From 29b46a8726fdd5ce09b8a7a29f15e5a2308037ea Mon Sep 17 00:00:00 2001
From: Bertache <samir.bertache@ens-lyon.fr>
Date: Fri, 25 Apr 2025 16:35:27 +0200
Subject: [PATCH] Add clean_parasplit_label
Don't take in account dangling end for parasplited reads input !
---
src/hiclassifier/processing.py | 104 ++++++++++++++++++++++++++++-----
1 file changed, 91 insertions(+), 13 deletions(-)
diff --git a/src/hiclassifier/processing.py b/src/hiclassifier/processing.py
index f3b9d35..90fca09 100644
--- a/src/hiclassifier/processing.py
+++ b/src/hiclassifier/processing.py
@@ -1,3 +1,11 @@
+# import logging
+# import multiprocessing
+
+# logger = multiprocessing.get_logger()
+# logger.setLevel(logging.DEBUG)
+from typing import Pattern
+
+
def get_label(current_pair, dict_digested, tolerance):
"""
Annotation spécifique Intra et Interchromosomique
@@ -88,6 +96,53 @@ def qualitad(current_pair, mapq_score_max):
return False
+def clean_parasplit_label(
+ label: str,
+ query_name: str,
+ pair_suffix_re: Pattern[str],
+) -> str:
+ """
+ Remplace « DL » par « None » dans *label* lorsque *query_name* se termine
+ par un suffixe « :ij ».
+
+ En effet, Parasplit, comme Cutsite est un outil qui coupe les lectures en
+ de la présence en site de LIGATION ! Il est donc normale de trouver un nombre
+ accru de Dangling ends. Cet fonction a pour but de corriger cela (en partie)
+
+ Examples
+ --------
+ >>> clean_parasplit_label("DL-border", "read:01")
+ 'None-border'
+ >>> clean_parasplit_label("read-DL", "read:03")
+ 'read-None'
+ >>> clean_parasplit_label("read-border", "read:03")
+ 'read-border'
+ >>> clean_parasplit_label("DL-border", "read")
+ 'DL-border'
+ """
+ # 1) Vérifier le suffixe « :ij » dans le query_name
+ if pair_suffix_re.search(query_name) is None:
+ return label
+
+ # 2) Vérifier que label est une str contenant « DL »
+ if not (isinstance(label, str) and "DL" in label):
+ return label
+
+ # 3) Séparer le label
+ try:
+ read_part, border_part = label.split("-", maxsplit=1)
+ except ValueError: # format inattendu
+ return label
+
+ # 4) Remplacer « DL » par « None » dans chaque partie
+ if "DL" in read_part:
+ read_part = "None"
+ if "DL" in border_part:
+ border_part = "None"
+
+ return f"{read_part}-{border_part}"
+
+
def process_items(
input_queue,
output_valide_queue,
@@ -112,6 +167,8 @@ def process_items(
output_queue (Queue): Queue to put processed read pairs.
"""
+ import re
+
from .auxiliary import check_data, pair_complete, terminaison
from .event_collector import (
update_restriction_counts,
@@ -119,6 +176,10 @@ def process_items(
)
from .readmanager import PairedRead
+ _DEFAULT_PAIR_SUFFIX_RE = re.compile(
+ r":(\d{2})$"
+ ) # cf Parasplit (gitbio.ens-lyon.fr)
+
loop_count = 0
# Création des dictionnaire de comptage du processus courant :
@@ -135,6 +196,17 @@ def process_items(
while True:
loop_count += 1
data = input_queue.get()
+
+ # if loop_count % 10 == 0:
+ # logger.warning(
+ # "loop=%d\tin=%s\tval=%s\toth=%s\tdel=%s",
+ # loop_count,
+ # safe_qsize(input_queue),
+ # safe_qsize(output_valide_queue),
+ # safe_qsize(output_other_queue),
+ # safe_qsize(output_del_queue),
+ # )
+
label = False
if data is None:
@@ -179,21 +251,26 @@ def process_items(
undigested, Counts_Undigested_Site
)
- if label and ("Other" not in label):
- (
- Counts_Biotin_Intra,
- Counts_Dangling_Intra,
- Counts_Biotin_Inter,
- Counts_Dangling_Inter,
- ) = update_restriction_counts(
- paired_read=current_pair,
- label=label,
- counts_vl_intra=Counts_Biotin_Intra,
- counts_dl_intra=Counts_Dangling_Intra,
- counts_vl_inter=Counts_Biotin_Inter,
- counts_dl_inter=Counts_Dangling_Inter,
+ if label:
+ label = clean_parasplit_label(
+ label, current_pair.query_name_1, _DEFAULT_PAIR_SUFFIX_RE
)
+ if isinstance(label, str) and ("Other" not in label):
+ (
+ Counts_Biotin_Intra,
+ Counts_Dangling_Intra,
+ Counts_Biotin_Inter,
+ Counts_Dangling_Inter,
+ ) = update_restriction_counts(
+ paired_read=current_pair,
+ label=label,
+ counts_vl_intra=Counts_Biotin_Intra,
+ counts_dl_intra=Counts_Dangling_Intra,
+ counts_vl_inter=Counts_Biotin_Inter,
+ counts_dl_inter=Counts_Dangling_Inter,
+ )
+
# # Send in queue
sending_to_queue(
data,
@@ -202,6 +279,7 @@ def process_items(
output_other_queue=output_other_queue,
output_del_queue=output_del_queue,
)
+
finally:
terminaison(
output_valide_queue=output_valide_queue,
--
GitLab