DeepCombDTI¶
DeepCombDTI is a model for drug-target interaction prediction using deep learning and various molecular fingerprints. Drugs can be represented by a molecular fingerprint like ECFP and Protein can be represented a feature vector like CTD.
Overview¶
Dataset¶
Test dataset types¶
All
All drugs and targets in the test dataset
Unseen drug
Drugs are not included in the training dataset
Unseen prot
Proteins are not included in the training dataset
Unseen both
Drugs and proteins are not included in the training dataset
Features¶
Molecular fingerprint¶
Molecular fingerprint는 약물을 표현하는 Feature이고, 예로는 ECFP, Mol2vec, Neural fingerprint, Seq2seq fingerprint 등이 있다.
Results¶
SFDTIP vs. Random forest¶
SFDTIP vs. EFDTIP¶
Property prediction¶
Performance measure¶
RMSE: 편차 제곱의 평균
Pearson correlation: 두 변수의 공분산을 표준 편차의 곱으로 나눈 값
Qualitative Estimate of Drug-likeness (QED)¶
Qualitative Estimate of Drug-likeness (QED) is a measure of a binding abilitity. This is a equation of QED:
The QED is achieved by taking the geometric mean of the individual functions. A series of desirability functions (d) are derived, each corresponding to a different molecular descriptor:
Molecular weight (MW)
Octanol-water partition coefficient (ALOGP)
The number of hydrogen bond donors (HBD)
The number of hydrogen bond acceptors (HBA)
The molecular polar surface area (PSA)
The number of rotatable bonds (ROTB)
The number of aromatic rings (AROM)
The number of structural alerts (ALERTS)
This is a result of QED predictions:
Solubility¶
Melting point¶
LogP¶
LogP는 섞이지 않는 두 용매인 물과 옥탄올(octanol)에 화합물을 녹였을 때 물과 옥탄올층에 녹아있는 화합물 농도의 비를 분배계수로 나타낸 것입니다.
이 값을 통하여 화합물이 친수성(hydrophilicity)을 갖는지 소수성(hydrophobicity)을 갖는지 예상할 수 있으며 값이 클수록 소수성이 강한 옥탄올층에 화합물의 분배가 많이 되어있는 것이므로 소수성이 큰 것으로 예상할 수 있습니다.