### Yet Another Numerical Output Example

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/optimize_sampling_algorithm_settings.ipynb

This snippet shows a further numerical output, continuing the pattern of performance metrics being logged.

```text
0.8111033333364384
```

--------------------------------

### Another Numerical Output Example

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/optimize_sampling_algorithm_settings.ipynb

This snippet displays another numerical output, potentially a different metric or a result from a varied configuration.

```text
0.7509776432384816
```

--------------------------------

### Download and Unzip Data

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/case_studies_ms2deepscore_2/Blood_case_study/Create_molecular_networks.ipynb

Downloads necessary files from Zenodo and unzips the case study data. Ensure the 'requests', 'tqdm', and 'zipfile' libraries are installed.

```python
import requests
from tqdm import tqdm
import zipfile
import os

def unzip_file(zip_path, extract_to="."):
    """Unzips a zip file to the specified directory."""
    with zipfile.ZipFile(zip_path, 'r') as zip_ref:
        zip_ref.extractall(extract_to)
        print(f"Unzipped {zip_path} to {extract_to}")

def download_file(link, file_name):
    response = requests.get(link, stream=True)
    if os.path.exists(file_name):
        print(f"The file {file_name} already exists, the file won't be downloaded")
        return
    total_size = int(response.headers.get('content-length', 0))
    
    with open(file_name, "wb") as f, tqdm(desc="Downloading file", total=total_size, unit='B', unit_scale=True, unit_divisor=1024,) as bar:
        for chunk in response.iter_content(chunk_size=1024):
            if chunk:
                f.write(chunk)
                bar.update(len(chunk))  # Update progress bar by the chunk size
    
model_file_name = "ms2deepscore_model.pt"
case_study_spectra_folder = "Blood_case_study"

download_file("https://zenodo.org/records/14290920/files/settings.json?download=1", "ms2deepscore_settings.json")
download_file("https://zenodo.org/records/14290920/files/ms2deepscore_model.pt?download=1", model_file_name)
download_file("https://zenodo.org/records/16311735/files/Blood plasma case study.zip?download=1", "Blood_plasma_case_study.zip")
unzip_file("Blood_plasma_case_study.zip", extract_to="Blood_case_study")
```

--------------------------------

### Numerical Output Example

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/optimize_sampling_algorithm_settings.ipynb

This snippet shows a numerical output, likely a performance score or metric, obtained from a model benchmarking process.

```text
0.6714763613634331
```

--------------------------------

### Install MatchMS and MS2DeepScore via Conda

Source: https://github.com/matchms/ms2deepscore/blob/main/README.md

Install matchms from bioconda and conda-forge channels, followed by MS2DeepScore via pip.

```bash
conda create --name ms2deepscore python=3.12
conda activate ms2deepscore
conda install --channel bioconda --channel conda-forge matchms
pip install ms2deepscore
```

--------------------------------

### Install MS2DeepScore via Pip

Source: https://github.com/matchms/ms2deepscore/blob/main/README.md

Install the MS2DeepScore package using pip within your activated environment.

```bash
pip install ms2deepscore
```

--------------------------------

### Figure Output Example

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/optimize_sampling_algorithm_settings.ipynb

This snippet represents the output of a figure, typically a plot or visualization generated during the benchmarking process.

```text
<Figure size 640x480 with 1 Axes>
```

--------------------------------

### Load Data and Predictions for Comparison

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/compare_with_cosine.ipynb

Loads true values, validation spectra, and pre-calculated cosine and MS2DeepScore predictions from specified file paths. This setup is crucial for subsequent performance evaluation.

```python
import os
from ms2deepscore.MS2DeepScore import MS2DeepScore
from ms2deepscore.models.load_model import load_model
from matchms.importing.load_from_mgf import load_from_mgf
from ms2deepscore.utils import load_pickled_file

data_folder = os.path.join("../../../data/pytorch/gnps_21_08_23_min_5_at_5_percent/")
both_mode_folder = "trained_models/both_mode_precursor_mz_2000_2000_2000_layers_500_embedding_2024_01_31_11_51_10/"

# True values
# pos_true_values = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "positive_positive_true_values.pickle"))
# neg_true_values = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "negative_negative_true_values.pickle"))
neg_pos_true_values = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "negative_positive_true_values.pickle"))

# Validation spectra
pos_spectra = list(load_from_mgf(os.path.join(data_folder, "training_and_validation_split", "positive_validation_spectra.mgf")))
neg_spectra = list(load_from_mgf(os.path.join(data_folder, "training_and_validation_split", "negative_validation_spectra.mgf")))

# Cosine predictions
# neg_cosine_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "cosine_scores", "negative_scores.pickle"))["score"]
# pos_cosine_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "cosine_scores", "positive_scores.pickle"))["score"]
neg_pos_cosines_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "cosine_scores", "negative_positive_scores.pickle"))["score"]

# Mod Cosine predictions
# neg_mod_cosine_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "modified_cosine_scores", "negative_scores.pickle"))["score"]
# pos_mod_cosine_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "modified_cosine_scores", "positive_scores.pickle"))["score"]
neg_pos_mod_cosines_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "cosine_scores", "negative_positive_scores.pickle"))["score"]

# Both models predictions
# pos_predictions_both_mode = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "positive_positive_predictions.pickle"))
# neg_predictions_both_mode = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "negative_negative_predictions.pickle"))
neg_pos_predictions_both_mode = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "negative_positive_predictions.pickle"))

# Pos mode predictions
pos_mode_folder = "trained_models/positive_mode_precursor_mz_2000_2000_2000_layers_500_embedding_2024_02_07_10_27_04/"
# pos_predictions_pos_mode = load_pickled_file(os.path.join(data_folder, pos_mode_folder, "benchmarking_results", "positive_positive_predictions.pickle"))
# neg_predictions_pos_mode = load_pickled_file(os.path.join(data_folder, pos_mode_folder, "benchmarking_results", "negative_negative_predictions.pickle"))
neg_pos_predictions_pos_mode = load_pickled_file(os.path.join(data_folder, pos_mode_folder, "benchmarking_results", "negative_positive_predictions.pickle"))

# neg mode predictions
neg_mode_folder = "trained_models/negative_mode_precursor_mz_2000_2000_2000_layers_500_embedding_2024_02_07_11_53_37/"
# pos_predictions_neg_mode = load_pickled_file(os.path.join(data_folder, neg_mode_folder, "benchmarking_results", "positive_positive_predictions.pickle"))
# neg_predictions_neg_mode = load_pickled_file(os.path.join(data_folder, neg_mode_folder, "benchmarking_results", "negative_negative_predictions.pickle"))
neg_pos_predictions_neg_mode = load_pickled_file(os.path.join(data_folder, neg_mode_folder, "benchmarking_results", "negative_positive_predictions.pickle"))
```

