A.A., Filimonov D.A., Poroikov V.V.
of Biomedical Chemistry RAMS, Russia, Moscow
Alexander Kos 15.10.02
biologically active compound reveals a wide variety of biological actions in
biological systems (human organisms, animals,
in vivo and in vitro assays). It is practically impossible to study each
compound in all tests currently available. Therefore, the ability to select
compounds with required types of biological activity and without unwanted
adverse effects and toxicity is very desirable.
software PASS (Prediction
of Activity Spectra
for Substances) and PharmaExpert
was developed toward this purpose. PASS predicts biological activity spectra
on the basis of structural formulae of chemical compounds. The biological
activity of compounds is predicted on the basis of structure-activity
relationships of known biological active substances presented in the training
set. PASS 1.602 training set includes 45649 substances. PASS 1.602 can predict
1043 different types of biological activity including pharmacological effects,
biochemical mechanisms, carcinogenicity, mutagenicity and teratogenicity. The
mean prediction of accuracy in leave one out cross-validations of PASS is
determines the existing relationships between pharmacological effects and
biochemical mechanisms. The current version of PharmaExpert covers 1587
mechanisms of action, 418 pharmaco-therapeutical effects and 2664 types of
relationships between them.
The above screenshot shows the relationship of the activities in the area 1. With the button 2 you can toggle the hierarchy looking once at mechanisms, or at effects.
can I select compounds which have the required biological activities without
having unwanted effects?
the search functionality (1) you can select the desired activities (2),
and you can define the search criteria, you also can search by Pi (predicted
inactivity - useful for adverse effects). In the result window you will find
the actual values (3), the compound IDs (4) and highlighting the ID will
display the corresponding
can I find a profile of important activities for a class of compounds?
The "Statistic" window in PharmaExpert lets you sort the activities by name, or by numbers. For this group of known AChEIs the most important activities are given below.
No Pa>Pi Pa>30% Pa>50% Pa>70% Types of Activity 1 134 134 124 86 Nootropic 2 133 129 110 66 Cognition disorders treatment 3 134 120 89 52 Cholinergic 4 130 119 84 49 Acetylcholinesterase inhibitor 5 95 95 84 41 Antiamyloidogenic 6 86 86 85 38 Antieczematic atopic 7 73 71 40 32 Dependence treatment 8 86 86 72 26 Neurotransmitter agonist 9 91 73 40 24 Miotic 10 63 63 53 23 Hepatotoxic 11 76 76 65 22 Hydrolase inhibitor 12 80 80 73 19 Antiurolithic 13 58 58 45 18 Immunomodulator (HIV) 14 52 32 19 17 Butyrylcholinesterase inhibitor 15 67 44 19 17 Beta amyloid precursor protein antagonist 16 118 118 46 14 Alzheimer's disease treatment 17 91 91 71 13 Adrenaline release inhibitor 18 90 90 57 12 Alpha 1b adrenoreceptor antagonist 19 74 74 54 11 Apoptosis antagonist 20 44 44 22 10 Fibrinogen receptor antagonist
We have to emphasize that we have no proof yet, if such a selection correspond to the real situation of a drug.
I want to develop new drugs with new structural features, or a new skeleton. How can PASS help me, if its prediction is dependent on a training set of known compounds?
We have two approaches to this problem. (1) We described the use of profiles, and clustering with neural networks, see slides 17 and 18 from a presentation at an ACS conference.
(2) You can use PharmaExpert.
In the screenshot above we show the PASS parameters for the activity "Antiparkinsonian" (1) that were calculated, and (2) you see the complete relationship of "Antiparkinsonian" to other effects. If your desired activity would be "Adrenaline Agonist" you might not find any prediction for it because your structure class and this activity was not covered in the training set. You might find that your compound is "Antiparkinsonian". The relationship of effects and mechanism tells you that this compound might have your desired activity "Adrenaline Agonist". Through the relationship of effects you bridge the gap that occurs because many structure classes were not investigated for a certain activity. There is still the chance that your structure class is completely new, and PASS might miss it. However, the MNA descriptors of PASS are sensitive to small fragments, and your structure class must be so new that it does not contain any small similar fragments to already observed compounds.
PASS and PharmaExpert can give you an early indication, obtained cheap and fast if your compound might be a good or bad drug candidate!