))
+ Traceback (most recent call last):
+ ...
+ BadIDError: The exon cd_10 as a bad id format.
+ It should have the form id_pos where
+ 1. id: Is the fasterDB gene id of the exons
+ 2. pos: Is the position of the exon within the gene
+ >>> check_id(pd.Series({"chr": 1, "start": 0, "stop": 50, "id": "17",
+ ... "score": 0, "strand": '-'}))
+ Traceback (most recent call last):
+ ...
+ BadIDError: The exon 17 as a bad id format.
+ It should have the form id_pos where
+ 1. id: Is the fasterDB gene id of the exons
+ 2. pos: Is the position of the exon within the gene
+ """
+ res = mserie["id"].split("_")
+ if len(res) != 2 or not res[0].isdigit() or not res[1].isdigit():
+ raise BadIDError(f"The exon {mserie['id']} as a bad id format.\n"
+ f"It should have the form id_pos where\n"
+ f" 1. id: Is the fasterDB gene id of the exons\n"
+ f" 2. pos: Is the position of the exon within the "
+ f"gene")
+
+
+def load_exon_bed(exon_bed: Path) -> pd.DataFrame:
+ """
+ Load a bed file containing exons. The columns containing the \
+ names of intervals must have the following structure: id_pos where:
+ 1. id is the fasterDB id of a gene
+ 2. pos: Is the exons position on the gene
+
+ :param exon_bed: A bed file containing exons
+ :return: The loaded exons bed file as a dataframe
+
+ >>> load_exon_bed(Config.tests_files / "exons.bed")
+ chr start stop id score strand
+ 0 18 28681865 28682388 1_1 0 -
+ 1 18 28681183 28681432 1_2 0 -
+ 2 18 28673521 28673606 1_3 0 -
+ 3 18 28672063 28672263 1_4 0 -
+ 4 18 28671489 28671530 1_5 0 -
+ 5 18 28670990 28671110 1_6 0 -
+ 6 18 28669401 28669557 1_7 0 -
+ 7 18 28667631 28667776 1_8 0 -
+ 8 18 28666538 28666705 1_9 0 -
+ """
+ df = pd.read_csv(exon_bed, sep="\t", names=["chr", "start", "stop", "id",
+ "score", "strand"])
+ df.apply(check_id, axis=1)
+ return df
+
+
+def get_first_and_last_exons(df_exon: pd.DataFrame) -> pd.DataFrame:
+ """
+ Return a dataframe indicating for each gene, it's first or last exons.
+
+ :param df_exon: A dataframe containing exons in bed format.
+ :return: dataframe indicating for each gene, it's first or last exons.
+
+ >>> d = pd.DataFrame({
+ ... 'chr': {0: 18, 1: 18, 2: 18, 3: 18, 4: 18, 5: 18, 6: 18, 7: 18, 8: 18},
+ ... 'start': {0: 28681865, 1: 28681183, 2: 28673521, 3: 28672063,
+ ... 4: 28671489, 5: 28670990, 6: 28669401, 7: 28667631,
+ ... 8: 28666538},
+ ... 'stop': {0: 28682388, 1: 28681432, 2: 28673606, 3: 28672263,
+ ... 4: 28671530, 5: 28671110, 6: 28669557, 7: 28667776,
+ ... 8: 28666705},
+ ... 'id': {0: '1_1', 1: '1_2', 2: '1_3', 3: '1_4', 4: '1_5', 5: '1_6',
+ ... 6: '1_7', 7: '1_8', 8: '1_9'},
+ ... 'score': {0: 0, 1: 0, 2: 0, 3: 0, 4: 0, 5: 0, 6: 0, 7: 0, 8: 0},
+ ... 'strand': {0: '-', 1: '-', 2: '-', 3: '-', 4: '-', 5: '-', 6: '-',
+ ... 7: '-', 8: '-'}})
+ >>> get_first_and_last_exons(d)
+ chr start_first stop_first start_last stop_last gene_id strand
+ 0 18 28681865 28682388 28666538 28666705 1 -
+ """
+ df_exon["gene_id"] = df_exon["id"].str.replace(r"_\d+", "", regex=True)\
+ .astype(int)
+ df_exon["pos"] = df_exon["id"].str.replace(r"\d+_", "", regex=True)\
+ .astype(int)
+ df_exon.sort_values(["gene_id", "pos"], ascending=True, inplace=True)
+ df_first = df_exon.drop_duplicates(subset="gene_id", keep="first")
+ df_last = df_exon.drop_duplicates(subset="gene_id", keep="last")
+ good_col = ["chr", "start_first", "stop_first", "start_last", "stop_last",
+ "gene_id", "strand"]
+ df = pd.merge(df_first, df_last, how="outer",
+ on=["gene_id", "chr", "strand", "score"],
+ suffixes=["_first", "_last"])
+ return df[df["pos_first"] != df["pos_last"]][good_col]
+
+
+def get_gene_row(row: pd.Series) -> pd.Series:
+ """
+ From a row containing the position of the first and the last exons of \
+ a gene, return another row containing the intervals of a gene without \
+ this first and last exons.
