src/preprocessing.py

@@ -11,7 +11,7 @@ from tensorflow.train import Example, Feature, Features, FloatList, Int64List
 from constants import *
 
 
-def generate_example(sequence, label, weight_matrix) -> bytes:
+def generate_example(sequence, reference_sequence, weight_matrix) -> bytes:
     """
     Create a binary-string for each sequence containing the sequence and the bases' frequency
     """
@@ -19,7 +19,9 @@ def generate_example(sequence, label, weight_matrix) -> bytes:
         "sequence": Feature(
             int64_list=Int64List(value=list(encode_sequence(sequence)))
         ),
-        "label": Feature(int64_list=Int64List(value=list(encode_sequence(label)))),
+        "reference_sequence": Feature(
+            int64_list=Int64List(value=list(encode_sequence(reference_sequence)))
+        ),
         "A_counts": Feature(float_list=FloatList(value=weight_matrix["A"])),
         "C_counts": Feature(float_list=FloatList(value=weight_matrix["C"])),
         "G_counts": Feature(float_list=FloatList(value=weight_matrix["G"])),
@@ -47,27 +49,23 @@ def read_fastq(data_file, label_file) -> List[bytes]:
             motifs = create([element.seq])
             example = generate_example(
                 sequence=str(element.seq),
-                label=str(label.seq),
+                reference_sequence=str(label.seq),
                 weight_matrix=motifs.pwm,
             )
             examples.append(example)
     return examples
 
 
-def create_dataset(data_file, label_file) -> None:
+def create_dataset(filepath) -> None:
     """
     Create a training and test dataset with a 70/30 split respectively
     """
     data = read_fastq(data_file, label_file)
-    train_eval_test_split = [0.8, 0.1, 0.1]
-    with TFRecordWriter(TRAIN_DATASET) as training, TFRecordWriter(
-        TEST_DATASET
-    ) as test, TFRecordWriter(EVAL_DATASET) as evaluation:
+    train_test_split = 0.7
+    with TFRecordWriter(TRAIN_DATASET) as train, TFRecordWriter(TEST_DATASET) as test:
         for element in data:
-            if random() < train_eval_test_split[0]:
-                training.write(element)
-            elif random() < train_eval_test_split[0] + train_eval_test_split[1]:
-                evaluation.write(element)
+            if random() < train_test_split:
+                train.write(element)
             else:
                 test.write(element)
 
@@ -78,7 +76,7 @@ def process_input(byte_string) -> Example:
     """
     schema = {
         "sequence": FixedLenFeature(shape=[], dtype=int64),
-        "label": FixedLenFeature(shape=[], dtype=int64),
+        "reference_sequence": FixedLenFeature(shape=[], dtype=int64),
         "A_counts": FixedLenFeature(shape=[], dtype=float32),
         "C_counts": FixedLenFeature(shape=[], dtype=float32),
         "G_counts": FixedLenFeature(shape=[], dtype=float32),
@@ -98,11 +96,8 @@ def read_dataset(filepath) -> TFRecordDataset:
     return batched_dataset
 
 
-def dataset_creation(
-    data_file, label_file
-) -> Tuple[TFRecordDataset, TFRecordDataset, TFRecordDataset]:
+def dataset_creation(data_file, label_file) -> Tuple[TFRecordDataset, TFRecordDataset]:
     create_dataset(data_file, label_file)
     train_data = read_dataset(TRAIN_DATASET)
-    eval_data = read_dataset(EVAL_DATASET)
     test_data = read_dataset(TEST_DATASET)
-    return train_data, eval_data, test_data
+    return train_data, test_data