diff --git a/crisprbact/predict.py b/crisprbact/predict.py
index 0c395ebefe923bf5f408259311d96f355ea2d55b..3a10c6f7e6a40fe73b8a9b31032b1bda7d4c7d7f 100644
--- a/crisprbact/predict.py
+++ b/crisprbact/predict.py
@@ -48,7 +48,6 @@ def find_targets(seq):
yield dict(
[
("target", matching_target),
- # ("guide", matching_target[:20]),
("guide", guide),
("start", pos_seq_start),
("stop", pos_seq_stop),
@@ -79,13 +78,9 @@ def on_target_predict(seq, genome=None, seed_sizes=[8, 9, 10, 11, 12, GUIDE_LEN]
genome_features = extract_features(records)
alltargets = list(find_targets(seq))
if alltargets:
- X = np.array(
- [
- encode(target["target"][:7] + target["target"][9:])
- for target in alltargets
- ] # encode and remove GG of PAM
- )
- X = X.reshape(X.shape[0], -1)
+ gen_targets = (target["target"] for target in alltargets)
+ gen_clean_targets = remove_GG_of_PAM(gen_targets)
+ X = reshape_targets(gen_clean_targets)
preds = predict(X)
for i, target in enumerate(alltargets):
target_id = i + 1
@@ -100,20 +95,8 @@ def on_target_predict(seq, genome=None, seed_sizes=[8, 9, 10, 11, 12, GUIDE_LEN]
)
off_targets_list = []
if off_target_df is not None and not off_target_df.empty:
- off_targets = off_target_df.loc[
- 0:,
- [
- "start",
- "end",
- "pampos",
- "strand",
- "recid",
- "longest_perfect_match",
- "pam_seq",
- "features",
- ],
- ]
- for j, off_t in enumerate(off_targets.values.tolist()):
+ off_targets = slice_off_targets_results(off_target_df)
+ for j, off_t in enumerate(off_targets):
off_target_dict = {
"off_target_id": str(target_id)
+ "-"
@@ -129,38 +112,28 @@ def on_target_predict(seq, genome=None, seed_sizes=[8, 9, 10, 11, 12, GUIDE_LEN]
"off_target_pam_seq": off_t[6],
"off_target_good_orientation": None,
}
- index_features = 7
- # Loop through features associated to an off-target position
- if len(off_t[index_features]) > 0:
- # Loop for each feature
- for feat in off_t[index_features]:
- if get_strand_value(
- off_target_dict["off_target_strand"]
- ) != get_strand_value(feat.location.strand):
- off_target_dict[
- "off_target_good_orientation"
- ] = True
- else:
- off_target_dict[
- "off_target_good_orientation"
- ] = False
-
- feature_dict = {
- "off_target_feat_strand": feat.location.strand,
- "off_target_feat_start": feat.location.start,
- "off_target_feat_end": feat.location.end,
- "off_target_feat_type": feat.type,
- }
- for k, feat in feat.qualifiers.items():
- if k != "translation":
- feature_dict["off_target_" + k] = "::".join(
- feat
- )
+ if (
+ seed_size == GUIDE_LEN
+ or off_target_dict["off_target_longest_perfect_match"]
+ != GUIDE_LEN
+ ):
+ index_features = 7
+ # Get feature details associated
+ # to an off-target position
+ if len(off_t[index_features]) > 0:
+ feat = off_t[index_features][0]
+ off_target_dict[
+ "off_target_good_orientation"
+ ] = is_good_orientation(
+ feat, off_target_dict["off_target_strand"]
+ )
+ off_target_feature = get_off_target_feature(feat)
off_targets_list.append(
- {**feature_dict, **off_target_dict}
+ {**off_target_feature, **off_target_dict}
)
- else:
- off_targets_list.append(off_target_dict)
+ else:
+ off_targets_list.append(off_target_dict)
+
off_targets_per_seed.append(
{
"id": str(i) + "-" + str(seed_size),
@@ -182,3 +155,53 @@ def on_target_predict(seq, genome=None, seed_sizes=[8, 9, 10, 11, 12, GUIDE_LEN]
return alltargets
else:
return []
+
+
+def remove_GG_of_PAM(targets):
+ for target in targets:
+ yield target[:7] + target[9:]
+
+
+def reshape_targets(targets):
+ X = np.array([encode(target) for target in targets]) # encode and remove GG of PAM
+ return X.reshape(X.shape[0], -1)
+
+
+def slice_off_targets_results(off_target_df):
+ return off_target_df.loc[
+ 0:,
+ [
+ "start",
+ "end",
+ "pampos",
+ "strand",
+ "recid",
+ "longest_perfect_match",
+ "pam_seq",
+ "features",
+ ],
+ ].values.tolist()
+
+
+def is_good_orientation(feat, off_target_strand):
+ if get_strand_value(off_target_strand) != get_strand_value(feat.location.strand):
+ return True
+ else:
+ return False
+
+
+def get_off_target_feature(feat):
+
+ feature_dict = {
+ "off_target_feat_strand": feat.location.strand,
+ "off_target_feat_start": feat.location.start,
+ "off_target_feat_end": feat.location.end,
+ "off_target_feat_type": feat.type,
+ }
+ for k, feat in feat.qualifiers.items():
+ if k != "translation":
+ feature_dict["off_target_" + k] = "::".join(feat)
+ return feature_dict
+
+
+# def gen_off_target_per_seed_size(seed_sizes):