PASS APPLICATION IN R&D OF NEW
PHARMACEUTICALS (adaptation of a text from Prof. Vladimir Poroikov, January
2005)
A new drug entity has
to pass successfully the following steps:
-
Ligand design (hits).
-
Chemical synthesis and/or purchase of
samples for biological testing.
-
Ligand finding (leads): in vitro testing
of the required specific biological activity.
-
Ligand optimization (drug-candidates):
in vivo confirmation of the required specific biological activity;
investigation of general pharmacological/toxicological profile (no
adverse/toxic effects at the appropriate doses) of the selected
substances; investigation of pharmacokinetics of the selected substances
(favorable absorption, distribution, metabolism and excretion
characteristics).
-
Submitting IND to get a permission of
Drug Authority for clinical trials.
-
Clinical trials, final proof of the
concept.
-
Submitting NDA to get an approval of
Drug Authority for medical application of the drug-candidate.
At any stage, project
may failure due to different reasons. More than 30% of failures in
pharmaceutical R & D projects are due to the adverse/toxic effects, which
are found at the later stages of the project when a lot of time and money
are already spent (for nothing).
Typically,
any chemical compound exhibits several or many kinds of biological activity,
and the final goal of R&D is to select the compounds with the required
pharmacological action but without unwanted adverse/toxic effects.
The whole complex of biological activities
that might be revealed by chemical compound during its interaction with the
human organism is called biological activity spectrum. It is not possible to
test experimentally millions of available compounds against thousands known
kinds of biological activity.
Computer program PASS predicts biological
activity spectrum of compound based on its structural formula. In version of
PASS 1.913 (December 2004) predicted biological activity spectrum includes
986 kinds of biological activity including:
-
677 actions on particular targets (e.g., 5
Hydroxytryptamine 2A antagonist, Acetylcholine M4 receptor agonist,
Adenosine deaminase inhibitor, Alpha glucosidase inhibitor, Calcium channel
N-type antagonist, Dipeptidyl peptidase IV inhibitor, Endothelin A receptor
antagonist, Growth hormone release inhibitor, Insulin sensitizer,
Leukotriene E4 antagonist, etc.);
-
44 actions on a particular infectious agent (Acaricide,
Antifungal, Antihelmintic, Antimycobacterial, Anti-HIV, Anti-HCV, etc.);
-
226 pharmacotherapeutic effects (e.g., Analgesic,
Antiarrhythmic, Bone formation stimulant, Bronchodilator, Cognition
disorders treatment, Diuretic, Immunomodulator, Male reproductive
disfunction treatment, Prostatic benign hyperplasia treatment, etc.);
-
44 adverse effects and toxicities (e.g., Arrhythmogenic,
Cardiotoxic, Convulsant, Hypertensive, Mutagenic, Carcinogenic, Embryotoxic,
etc.).
Application of
computer programs PASS (Prediction of Activity Spectra for Substances) and
PharmaExpert provide the following opportunities: (1) to increase the number
of hits in the sub-set of compounds selected for synthesis and biological
testing, and (2) to filter out the hits with likely unwanted adverse/toxic
action.
Let us consider two
examples of such PASS application:
(1) Selection of
compounds with anti-HIV activity.
To analyze how PASS
predictions can enrich the number of active compounds in the subset selected
on the basis of computer prediction from the database of chemical compounds,
we compared the results of anti-HIV activity prediction for the compounds
from the Open NCI Database with the results of anti-HIV screening. Within
the 250000 compounds from the Open NCI Database, the subset of 42689
compounds was tested versus anti-HIV activity, and the number of active
compounds was found to be 1504. Thus, the percentage of actives in the
tested subset of open NCI compounds is 3.52% (1504/42689). A random
selection would therefore preserve this ratio. Using PASS prediction, even
if the value Pa>10% is used as a threshold for selecting active compounds,
the fraction of “actives” is enriched to a factor 2.2. At the highest
threshold Pa>90%, the enrichment gets close to a factor of 17 [Poroikov et
al., 2003].
(2) Filtering
compounds from the Prestwick Chemical Library.
Prestwick Chemical Library [http://www.prestwickchemical.com]
is a collection consisted of 880 carefully selected compounds, which are
highly diverse in structure and cover many therapeutic areas – from
neuropsychiatry to cardiology, immunology, anti-inflammatory, and more.
Over 85% of these compounds are marketed drugs, for which
both main pharmacological actions and some adverse/toxic effects are known.
In particular, the convulsant effect is found for 49 compounds from this
collection (e.g., acetazolamide, amitryptiline, arecoline, chlotpromazine,
baclofen, buflomedil, bupivacaine, bupropion, clozapine, enoxacin, ethamivan,
fenfluramine, haloperidol, hydrastine, iohexol, laudanosine, lidocaine,
maprotiline, mefenamic acid, methazolamide, mianserine, mefloquine,
metrizamide, naloxone, nefopam, orphenadrine, pimozide, propafenone,
quinidine, quinacrine, terfenadine, theophylline, etc.).
We obtained PASS predictions for all 880 compounds, and
select 99 compounds, which are likely 5 hydroxytryptamine release stimulants
and, therefore, might be applied as antidepressants. Analyzing the subset,
we show that 74 of these compounds are also predicted as being convulsants
with probability more than 40%.
Using computer program PharmaExpert we retrieve the results
of prediction for the Prestwick Chemical Library based on the following
query: “5 hydroxytryptamine release stimulant with probability more than 50%
NOT convulsant”. As a result, we obtain 12 compounds that correspond to this
query. No one known convulsant was included into this sub-set.
