Checkpoint Contents Pension & Benefits Library Pension & Benefits Editorial Materials EBIA Benefits Compliance Library Cafeteria Plans XX. What Expenses Can Be Reimbursed Under a Health FSA? XX.M. Table of Common Expenses, Showing Whether They Are for "Medical Care" M. Table of Common Expenses, Showing Whether They Are for "Medical
Bioorganic & Medicinal Chemistry 14 (2006) 7011–7022 Drug Guru: A computer software program for drug design using medicinal chemistry rules Kent D. Stewart,a,* Melisa Shirodaa and Craig A. Jamesb aAbbott Laboratories, Global Pharmaceuticals Research and Development, Abbott Park, IL 60064, USA bMoonview Consulting, LLC, San Diego, CA, USA Received 27 April 2006; revised 6 June 2006; accepted 8 June 2006 Available online 25 July 2006 Abstract—Drug GuruTM (drug generation using rules) is a new web-based computer software program for medicinal chemists thatapplies a set of transformations, that is, rules, to an input structure. The transformations correspond to medicinal chemistry designrules-of-thumb taken from the historical lore of drug discovery programs. The output of the program is a list of target analogs thatcan be evaluated for possible future synthesis. A discussion of the features of the program is followed by an example of the softwareapplied to sildenaﬁl (ViagraÒ) in generating ideas for target analogs for phosphodiesterase inhibition. Comparison with other com-puter-assisted drug design software is given.
Ó 2006 Elsevier Ltd. All rights reserved.
A rich tradition of analog design strategies has evolved for creating new compounds within medicinal chemistry research for biological evaluation. When similar physi- cal properties between lead compound and analog are desired, ‘bioisosteric' replacements are commonly em- ployed. Where more structurally altered yet still compat- ible diﬀerences between lead compound and analog are desired, non-classical replacements are considered. This latter situation occurs when a chemist desires structures Figure 1. Two well-known examples of rule-of-thumb for designing that are outside of the intellectual property of a compet- analogs: (a) the carboxylate-to-tetrazole replacement, (b) the amide-to- itor or when attempting to achieve more dramatic retroamide switch. Other examples are listed in or described in changes in potency or bioavailability. Collectively, these replacements are known to experienced medicinal chem-ists as ‘rules-of-thumb' for drug design. Two examplesof well-known design rules-of-thumb, the carboxylate- sent a useful starting place in a drug discovery eﬀort, particularly when other knowledge such as pharmaco- illustrated in have historically been considered phore models, 3D receptor structure, or structure– to yield analogs of high interest in medicinal chemistry activity relationship data is limited, low quality, or programs. Hundreds of these structural transformations have been reported in the literature and have potentialfor general applicability and acceptance as design We have written a web-based computer application, rulesWhile no single rule is ever guaranteed to called ‘Drug Guthat contains the historical rules- achieve the desired endpoint, the traditional rules repre- of-thumb as lines of SMIRKS code.The name ofthe program is an acronym for drug generation usingrules. The program applies a library of rules to any inputstructure and then permits visual or computational eval- Keywords: Computational chemistry; Drug design.
* Corresponding author. Tel.: +1 847 937 1205; fax: +1 847 937 uation of the output structures. To our knowledge, no previously described or commercially available software 0968-0896/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2006.06.024 K. D. Stewart et al. / Bioorg. Med. Chem. 14 (2006) 7011–7022 accomplishes all of the tasks of Drug Guru, and we pro- The ring-break transformation (rule #9 in ﬁnds pose that it nicely complements other computational its basis in the general strategy to take atoms covalently approaches. In this article, we describe the basic features connected within a ring and replace with atoms that are of the software, illustrate its use in a retrospective study intramolecularly H-bonded to form a ‘virtual' ring sys- of sildenaﬁl, and end with a comparison to other com- tem. Drug Guru has several of these ring-break rules.
puter-aided drug design software.
In the example described here, one of the aryl rings ofa fused aromatic ring is replaced with a carbonyl andamine that project from the other aryl ring. This ring- 2. Materials and methods break rule was recently illustrated by Novartis in theirscaﬀold morphing discovery of the anthranilic amide 2.1. Drug Guru rules inhibitors of KDR kinase.This rule and example areillustrated in The most important and novel aspect of Drug Guru isthe collection of rules. We found the SMIRKS transfor- The NC-switch transformation, rule #10 in , mation protocol useful for encoding rules.SMIRKS is ﬁnds its basis within the general ‘carbon-to-nitrogen' a linear text string that represents a graph transforma- replacement strategy used in discovery of new heterocy- tion which, when applied, converts a representation of cles as core pieces in drugs. The general notion is to take an input structure (a SMILES code) into a new structure every carbon in a structure and sequentially replace each (a new SMILES code). Ten illustrative rules and their with nitrogen. In the speciﬁc case of rule #10, two atoms corresponding SMIRKS codes are given in are ‘exchanged' in the situation where a vinyl amine ispart of the aromatic system, leading to an ‘NC-switch' In practice, two general kinds of SMIRKS are required rule. A good historical illustration of an application of to cover the transformations encountered in most this NC-switch rule may be found in the research pub- medicinal chemistry programs: functional group trans- lished by Abbott Laboratories that led to the pyridone formations and molecular framework modiﬁcations. In class of antibacterial agents typiﬁed by ABT-719.This the version of Drug Guru described here, there are 133 rule and example are illustrated in and 53 rules in each of these general classes, respectively,for a total of 186 rules. A list of rule categories is given 2.2. Structure input in Fourteen functional groups were empiricallyselected as most frequently encountered in medicinal Input of a chemical structure can be via corporate iden- chemistry research programs. Rules corresponding to tiﬁer code, drawing program, or by coordinate ﬁle, entries 1–7 of are typical functional group trans- either individually or in batch. Typical input structures formations and will be familiar to most experienced include the current lead structure for a particular pro- chemists. A text mnemonic is assigned to each rule to ject, a competitor compound or a natural ligand. After permit ready comprehension of the general nature of entering the input structure, the user selects the run but- the rule , column 1). An extensive literature sur- ton with no additional user information needed. All gen- vey of medicinal chemistry reports is currently underway eral transformations are applied in this mode. An to produce a more expanded and comprehensive listing ‘expert user' page is optionally accessed to allow some of rules and their corresponding SMIRKS. In addition additional features (described below) to be explored if to the functional group transformation rules, a variety desired. The input web page is illustrated in .