+
+ :param row: A row containing the position of the first and the \
+ last exons of a gene
+ :return: another row containing the intervals of a gene without \
+ this first and last exons.
+
+ >>> get_gene_row(pd.Series(
+ ... {'chr': 18, 'start_first': 900, 'stop_first': 1000,
+ ... 'start_last': 0, 'stop_last': 100, 'gene_id': 1,
+ ... 'strand': '-'})).to_dict() == {'chr': 18, 'start': 100,
+ ... 'stop': 900, 'id': 1, 'score': 0, 'strand': '-'}
+ True
+ >>> get_gene_row(pd.Series(
+ ... {'chr': 18, 'start_first': 0, 'stop_first': 100,
+ ... 'start_last': 900, 'stop_last': 1000, 'gene_id': 1,
+ ... 'strand': '+'})).to_dict() == {'chr': 18, 'start': 100,
+ ... 'stop': 900, 'id': 1, 'score': 0, 'strand': '+'}
+ True
+ """
+ if row["strand"] == "+":
+ return pd.Series({"chr": row["chr"],
+ "start": row["stop_first"],
+ "stop": row["start_last"],
+ "id": row["gene_id"],
+ "score": 0,
+ "strand": row["strand"]})
+ return pd.Series({"chr": row["chr"],
+ "start": row["stop_last"],
+ "stop": row["start_first"],
+ "id": row["gene_id"],
+ "score": 0,
+ "strand": row["strand"]})
+
+
+def create_new_bed(df_exons: pd.DataFrame) -> pd.DataFrame:
+ """
+ From a dataframe containing the first and the last exons of \
+ each gene create dataframe of genes without the first and the last exons.
+
+ :param df_exons: a dataframe containing the first and the last exons of \
+ each gene create
+ :return: The dataframe in bed format of genes without it's first or last \
+ exons
+
+ >>> d = pd.DataFrame({'chr': [18, 18], 'start_first': [900, 0],
+ ... 'stop_first': [1000, 100],
+ ... 'start_last': [0, 900], 'stop_last': [100, 1000], 'gene_id': [1, 1],
+ ... 'strand': ['-', '+']})
+ >>> create_new_bed(d)
+ chr start stop id score strand
+ 0 18 100 900 1 0 -
+ 1 18 100 900 1 0 +
+ """
+ df_exons = df_exons.apply(get_gene_row, axis=1)
+ df_exons["chr"] = "chr" + df_exons["chr"]
+ return df_exons
+
+
+def create_gene_bed():
+ """
+ Create a gene files without first and last exons.
+ """
+ df_exon = load_exon_bed(Config.bed_exon)
+ df_fl = get_first_and_last_exons(df_exon)
+ bed_gene = create_new_bed(df_fl)
+ ConfigEncodeClip.output_gene.parent.mkdir(exist_ok=True)
+ bed_gene.to_csv(ConfigEncodeClip.output_gene, sep="\t", index=False,
+ header=False)
+
+
+if __name__ == "__main__":
+ doctest.testmod()