--------------------------------

### Initialize Scoring Objects

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/Compare balanced cross ion mode sampling.ipynb

Initializes CalculateScoresBetweenAllIonmodes objects for both a normal and a balanced model. This setup is crucial for comparing scoring performance.

```python
from ms2deepscore.benchmarking.CalculateScoresBetweenAllIonmodes import CalculateScoresBetweenAllIonmodes
scores_normal_model = CalculateScoresBetweenAllIonmodes(normal_model_file_name, pos_test, neg_test, fingerprint_type="daylight", n_bits_fingerprint=4096)
scores_balanced_model = CalculateScoresBetweenAllIonmodes(balanced_model_file_name, pos_test, neg_test, fingerprint_type="daylight", n_bits_fingerprint=4096)
```

--------------------------------

### Load Data and Predictions for Comparison

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/compare_with_cosine.ipynb

Loads true values, validation spectra, and pre-calculated cosine and MS2DeepScore predictions from pickle and mgf files. This setup is essential for evaluating different scoring methods.

```python
import os
from ms2deepscore.MS2DeepScore import MS2DeepScore
from ms2deepscore.models.load_model import load_model
from matchms.importing.load_from_mgf import load_from_mgf
from ms2deepscore.utils import load_pickled_file

data_folder = os.path.join("../../../data/pytorch/gnps_21_08_23_min_5_at_5_percent/")
both_mode_folder = "trained_models/both_mode_precursor_mz_ionmode_2000_2000_2000_layers_500_embedding_2024_01_31_11_51_10/"

# True values
pos_true_values = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "positive_positive_true_values.pickle"))
# neg_true_values = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "negative_negative_true_values.pickle"))
# neg_pos_true_values = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "negative_positive_true_values.pickle"))

# Validation spectra
pos_spectra = list(load_from_mgf(os.path.join(data_folder, "training_and_validation_split", "positive_validation_spectra.mgf")))
# neg_spectra = list(load_from_mgf(os.path.join(data_folder, "training_and_validation_split", "negative_validation_spectra.mgf")))

# Cosine predictions
# neg_cosine_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "cosine_scores", "negative_scores.pickle"))["score"]
pos_cosine_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "cosine_scores", "positive_scores.pickle"))["score"]
# neg_pos_cosines_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "cosine_scores", "negative_positive_scores.pickle"))["score"]

# Mod Cosine predictions
# neg_mod_cosine_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "modified_cosine_scores", "negative_scores.pickle"))["score"]
pos_mod_cosine_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "modified_cosine_scores", "positive_scores.pickle"))["score"]
# neg_pos_mod_cosines_predictions = load_pickled_file(os.path.join(data_folder, "training_and_validation_split", "cosine_scores", "negative_positive_scores.pickle"))["score"]

# Both models predictions
pos_predictions_both_mode = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "positive_positive_predictions.pickle"))
# neg_predictions_both_mode = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "negative_negative_predictions.pickle"))
# neg_pos_predictions_both_mode = load_pickled_file(os.path.join(data_folder, both_mode_folder, "benchmarking_results", "negative_positive_predictions.pickle"))

# Pos mode predictions
pos_mode_folder = "trained_models/positive_mode_precursor_mz_2000_2000_2000_layers_500_embedding_2024_02_07_10_27_04/"
pos_predictions_pos_mode = load_pickled_file(os.path.join(data_folder, pos_mode_folder, "benchmarking_results", "positive_positive_predictions.pickle"))
# neg_predictions_pos_mode = load_pickled_file(os.path.join(data_folder, pos_mode_folder, "benchmarking_results", "negative_negative_predictions.pickle"))
# neg_pos_predictions_pos_mode = load_pickled_file(os.path.join(data_folder, pos_mode_folder, "benchmarking_results", "negative_positive_predictions.pickle"))

# neg mode predictions
neg_mode_folder = "trained_models/negative_mode_precursor_mz_2000_2000_2000_layers_500_embedding_2024_02_07_11_53_37/"
pos_predictions_neg_mode = load_pickled_file(os.path.join(data_folder, neg_mode_folder, "benchmarking_results", "positive_positive_predictions.pickle"))
# neg_predictions_neg_mode = load_pickled_file(os.path.join(data_folder, neg_mode_folder, "benchmarking_results", "negative_negative_predictions.pickle"))
# neg_pos_predictions_neg_mode = load_pickled_file(os.path.join(data_folder, neg_mode_folder, "benchmarking_results", "negative_positive_predictions.pickle"))