Some advantages
of PASS use.
Possibility
of application at early stages of the research.
Because only structural formula of compound (hit) is necessary as input for
PASS, computer prediction can be obtained at the very early step of
pharmaceutical R &D (ligand design) when no time & money are yet spent on
chemical synthesis, biological testing, etc.
Reasonable accuracy of prediction.
Average accuracy of prediction in leave one out cross-validation (for
~57.000 compounds and ~1.000 kinds of biological activity from the PASS
training set) is about 85%. PASS algorithm produce robust estimates of
structure-activity relationships despite the incompleteness of the training
set [Poroikov et al., 2000].
Predictions
are rather fast.
Calculation of biological activity spectra for 10.000 compounds on an
ordinary PC takes about 5 min; therefore PASS can be effectively used to
analyze the databases consisted of millions of structures.
Standard
structure format is used.
Standard SDF-file format (http://www.mdli.com) is used as input for PASS;
therefore, the existing databases of chemical structures can be easily
retrieved.
Possibility
of creating the exclusive knowledgebase.
The user can add new biologically active compounds and new kinds of
biological activity to the training set, and create his knowledgebase(s);
therefore, the “in house” proprietary data can be effectively applied for
this purpose on the exclusive basis.
Possibility of free testing.
PASS prediction abilities can be freely tested via Internet [Sadym et al.,
2003].
References for
further reading
Anzali S., Barnickel G., Cezanne B., Krug M., Filimonov D.,
Poroikov V. (2001). Discriminating between drugs and nondrugs by Prediction
of Activity Spectra for Substances (PASS). J. Med. Chem., 44 (15),
2432-2437.
Filimonov D., Poroikov V., Borodina Yu., Gloriozova T.
(1999). Chemical Similarity Assessment through multilevel neighborhoods of
atoms: definition and comparison with the other descriptors.
J.Chem.Inf.Comput. Sci., 39 (4), p.666-670.
Geronikaki A., Babaev E., Dearden J., Dehaen W., Filimonov
D., Galaeva I., Krajneva V., Lagunin A., Macaev F., Molodavkin G., Poroikov
V., Saloutin V., Stepanchikova A., Voronina T. (2004). Design of new
anxiolytics: from computer prediction to synthesis and biological
evaluation. Bioorg. Med. Chem.,
12 (24),
6559-6568.
Geronikaki A., Dearden J., Filimonov D., Galaeva I., Garibova
T., Gloriozova T., Krajneva V., Lagunin A., Macaev F., Molodavkin G.,
Poroikov V., Pogrebnoi S., Shepeli F., Voronina T., Tsitlakidou
M., Vlad L. (2004). Design of new cognition enhancers: from computer
prediction to synthesis and biological evaluation. J. Med. Chem., 47
(11), 2870-2876.
Lagunin A.A., Gomazkov O.A., Filimonov D.A., Gureeva T.A.,
Dilakyan E.A., Kugaevskaya E.V., Elisseeva Yu.E., Solovyeva N.I., Poroikov
V.V. (2003). Computer-aided selection of potential antihypertensive
compounds with dual mechanisms of action. J. Med. Chem., 46 (15),
3326-3332.
Poroikov V., Akimov D., Shabelnikova E., Filimonov D. (2001).
Top 200 medicines: can new actions be discovered through computer-aided
prediction? SAR and QSAR in Environmental Research, 12 (4), 327-344.
Poroikov V.,
Filimonov D. (2001). Computer-aided prediction of biological activity
spectra. Application for finding and optimization of new leads. Rational
Approaches to Drug Design, Eds. H.-D. Holtje, W.Sippl, Prous Science,
Barcelona, p.403-407.
Poroikov V.V., Filimonov D.A. (2002). How to acquire new
biological activities in old compounds by computer prediction. J. Comput.
Aid. Molec. Des.,
16
(11),
819-824.
Poroikov V., Lagunin A. (2002). PharmaExpert: knowledge-based
computer system for interpretation of biological activity spectrum for
substance. Newsletter of The QSAR and Modelling Society, No.13, p.23-24.
Poroikov V.V., Filimonov D.A., Ihlenfeldt W.-D., Gloriozova
T.A., Lagunin A.A., Borodina Yu.V., Stepanchikova A.V., Nicklaus M.C.
(2003). PASS Biological Activity Spectrum Predictions in the Enhanced Open
NCI Database Browser. J. Chem. Inform. Comput. Sci.,
43
(1) 228-236.
Poroikov V., Filimonov D. (2005). PASS: Prediction of
Biological Activity Spectra for Substances. In: Predictive Toxicology. Ed.
by Christoph Helma. N.Y.: Marcel Dekker, 459-478.
Sadym A., Lagunin A., Filimonov D., Poroikov V. (2003).
Prediction of biological activity spectra via Internet. SAR and QSAR in
Environmental Research, 14 (5-6), 339-347.
Stepanchikova A.V., Lagunin A.A., Filimonov D.A., Poroikov
V.V. (2003). Prediction of biological activity spectra for substances:
Evaluation on the diverse set of drugs-like structures. Current Med. Chem.,
10 (3), 225-233.
To simplify the scheme, some technological stages (e.g., process
development, dosage form development, etc.) are omitted in the steps
given above.
|
Further applications with PASS
-
How to select promising candidates from
supplier databases! Follow
this hyperlink.
-
See
http://cactus.nci.nih.gov/ncidb2.2/
for a WWW browser to the new and enlarged collection of open NCI database
compounds (>250,000 structures) which includes the PASS parameters for
all these compounds.
|