Some debug scripts modified / added

inference_test.py

 from nns.keras_metrics_from_logits import precision, recall, binary_accuracy
 import pandas as pd
+import random
 from pathlib import Path
 import h5py
 from inference.InferenceModel import InferenceModel
@@ -21,22 +22,33 @@ def loss_fun(y_true, y_pred):  # just dummy to satisfy the stupid thing being th
     return K.binary_crossentropy(K.cast(y_true, K.floatx()), y_pred_single,
                                  from_logits=True)
 
-def precision_recall_at_threshold(threshold, ax):
+def precision_recall_at_threshold(threshold, target_model, target_name, ax):
+    """
+
+    :param threshold: Logit threshold to use
+    :param target_model: Path to target model
+    :param target_name: Name of target species in ground_truth csv
+    :param ax: ax to plot on
+    :return:
+    """
     stride = 125
-    num_of_reads_to_check = 100
+    num_of_reads_to_check = 500
     ground_truths_list = []
     frac_of_kmers_found_list = []
-    model_path = '/lustre/BIF/nobackup/noord087/mscthesis/baseless/baseless_250_hyperparam/e_coli_from16s_compiled.tar'
-    mod = InferenceModel(model_path, threshold=threshold)
+    mod = InferenceModel(target_model, threshold=threshold)
+    print(f'{len(mod.kmers)=}')
     ground_truth_df = pd.read_csv(
         '/lustre/BIF/nobackup/noord087/HoiCarlos/16Sreads_mockcommunity/ground_truth_with_read_id_and_perc_id.csv')
     start = time.time()
     for counter, file in enumerate(
             Path('/home/noord087/lustre_link/HoiCarlos/16Sreads_mockcommunity/demultiplexed_reads/files_for_initial_training/test').iterdir()):
         file_info = ground_truth_df[ground_truth_df['file name'] == file.name]
-        read_is_target_species = 'Escherichia coli' in str(
+        read_is_target_species = target_name in str(
             file_info['species'])
         print(f"Target in read: {read_is_target_species}")
+        # if read_is_target_species and random.random() > 0.9:
+        #     # Reduce to 1/10 of what is normal
+        #     continue
         with h5py.File(file, 'r') as f:
             raw_read_split_reads = Read(f, 'median').get_split_raw_read(250,
                                                                         stride)
@@ -57,9 +69,14 @@ def precision_recall_at_threshold(threshold, ax):
 
 
 def precision_recall_wrapper():
+    # # Now doing it for rhodobacter
+    # model_path = '/lustre/BIF/nobackup/noord087/mscthesis/baseless/baseless_250_hyperparam/r_sphaeroides_from16s_compiled.tar'
+    # target_name = 'Rhodobacter sphaeroides'
+    model_path = '/lustre/BIF/nobackup/noord087/mscthesis/baseless/baseless_250_hyperparam/e_coli_from16s_compiled.tar'
+    target_name = 'Escherichia coli'
     fig, ax = plt.subplots()
-    for cutoff in np.arange(3, 7):
-        precision_recall_at_threshold(cutoff, ax=ax)
+    for cutoff in np.arange(4, 6):
+        precision_recall_at_threshold(cutoff, model_path, target_name, ax=ax)
     plt.show()
 
 

k_mer_design/kmers_in_bacterial_species_plot.py

+import pandas as pd
+from pathlib import Path
+from inference.compile_model import reverse_complement
+
+
+kmer_freqs_df = pd.read_parquet(Path(f'{__file__}/../data/ncbi_16S_bacteria_archaea_kmer_counts.parquet').resolve())
+kmer_path = '/lustre/BIF/nobackup/noord087/mscthesis/baseless/baseless_250_hyperparam/kmers_in_dbs.txt'
+ground_truth_df = pd.read_csv(
+        '/lustre/BIF/nobackup/noord087/HoiCarlos/16Sreads_mockcommunity/ground_truth_with_read_id_and_perc_id.csv')
+
+with open(kmer_path, 'r') as f:
+    kmers = f.read().splitlines()
+
+kmers.extend([reverse_complement(kmer) for kmer in kmers])
+valid_kmers = [kmer for kmer in kmers if kmer in kmer_freqs_df.columns]
+df_kmers_selected = kmer_freqs_df[valid_kmers]
+
+# TODO convert between species in ground truth and species in kmers_freq_df
\ No newline at end of file