```

--------------------------------

### Download and Unzip Spectral Data

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/case_studies_ms2deepscore_2/Rumex_case_study/case_study_mol_network_Rumex.ipynb

Downloads necessary model files, settings, and spectral data for the Rumex case study. It then unzips the spectral data to the specified directory. Ensure 'requests' and 'tqdm' are installed.

```python
import os
import zipfile

def unzip_file(zip_path, extract_to="."):
    """Unzips a zip file to the specified directory."""
    with zipfile.ZipFile(zip_path, 'r') as zip_ref:
        zip_ref.extractall(extract_to)
        print(f"Unzipped {zip_path} to {extract_to}")

def download_file(link, file_name):
    response = requests.get(link, stream=True)
    if os.path.exists(file_name):
        print(f"The file {file_name} already exists, the file won't be downloaded")
        return
    total_size = int(response.headers.get('content-length', 0))

    with open(file_name, "wb") as f, tqdm(desc="Downloading file", total=total_size, unit='B', unit_scale=True, unit_divisor=1024,) as bar:
        for chunk in response.iter_content(chunk_size=1024):
            if chunk:
                f.write(chunk)
                bar.update(len(chunk))  # Update progress bar by the chunk size

model_file_name = "ms2deepscore_model.pt"
case_study_spectra_folder = ""

download_file("https://zenodo.org/records/14290920/files/settings.json?download=1", "ms2deepscore_settings.json")
download_file("https://zenodo.org/records/14290920/files/ms2deepscore_model.pt?download=1", model_file_name)
download_file("https://zenodo.org/records/16311735/files/Rumex_case_study.zip?download=1", "Rumex_case_study.zip")
unzip_file("Rumex_case_study.zip", extract_to=case_study_spectra_folder)
```

--------------------------------

### Plotting Negative Prediction Distribution

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/positive_vs_negative_examples.ipynb

Generates a histogram to visualize the distribution of predictions for negative examples within the combined dataset. Requires matplotlib.

```python
plt.hist(neg_in_combined_predictions)
```

--------------------------------

### Search Progress Indicator

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/positive_vs_negative_examples.ipynb

This output indicates that a search operation has started and is currently at 0 iterations, with no time elapsed. It suggests the search process has just begun.

```text
0it [00:00, ?it/s]
```

--------------------------------

### Example SMILES String and Associated Data

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/positive_vs_negative_examples.ipynb

Displays a SMILES string representing a molecule, its classification (positive/negative), and associated numerical identifiers and precursor ions.

```text
CC1(C)CC[C@]2(C(=O)O[C@@H]3O[C@H](CO)[C@@H](O)[C@H](O)[C@H]3O)CC[C@]3(C)C(=CCC4[C@@]5(C)C[C@H](O)[C@H](O[C@@H]6O[C@H](C(=O)O)[C@@H](O)[C@H](O)[C@H]6O)[C@](C)(C=O)C5CC[C@]43C)[C@@H]2C1
positive
negative
1486
159035
[M-H]-
[M+K]+
```

```text
C[C@@H](C1=C[C@@]2(C=C[C@H]3[C@@H]2[C@@H](OC=C3C(=O)OC)O[C@H]4[C@@H]([C@H]([C@@H]([C@H](O4)CO)O)O)O)OC1=O)O
positive
negative
1546
164206
[M-H]-
[M+H]+
```

```text
C[C@@H]1O[C@@H](OC[C@H]2O[C@@H](Oc3c(-c4ccc(O)c(O)c4)oc4cc(O)cc(O)c4c3=O)[C@H](O)[C@@H](O)[C@@H]2O)[C@H](O)[C@H](O)[C@H]1O
negative
positive
7181
201967
[M+H]+
[M-H]-
```

```text
CC(C)(O)CCC(=O)C(C)(O)C1C(O)CC2(C)C3CC=C4C(CC(O)C(OC5OC(CO)C(O)C(O)C5O)C4(C)C)C3(C)C(=O)CC12C
negative
positive
429
1186
[M+NH4]+
[M+FA-H]-
```

```text
CC(C)CC(=O)OC[C@@]1(O)C[C@@]23CC[C@H]4[C@@](C)(CCC[C@@]4(C)C(=O)O)[C@@H]2CC[C@@H]1C3
negative
positive
551
28593
[M-H2O+H]+
[M-H]-
```

```text
C[C@H]1C/C=C/[C@H]2[C@@H]3O[C@]3(C)[C@@H](C)[C@H]4[C@H](CC5=CC=CC=C5)NC(=O)[C@@]24OC(=O)O/C=C/[C@@](C)(O)C1=O
negative
positive
599
1930
[M+H]+
[M+FA-H]-
```

```text
CCCCCC(O)CC(O)CC(=O)OC(CCCCC)CC(O)CC(=O)OC(CCCCC)CC(O)CC(=O)OC(CCCCC)CC(O)CC(O)=O
negative
positive
639
1534
[M-H2O+H]+
[M-H]-
```

```text
CCCCCC(=O)OC(CC(=O)[O-])C[N+](C)(C)C
negative
positive
1410
7422
[M+H]+
[M-H]-
```

```text
CCCCCCCCCC(=O)OC(CC(=O)[O-])C[N+](C)(C)C
negative
positive
1411
30856
[M+H]+
[M-H]-
```

```text
C[C@H]1C/C=C/[C@H]2[C@@H]3O[C@]3(C)[C@@H](C)[C@H]4[C@H](CC5=CC=CC=C5)NC(=O)[C@@]24OC(=O)O/C=C/[C@@](C)(O)C1=O
negative
positive
1487
1930
[M-H2O+H]+
[M+FA-H]-
```

```text
COC(=O)CCC(C)C1CCC2C3CC[C@@H]4C[C@H](O)CC[C@]4(C)C3C[C@H](O)[C@]12C
negative
positive
1492
24881
[M-H2O+H]+
[M-H]-
```

```text
COc1ccc(-c2oc3cc(O)cc(O)c3c(=O)c2O)cc1O
negative
positive
1584
11203
[M+H]+
[M-H]-
```

