Discovery of Multitarget-Directed Ligands against Alzheimer’s Disease through Systematic Prediction of Chemical–Protein Interactions

To
determine chemical–protein interactions (CPI) is costly,
time-consuming, and labor-intensive. In silico prediction
of CPI can facilitate the target identification and drug discovery.
Although many in silico target prediction tools have
been developed, few of them could predict active molecules against
multitarget for a single disease. In this investigation, naive Bayesian
(NB) and recursive partitioning (RP) algorithms were applied to construct
classifiers for predicting the active molecules against 25 key targets
toward Alzheimer’s disease (AD) using the multitarget-quantitative
structure–activity relationships (mt-QSAR) method. Each molecule
was initially represented with two kinds of fingerprint descriptors
(ECFP6 and MACCS). One hundred classifiers were constructed, and their
performance was evaluated and verified with internally 5-fold cross-validation
and external test set validation. The range of the area under the
receiver operating characteristic curve (ROC) for the test sets was
from 0.741 to 1.0, with an average of 0.965. In addition, the important
fragments for multitarget against AD given by NB classifiers were
also analyzed. Finally, the validated models were employed to systematically
predict the potential targets for six approved anti-AD drugs and 19
known active compounds related to AD. The prediction results were
confirmed by reported bioactivity data and our in vitro experimental validation, resulting in several multitarget-directed
ligands (MTDLs) against AD, including seven acetylcholinesterase (AChE)
inhibitors ranging from 0.442 to 72.26 μM and four histamine
receptor 3 (H3R) antagonists ranging from 0.308 to 58.6
μM. To be exciting, the best MTDL DL0410 was identified as an
dual cholinesterase inhibitor with IC50 values of 0.442
μM (AChE) and 3.57 μM (BuChE) as well as a H3R antagonist with an IC50 of 0.308 μM. This investigation
is the first report using mt-QASR approach to predict chemical–protein
interaction for a single disease and discovering highly potent MTDLs.
This protocol may be useful for in silico multitarget
prediction of other diseases.