try:
mass = CalcExactMolWt(Chem.MolFromSmiles(smi))
except:
continue
molwt.append(mass)
smiles.append(smi)
spec.append(ms2vec(ms[:, 0], ms[:, 1]))
mlp = pd.read_csv('Fingerprint/results/mlp_result.txt',
sep='\t',
header=None)
mlp.columns = ['id', 'accuracy', 'precision', 'recall', 'f1']
fpkeep = mlp['id'][np.where(mlp['f1'] > 0.5)[0]]
spec = np.array(spec)
pred_fps = predict_fingerprint(
spec, fpkeep) # predict fingerprint of the "unknown"
nist_smiles = np.array(json.load(open('DeepEI/data/all_smiles.json')))
nist_masses = np.load('DeepEI/data/molwt.npy')
nist_fps = load_npz('DeepEI/data/fingerprints.npz')
nist_fps = csr_matrix(
nist_fps)[:, fpkeep].todense() # fingerprints of nist compounds
nist_spec = load_npz('DeepEI/data/peakvec.npz').todense()
pred_spec = np.load(
'DeepEI/data/neims_spec_massbank.npy') # spectra predicted by NEIMS
output = pd.DataFrame(columns=['smiles', 'mass', 'score', 'rank'])
for i in tqdm(range(len(smiles))):
smi = smiles[i]
std_smi = Chem.MolToSmiles(Chem.MolFromSmiles(smi))
keep, :]
neims_msbk_smiles = np.array(
json.load(open('DeepEI/data/neims_msbk_smiles.json')))
neims_msbk_masses = np.load('DeepEI/data/neims_msbk_masses.npy')
neims_msbk_spec = load_npz('DeepEI/data/neims_spec_msbk.npz').todense()
neims_msbk_cdkfps = load_npz('DeepEI/data/neims_msbk_cdkfps.npz').todense()
msbk_smiles = np.array(json.load(open('DeepEI/data/msbk_smiles.json')))
msbk_masses = np.load('DeepEI/data/msbk_masses.npy')
msbk_spec = load_npz('DeepEI/data/msbk_spec.npz').todense()
mlp = pd.read_csv('Fingerprint/results/mlp_result.txt', sep='\t', header=None)
mlp.columns = ['id', 'accuracy', 'precision', 'recall', 'f1']
fpkeep = mlp['id'][np.where(mlp['f1'] > 0.5)[0]]
pred_fps = predict_fingerprint(msbk_spec, fpkeep)
db_smiles = np.array(list(nist_smiles) + list(neims_msbk_smiles))
db_masses = np.append(nist_masses, neims_msbk_masses)
db_spec = np.append(neims_nist_spec, neims_msbk_spec, axis=0)
db_fingerprints = np.append(nist_fingerprint, neims_msbk_cdkfps,
axis=0)[:, fpkeep]
if __name__ == '__main__':
output = pd.DataFrame(
columns=['smiles', 'mass', 'score', 'rank', 'inNIST'])
for i, smi in enumerate(tqdm(msbk_smiles)):
specr = msbk_spec[i] # true spectrum
mass = msbk_masses[i] # true mol weight
test_rindex = test_ri[i,0]
test_spec = test_spec[i,:]
test_mass = test_masses[i]
test = i
# only keep fingerprint with f1 > 0.5
mlp = pd.read_csv('Fingerprint/results/mlp_result.txt', sep='\t', header=None)
mlp.columns = ['id', 'accuracy', 'precision', 'recall', 'f1']
fpkeep = mlp['id'][np.where(mlp['f1'] > 0.5)[0]]
cdk_fp = load_npz('DeepEI/data/fingerprints.npz')
cdk_fp = csr_matrix(cdk_fp)[:, fpkeep].todense()
# predict fingerprints via ms
pred_fp = predict_fingerprint(test_spec, fpkeep)
# rank
output = pd.DataFrame(columns=['smiles', 'mass', 'true RI', 'predict RI', 'mass filter', 'RI filter', 'Without filter'])
for i in tqdm(range(len(test))):
smi = test_smiles[i]
mass = test_mass[i]
ri = test_rindex[i]
pred_fpi = pred_fp[i,:]
trueindex = np.where(all_smiles == smi)[0][0]
# mass filter
candidate = np.where(np.abs(all_masses - mass) < 5)[0]
w_true = np.where(candidate==trueindex)[0]
if len(w_true)==0:
rank_mass = 99999