--------------------------------

### Setup for Plotting Comparison Violinplots

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/Comparison_to_single_ion_mode_models.ipynb

Initializes a figure and axes for creating violin plots to compare score distributions across multiple panels. It sets up shared x-axes and specific height ratios for subplots.

```python
from scipy.stats import gaussian_kde
import matplotlib.pyplot as plt
import numpy as np
from ms2deepscore.utils import create_evenly_spaced_bins
from matplotlib import pyplot as plt

from ms2deepscore.benchmarking.CalculateScoresBetweenAllIonmodes import CalculateScoresBetweenAllIonmodes
from ms2deepscore.utils import create_evenly_spaced_bins
import pandas as pd
def get_predictions_per_bin(predictions_and_tanimoto_scores,
                            average_per_inchikey_pair: pd.DataFrame,
                            tanimoto_bins: np.ndarray):
    """Compute average loss per Tanimoto score bin

    Parameters
    ----------
    average_per_inchikey_pair
        Precalculated average (prediction or loss) per inchikey pair
    ref_score_bins
        Bins for the reference score to evaluate the performance of scores. in the form [(0.0, 0.1), (0.1, 0.2) ...]
    """
    average_predictions = average_per_inchikey_pair.to_numpy()

    sorted_bins = sorted(tanimoto_bins, key=lambda b: b[0])

    bins = [bin_pair[0] for bin_pair in sorted_bins]
    bins.append(sorted_bins[-1][1])

    digitized = np.digitize(predictions_and_tanimoto_scores.tanimoto_df, bins, right=True)
    predictions_per_bin = []
    for i, bin_edges in tqdm(enumerate(sorted_bins), desc="Selecting available inchikey pairs per bin"):
        row_idxs, col_idxs = np.where(digitized == i+ 1)
        predictions_in_this_bin = average_predictions[row_idxs, col_idxs]
        predictions_in_this_bin_not_nan = predictions_in_this_bin[~np.isnan(predictions_in_this_bin)]
        predictions_per_bin.append(predictions_in_this_bin_not_nan)
    return predictions_per_bin

def plot_comparison_violinplot_three_panels(
    list_a,
    list_b,
    nr_of_bins: int
):
    bins = create_evenly_spaced_bins(nr_of_bins)
    bin_labels = [f"{a:.1f}–<{b:.1f}" for (a, b) in bins]

    n_panels = 3
    assert len(list_a) == n_panels and len(list_b) == n_panels, f"Expected {n_panels} sets of scores in each input"

    fig, axes = plt.subplots(
        2, n_panels, figsize=(5 * n_panels, 8),
        sharex='col',
        gridspec_kw={'height_ratios': [1, 4]},
        constrained_layout=True
    )
```

--------------------------------

### Performing Library Search for Positive/Negative Examples

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/positive_vs_negative_examples.ipynb

This function searches a library using provided embeddings and spectra. It's used here to compare different combinations of training and validation data.

```python
library_search(combined_training_embeddings, combined_training_spectra, neg_validation_embeddings, neg_validation_spectra)
```

```python
library_search(combined_training_embeddings, combined_training_spectra, pos_validation_embeddings, pos_validation_spectra)
```

```python
library_search(neg_training_embeddings, neg_training_spectra, pos_validation_embeddings, pos_validation_spectra)
```

--------------------------------

### Get Tanimoto Score Between Spectra Pairs

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/positive_vs_negative_examples.ipynb

