Repository for NYCU JHHLab NPS detection project. The repository of the paper "PS²MS".

• Original paper: "Rapid Prediction of Electron–Ionization Mass Spectrometry Using Neural Networks"
• Source code: NEIMS Github
• Quickstart of retraining model from NEIMS: Model retrain quickstart

python3 make_train_test_split.py --main_sdf_name=path/to/mainlib_merge.SDF --replicates_sdf_name=path/to/test.SDF --output_master_dir=path/to/output/dir/spectra_tf_records

• --main_sdf_name: The file comprises mass spectra of all compounds, encompassing both the training and test sets, presented in SDF format.
• --replicates_sdf_name: The file comprises mass spectra of compounds of test set, presented in SDF format.
• --output_master_dir: The output directory of preprocessing data.

python3 molecule_estimator.py \
  --dataset_config_file=path/to/output/dir/spectra_tf_records/query_replicates_val_predicted_replicates_val.json \
  --train_steps=10000 \
  --model_dir=path/of/output/model/models/output \
  --hparams=make_spectra_plots=True,mask=False,mass_power=0 --alsologtostderr

• --dataset_config_file: The path of preprocessed dataset
• --model_dir: The path of model

python3 make_spectra_prediction.py \
  --input_file=path/to/test_SMILES.txt \
  --output_file=path/to/test.SDF \
  --weights_dir=path/of/output/model/models/output

• --input_file: The SMILES of test set, presented in txt format.
• --output_file: The predict mass spectrum of test set, presented in SDF format.
• --weights_dir: The path of model.

python3 sdf_to_msp.py ${input_file} ${is_predict_spectrum}

• Convert files from SDF format to msp format.
• ${is_predict_spectrum}: a boolean value. If the input file contains predict mass spectrum, this value should beset to True.

• Original paper: "Predicting a Molecular Fingerprint from an Electron Ionization Mass Spectrum with Deep Neural Networks"
• Source code: DeepEI Github

• Please refer to the README.md in the repository of DeepEI.

python3 merge_fp_into_msp.py \
  ${MS.msp} \
  ${FP.pkl} \
  ${result.msp}

• By Samuel

conda env create --name enumerate -f enumerate/environment.yml

cd enumerate 
conda activate enumerate

mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j

cd enumerate 
conda activate enumerate

./build/build_database <split_count> <split_idx> <output_dir>

# for example: ./build/build_database 200 0 /tmp

• Source code: cfm-id-code
  • We utilize the data type and the function responsible for calculating the cosine similarity between two mass spectra in this project.

Follow the step in INSTALL.txt

• Condition 1: We know what analytes actually are.
  • database_file: The file contains the mass spectrum and the fingerprint of enumerated compounds and is in msp format.
  • test_file: The file contains the mass spectrum and the fingerprint of the analytes and is in msp format.
  • result.txt: The file records the ranking performance of every analyte. For every analyte, the first line is the rank and the SMSF score of the answer in enumerate database. The following lines are the SMILES and SMSF score of molecular which has higher rank than the answer in enumerate database.
  • scores.txt: The file records the similarity scores of every analyte. For an analyte, it records the cosine similarity of mass spectrum, the Jaccard similarity of mass spectrum, the similarity of fingerprint and the SMSF score of with the corresponding one in enumerate database.
  • restrict_mw: A boolean value. If true, the system will solely compute the similarity for compounds whose molecular weight falls within the range of the analyte's molecular weight plus or minus one..
  • top_n_of_JS_of_MS: A positive integer. This value determines the number of the highest peaks used to compute the Jaccard similarity of mass spectra between two compounds. If set to zero, the system will not calculate the Jaccard similarity of mass sepctrum.

./build/drug-comparation/cfm_comparation \
  ${database_file} \
  ${test_file} \
  ${result.txt} \
  ${scores.txt} \
  ${restrict_mw} ${top_n_of_JS_of_MS}

• Condition 2: We don't know what analytes are.
  • database_file: The file contains the mass spectrum and the fingerprint of enumerated compounds and is in msp format.
  • test_file: The file contains the mass spectrum and the fingerprint of the analytes and is in msp format.
  • result.txt: The file records the result of every analyte. For every analyte, the top 100 compounds is recorded with the SMILES and SMSF score.
  • restrict_mw: A boolean value. If true, the system will solely compute the similarity for compounds whose molecular weight falls within the range of the analyte's molecular weight plus or minus one..
  • top_n_of_JS_of_MS: A positive integer. This value determines the number of the highest peaks used to compute the Jaccard similarity of mass spectra between two compounds. If set to zero, the system will not calculate the Jaccard similarity of mass sepctrum.

./build/drug-comparation/cfm_comparation_without_answer \
  ${database_file} \
  ${test_file} \
  ${result.txt} \
  ${restrict_mw} ${top_n_of_JS_of_MS}