Calculates the Tanimoto score between pairs of spectra using pre-computed embeddings. Requires the `get_tanimoto_score_between_spectra` function from `ms2deepscore.benchmarking.calculate_scores_for_validation`.

```python
from ms2deepscore.benchmarking.calculate_scores_for_validation import get_tanimoto_score_between_spectra

def get_pairs_tanimoto_score(spectrum_pairs):
    tanimoto_scores = []
    ref_spectra = [spectra[0] for spectra in spectrum_pairs]
    query_spectra = [spectra[1] for spectra in spectrum_pairs]
    tanimoto_scores = get_tanimoto_score_between_spectra(ref_spectra, query_spectra)
    return np.diagonal(tanimoto_scores)
```

--------------------------------

### Perform Library Search and Get Predictions

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/positive_vs_negative_examples.ipynb

Executes the library search function with combined training embeddings and positive validation embeddings to obtain prediction indices and scores.

```python
pos_in_combined_idx, pos_in_combined_predictions = library_search(combined_training_embeddings, pos_validation_embeddings)
```

--------------------------------

### Calculate Tanimoto Scores for Other Pairs

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/positive_vs_negative_examples.ipynb

Use this function to compute Tanimoto scores for a given set of spectral pairs. This is useful for evaluating similarity between spectra, especially when dealing with negative examples or dissimilar compounds.

```python
tanimoto_scores_other_pairs = get_pairs_tanimoto_score(other_pairs)
```

--------------------------------

### Get Unique InChIKeys Function

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/positive_vs_negative_examples.ipynb

This Python function `get_unique_inchikeys` extracts the first 14 characters of each InChIKey from a list of spectra and returns a set of unique InChIKeys. It's used to process spectral data for analysis.

```python
def get_unique_inchikeys(spectra):
    inchikey_list = []
    for spectrum in spectra:
        inchikey = spectrum.get("inchikey")[:14]
        inchikey_list.append(inchikey)
    return set(inchikey_list)

inchikeys_pos_val = get_unique_inchikeys(pos_validation_spectra)
inchikeys_neg_val = get_unique_inchikeys(neg_validation_spectra)
inchikeys_pos_train = get_unique_inchikeys(pos_training_spectra)
inchikeys_neg_train = get_unique_inchikeys(neg_training_spectra)
```

--------------------------------

### Print Hello World

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/Training_spectra_filtering.ipynb

A basic Python script to print 'hello'.

```python
print("hello")
```

--------------------------------

### Load Spectra from MGF

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/Mean_prediction_for_exact_matches.ipynb

Loads spectral data from a MGF file. This is used to get the list of validation spectra.

```python
from matchms.importing.load_from_mgf import load_from_mgf
val_spectra = list(load_from_mgf(os.path.join(results_dir, "./positive_validation_spectra.mgf")))
print(len(val_spectra))
```

--------------------------------

### Download Model, Spectra, and Annotations

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/case_studies_ms2deepscore_2/Urine_case_study/visualize_embedding_umap.ipynb

Downloads necessary files for the case study, including the MS2DeepScore model, spectra, and annotations. It checks if files already exist to avoid re-downloading.

```python
import requests
import os
from tqdm import tqdm

def download_file(link, file_name):
    response = requests.get(link, stream=True)
    if os.path.exists(file_name):
        print(f"The file {file_name} already exists, the file won't be downloaded")
        return
    total_size = int(response.headers.get('content-length', 0))
    
    with open(file_name, "wb") as f, tqdm(desc="Downloading file", total=total_size, unit='B', unit_scale=True, unit_divisor=1024,) as bar:
        for chunk in response.iter_content(chunk_size=1024):
            if chunk:
                f.write(chunk)
                bar.update(len(chunk))  # Update progress bar by the chunk size
    
model_file_name = "ms2deepscore_model.pt"
case_study_spectra_file_name = "case_study_spectra.mgf"
ms2query_annotations = "ms2query_annotations.csv"

download_file("https://zenodo.org/records/14290920/files/settings.json?download=1", "ms2deepscore_settings.json")
download_file("https://zenodo.org/records/14290920/files/ms2deepscore_model.pt?download=1", model_file_name)
download_file("https://zenodo.org/records/14535374/files/cleaned_spectra_pos_neg_with_numbering.mgf?download=1", case_study_spectra_file_name)
download_file("https://zenodo.org/records/14535374/files/ms2query_annotations.csv?download=1", ms2query_annotations)
```

--------------------------------

### Download Data Files

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/case_studies_ms2deepscore_2/Urine_case_study/Create_molecular_networks.ipynb

Downloads essential files for the case study, including the MS2DeepScore model and spectral data, from Zenodo. It checks if files already exist to avoid redundant downloads.

```python
import requests
import os
from tqdm import tqdm

def download_file(link, file_name):
    response = requests.get(link, stream=True)
    if os.path.exists(file_name):
        print(f"The file {file_name} already exists, the file won't be downloaded")
        return
    total_size = int(response.headers.get('content-length', 0))
    
    with open(file_name, "wb") as f, tqdm(desc="Downloading file", total=total_size, unit='B', unit_scale=True, unit_divisor=1024,) as bar:
        for chunk in response.iter_content(chunk_size=1024):
            if chunk:
                f.write(chunk)
                bar.update(len(chunk))  # Update progress bar by the chunk size
    
model_file_name = "ms2deepscore_model.pt"
case_study_spectra_file_name = "case_study_spectra.mgf"

download_file("https://zenodo.org/records/14290920/files/settings.json?download=1", "ms2deepscore_settings.json")
download_file("https://zenodo.org/records/14290920/files/ms2deepscore_model.pt?download=1", model_file_name)
download_file("https://zenodo.org/records/14535374/files/cleaned_spectra_pos_neg_with_numbering.mgf?download=1", case_study_spectra_file_name)
```

--------------------------------

### Create and Activate Conda Environment

Source: https://github.com/matchms/ms2deepscore/blob/main/README.md

Use this snippet to create a new Anaconda environment with Python 3.12 and activate it for MS2DeepScore installation.

```bash
conda create --name ms2deepscore python=3.12
conda activate ms2deepscore
```

--------------------------------

### Load Spectra and Models for Benchmarking

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/EmbeddingEvaluator_benchmarking.ipynb

Loads positive and negative test spectra from MGF files and initializes the MS2DeepScore model and embedding evaluator. Ensure the necessary model and spectra files are downloaded prior to execution.

```python
import os
from ms2deepscore.MS2DeepScore import MS2DeepScore
from ms2deepscore.models.load_model import load_model
from matchms.importing.load_from_mgf import load_from_mgf
from ms2deepscore.utils import load_pickled_file
from ms2deepscore.models.load_model import load_model, load_embedding_evaluator

# Validation spectra
pos_test_spectra = list(tqdm(load_from_mgf(pos_test_spectra_file_name)))
neg_test_spectra = list(tqdm(load_from_mgf(neg_test_spectra_file_name)))

ms2ds_model = load_model(model_file_name)
ms2ds_embedding_evaluator = load_embedding_evaluator(embedding_model)

```

--------------------------------

### Load Case Study Spectra from MGF

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/case_studies_ms2deepscore_2/Blood_case_study/Create_molecular_networks.ipynb

Loads the previously saved MGF file containing the case study spectra. Requires the 'matchms' library.

```python
from matchms.importing.load_from_mgf import load_from_mgf
import os
all_spectra = list(load_from_mgf(os.path.join(case_study_spectra_folder, "Blood plasma case study", "blood_case_study_spectra.mgf")))
```

--------------------------------

### Load Spectra from MGF Files

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/Modified_cosine_heatmap.ipynb

Loads positive and negative test spectra from MGF files using match.ms. It utilizes tqdm for progress visualization during loading.

```python
from matchms.importing import load_from_mgf
from tqdm import tqdm
import os

pos_test = list(tqdm(load_from_mgf(pos_test_spectra_file_name)))
neg_test = list(tqdm(load_from_mgf(neg_test_spectra_file_name)))
```

--------------------------------

### Importing Libraries

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/Compare_newer_with_older_gnps_spectra.ipynb

Imports necessary libraries for data manipulation and spectral analysis.

```python
import pandas as pd
import numpy as np
```

--------------------------------

### Assigning Labels to Score Objects

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/Comparison_to_single_ion_mode_models.ipynb

Assigns descriptive labels to score objects, likely for use in plot legends. This is a setup step before plotting.

```python
scores_neg_model.neg_vs_neg_scores.label = "negative single ion mode model"
neg_neg_scores_neg_model.neg_vs_neg_scores.label = "dual ion mode model"
```

--------------------------------

### Download Spectral Data

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/optimize_sampling_algorithm_settings.ipynb

Downloads negative and positive training spectra files in .mgf format if they do not already exist. Uses requests for downloading and tqdm for progress indication.

```python
import requests
import os
from tqdm import tqdm

def download_file(link, file_name):
    response = requests.get(link, stream=True)
    if os.path.exists(file_name):
        print(f"The file {file_name} already exists, the file won't be downloaded")
        return
    total_size = int(response.headers.get('content-length', 0))
    
    with open(file_name, "wb") as f, tqdm(desc="Downloading file", total=total_size, unit='B', unit_scale=True, unit_divisor=1024,) as bar:
        for chunk in response.iter_content(chunk_size=1024):
            if chunk:
                f.write(chunk)
                bar.update(len(chunk))  # Update progress bar by the chunk size
    
neg_train_spectra_file_name = "neg_training_spectra.mgf"
pos_train_spectra_file_name = "pos_training_spectra.mgf"

download_file("https://zenodo.org/records/13934470/files/negative_training_spectra.mgf?download=1", neg_train_spectra_file_name)
download_file("https://zenodo.org/records/13934470/files/positive_training_spectra.mgf?download=1", pos_train_spectra_file_name)



```

--------------------------------

### Create MS2DeepScore Workflow

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/case_studies_ms2deepscore_2/Blood_case_study/Create_molecular_networks.ipynb

Sets up a workflow for computing MS2DeepScore similarity scores. Requires pre-defined model and input MGF file. Logging level can be adjusted.

```python
from matchms.Pipeline import Pipeline, create_workflow
from ms2deepscore import MS2DeepScore

workflow = create_workflow(
    query_filters=[["require_minimum_number_of_peaks", {"n_required": 3}]]
    score_computations=[
        [MS2DeepScore, {"model": model}],
        ],
)
pipeline = Pipeline(workflow)
pipeline.logging_level = "ERROR"  # To define the verbosety of the logging
report = pipeline.run("blood_case_study_spectra.mgf")
```

--------------------------------

### Get shape of a NumPy array

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/compare_with_cosine.ipynb

This code snippet retrieves the dimensions (shape) of a NumPy array, commonly used to understand the size of data matrices.

```python
combined_true_values.shape
```

```python
Result:
(4130, 1009)
```

--------------------------------

### Filter Spectra by InchIKey Prefix

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/Modified_cosine_heatmap.ipynb

Filters a list of spectra to exclude those whose InchIKey starts with 'FTXGVGWYYREYSV'. This is a preliminary step before selecting unique spectra.

```python
pos_test_spectra = []
for spectrum in pos_test:
    if spectrum.get("inchikey")[:14] != "FTXGVGWYYREYSV":
        pos_test_spectra.append(spectrum)
```

--------------------------------

### Initialize RDKit and Plot Spectrum/Molecule (Positive Mode)

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/case_studies_ms2deepscore_2/Rumex_case_study/case_study_mol_network_Rumex.ipynb

Initializes RDKit and plots a spectrum and its corresponding molecule image for a given positive ion mode feature ID. Requires 'spectra' and 'merged_data' to be pre-defined.

```python
from rdkit import Chem
from rdkit.Chem import Draw

ionmode_feature_id = "pos_2289"

spectrum = [spectrum for spectrum in spectra if spectrum.get("ionmode_feature_id") == ionmode_feature_id][0]
smiles = merged_data[merged_data["ionmode_feature_id"] == ionmode_feature_id]["smiles"].values[0]
fig, ax = spectrum.plot(figsize=(4, 3))
mol = Chem.MolFromSmiles(smiles)
# Display molecule image
Draw.MolToImage(mol)
```

--------------------------------

### Download Model and Spectra Files

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/Modified_cosine_heatmap.ipynb

Downloads necessary files for MS2 DeepScore benchmarking, including model weights, settings, and test spectra. It checks if files already exist to avoid redundant downloads.

```python
import requests
import os
from tqdm import tqdm

def download_file(link, file_name):
    response = requests.get(link, stream=True)
    if os.path.exists(file_name):
        print(f"The file {file_name} already exists, the file won't be downloaded")
        return
    total_size = int(response.headers.get('content-length', 0))
    
    with open(file_name, "wb") as f, tqdm(desc="Downloading file", total=total_size, unit='B', unit_scale=True, unit_divisor=1024,) as bar:
        for chunk in response.iter_content(chunk_size=1024):
            if chunk:
                f.write(chunk)
                bar.update(len(chunk))  # Update progress bar by the chunk size

model_file_name = "ms2deepscore_model.pt"
neg_test_spectra_file_name = "neg_testing_spectra.mgf"
pos_test_spectra_file_name = "pos_testing_spectra.mgf"

download_file("https://zenodo.org/records/14290920/files/settings.json?download=1", "ms2deepscore_settings.json")
download_file("https://zenodo.org/records/14290920/files/ms2deepscore_model.pt?download=1", model_file_name)
download_file("https://zenodo.org/records/13934470/files/negative_testing_spectra.mgf?download=1", neg_test_spectra_file_name)
download_file("https://zenodo.org/records/13934470/files/positive_testing_spectra.mgf?download=1", pos_test_spectra_file_name)
```

--------------------------------

### Initialize RDKit and Plot Spectrum/Molecule (Positive Mode)

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/case_studies_ms2deepscore_2/Rumex_case_study/case_study_mol_network_Rumex.ipynb

Initializes RDKit and plots a spectrum and its corresponding molecule image for a given positive ion mode feature ID. Requires 'spectra' and 'merged_data' to be pre-defined.

```python
from rdkit import Chem
from rdkit.Chem import Draw

ionmode_feature_id = "pos_475"

spectrum = [spectrum for spectrum in spectra if spectrum.get("ionmode_feature_id") == ionmode_feature_id][0]
smiles = merged_data[merged_data["ionmode_feature_id"] == ionmode_feature_id]["smiles"].values[0]
fig, ax = spectrum.plot(figsize=(4, 3))
mol = Chem.MolFromSmiles(smiles)
# Display molecule image
Draw.MolToImage(mol)
```

--------------------------------

### Load Spectra from MGF File

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/case_studies_ms2deepscore_2/Rumex_case_study/case_study_mol_network_Rumex.ipynb

Loads spectral data from a specified MGF file using matchms. `tqdm` is used for progress visualization during loading. Ensure the file path is correct.

```python
from matchms.importing import load_from_mgf
from tqdm import tqdm

case_study_spectra_file_name = "./results/combined_neg_pos_rumex_spectra.mgf"
spectra = list(tqdm(load_from_mgf(case_study_spectra_file_name)))
```

--------------------------------

### Sampled Array Reshaping

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/compare_with_cosine.ipynb

Reshapes a sampled array using sorted indices, similar to the previous example. This is useful for comparing sampled data against the original structure.

```python
sampled_neg_pos_both_mode.flatten()[sorted_idx].reshape(sampled_neg_pos_both_mode.shape)
```

--------------------------------

### Load Spectra from MGF File

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/Find_bias_in_bad_predictions.ipynb

Loads spectral data from a specified MGF file and prints the total number of spectra loaded. Requires the matchms library.

```python
from matchms.importing.load_from_mgf import load_from_mgf
val_spectra = list(load_from_mgf("./positive_validation_spectra.mgf"))
print(len(val_spectra))
```

--------------------------------

### Load Spectral Data from MGF Files

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/case_studies_ms2deepscore_2/Rumex_case_study/case_study_mol_network_Rumex.ipynb

Loads positive and negative ion mode spectra from MGF files using matchms. Ensure the file paths are correct.

```python
from matchms.importing import load_from_mgf
spectra_neg = list(load_from_mgf("./Rumex_case_study/Rumex_neg_20230806_all_samples_GNPS_fbmn_quant.mgf"))
spectra_pos = list(load_from_mgf("./Rumex_case_study/Rumex_20230806_all_samples_2_GNPS_fbmn_quant_quant.mgf"))
```

--------------------------------

### Filter Spectra by InChIKey Prefix

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/EmbeddingEvaluator_benchmarking.ipynb

This code snippet filters a list of positive spectra, excluding those whose InChIKey starts with 'FTXGVGWYYREYSV'. This is a preprocessing step before further analysis.

```python
pos_spectra = []
for spectrum in pos_test_spectra:
    if spectrum.get("inchikey")[:14] != "FTXGVGWYYREYSV":
        pos_spectra.append(spectrum)
```

--------------------------------

### Define Model and Spectra File Paths

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/Compare balanced cross ion mode sampling.ipynb

Sets up the file paths for the MS2DeepScore models trained with different sampling strategies and the test spectra file. Ensure these paths are correct for your environment.

```python
from matchms.importing import load_from_mgf
from tqdm import tqdm

normal_model_file_name = "/lustre/BIF/nobackup/jonge094/ms2deepscore/data/library_22_07_2025/trained_models/both_mode_ionmode_precursor_mz_10000_layers_500_embedding_2025_08_18_14_48_59/ms2deepscore_model.pt"
balanced_model_file_name = "/lustre/BIF/nobackup/jonge094/ms2deepscore/data/library_22_07_2025/trained_models/both_mode_ionmode_precursor_mz_10000_layers_500_embedding_2025_08_24_00_04_17/ms2deepscore_model.pt"

test_spectra_file = "/lustre/BIF/nobackup/jonge094/ms2deepscore/data/library_22_07_2025/trained_models/test_merged_and_cleaned_libraries_1.mgf"
```

--------------------------------

### Execute Evaluation and Get Shapes

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/ms2deepscore_2_preprint/EmbeddingEvaluator.ipynb

Applies the spectrum selection and MSE calculation functions to the validation data. This snippet also prints the shape of the MS2DS scores matrix.

```python
selected_all_val_spectra = select_one_spectrum_per_inchikey(selected_all_val_spectra)
predicted_mse, ms2ds_scores, tanimoto_scores, val_MSEs, val_MAEs = calculate_MSE(selected_all_val_spectra, ms2ds_model, ms2ds_embedding_evaluator)
ms2ds_scores.shape

```

--------------------------------

### Progress Bar Output for Balanced Selection with Different Settings

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/optimize_sampling_algorithm_settings.ipynb

Shows the progress of balanced sampling for inchikey pairs per bin when using different `max_pair_resampling` settings. This output confirms the sampling process is running.

```text
Balanced sampling of inchikey pairs (per bin): 100%|█████████████████████████████████████████████████████████████████████████████████████████| 188085/188085 [00:05<00:00, 32025.95it/s]
Balanced sampling of inchikey pairs (per bin): 100%|█████████████████████████████████████████████████████████████████████████████████████████| 188085/188085 [00:05<00:00, 31595.47it/s]
Balanced sampling of inchikey pairs (per bin): 100%|█████████████████████████████████████████████████████████████████████████████████████████| 188085/188085 [00:04<00:00, 38599.67it/s]
Balanced sampling of inchikey pairs (per bin): 100%|█████████████████████████████████████████████████████████████████████████████████████████| 188085/188085 [00:05<00:00, 31645.88it/s]
Balanced sampling of inchikey pairs (per bin): 100%|█████████████████████████████████████████████████████████████████████████████████████████| 188085/188085 [00:05<00:00, 31566.31it/s]
Balanced sampling of inchikey pairs (per bin): 100%|█████████████████████████████████████████████████████████████████████████████████████████| 188085/188085 [00:05<00:00, 31702.47it/s]
Balanced sampling of inchikey pairs (per bin): 100%|█████████████████████████████████████████████████████████████████████████████████████████| 188085/188085 [00:05<00:00, 31898.22it/s]
Balanced sampling of inchikey pairs (per bin): 100%|█████████████████████████████████████████████████████████████████████████████████████████| 188085/188085 [00:05<00:00, 32297.54it/s]
Balanced sampling of inchikey pairs (per bin): 100%|█████████████████████████████████████████████████████████████████████████████████████████| 188085/188085 [00:05<00:00, 32580.08it/s]
Balanced sampling of inchikey pairs (per bin): 100%|█████████████████████████████████████████████████████████████████████████████████████████| 188085/188085 [00:05<00:00, 32427.66it/s]

```

--------------------------------

### Initialize Pair Generators Dictionary

Source: https://github.com/matchms/ms2deepscore/blob/main/notebooks/model_benchmarking/optimize_sampling_algorithm_settings.ipynb

Initializes an empty dictionary to store pair generators, likely for organizing results from different sampling settings. This is a setup step before performing balanced sampling.

```python
pair_generators_standard_order = {}
```