of molecular framework modiﬁcation rules are also Tautomerism of the input structure was found to be encoded, including ring break/form, ring contraction/ex- an important factor: diﬀerent tautomers give diﬀerent pansion, and ring replacement rules (Rules for output results. When tautomeric possibilities are found entries 8–10 in are representative framework by an automatic tautomer check within Drug Guru, modiﬁcations. Homologation rules, such as transforma- the user is queried to select the desired input tions that extend ring substituents by an oxygen, sulfur, carbon, or nitrogen atoms (such as entry 8, areexamples of rules that currently fall into a ‘miscella- 2.3. Evaluation of output neous' category (Also listed in withinthe category of framework modiﬁcations are rules for In Drug Guru, the primary method of evaluating the altering molecular conformation: for example, the addi- output structures is by visual inspection. Currently, we tion of geminal methyl groups to ﬂexible chains to ex- have implemented 186 rules and empirically observe that ploit the Thorpe–Ingold eﬀecTwo framework a typical medicinal chemistry request will result in 50– modiﬁcation rules from the ring-break rule, en- 150 output structures. Structures with higher structural try 9; and the NC-switch rule, entry 10, illustrate more complexity, that is, high number of functional groups complex and less obvious structural transformations or skeletal connections, yield more instances of rule and are described in detail below. A public web-based applicability, thus leading to a greater number of output utility program is available ( structures. Drug Guru requires only seconds for genera- that will allow readers to evaluate tion of results. The visual evaluation and thoughtful the SMIRKS supplied in Drug Guru is written analysis of the results by the end-user medicinal chemists in Perl programming language, and uses the DayPerl2 typically requires 10–30 min. With the number of output interface to the Daylight SMILES, SMARTS, and structures in the hundreds, we desired a database man- Reaction Toolkits.
agement tool that would conveniently permit casual Table 1. Illustrative listing of rules Rule illustration [C,c:1]-[OH] [C,c:1]-[O]-C 2]-[CH2]-[C,c:2] [C,c:1]-[S]-[C,c:2] K. D. Stewart et al. / Bioorg. Med. Chem. 14 (2006) 7011–7022 Table 2. Rule categories and number of rules not surprising since Drug Guru was programmed with Functional group transformations rules derived from the traditional medicinal chemistry knowledge base. Even at this level, the program pro- vides value due to its comprehensiveness. Importantly, in test situations, chemists have additionally indicated that non-obvious structures are also included, thus mak- ing the use of Drug Guru an even more fruitful experienced medicinal chemists gain insight from Drug Guru as a starting place for learning about analog de- sign. The references and example structures provided with each rule provide a starting place for learning prac- tical medicinal chemistry (see Rule History description Drug Guru will occasionally generate low quality, non- Molecular framework modiﬁcationsRing break drug-like structures (three contiguous heteroatoms as an extreme example). Such output structures result from extrapolating speciﬁc rules to general situations, not all of which are relevant, or a practical failure to consid- er every possible situation to which a SMIRKS code may apply. We have opted not to discard the low quality structures: (1) in practice, the number of low quality Total = 186 rules.
structures is small, less than 5% of the output, and nota major nuisance. (2) The unorthodox application ofrules sometimes results in very novel output which mightlead to an ‘outside-the-box' idea. Our current strategy with output lists of less than 200 entries is to let themedicinal chemists themselves judge ‘goodness/badness' of the structures, and the default usage is to not apply any structural or numerical ranking. Further reﬁnement of the rules and/or addition of an optional computation-al ﬁltering step will reduce the number of low quality structures. In cases of output lists with greater than 200 entries or where other scientiﬁc information is avail-able, some ﬁltering and/or ranking of output is needed.
Ranking based on calculated physical property is an op- tion which is further described below.
Since it is possible that diﬀerent rules can result in iden- tical output structures, duplicate entries are grouped in a ﬁnal step prior to display of all of the results to the user in web page format. A typical output page using tworules from is shown in . A ‘history' linkis available to provide information on the structural transformation, for example, scientiﬁc basis and scopeof the transformation, examples of application within medicinal chemistry history including proprietary cor-porate history, and any interesting unpublished lore Figure 2. (a) Illustration of the ring-break transformation (rule #9 in about the transformation. An example history page is (b) Speciﬁc example of the ring-break rule in converting shown in for the hydroxy-to-methoxy rule. In PTK787 to the anthranilic amide analog AAL993.
addition to automatically archiving results in a user areaafter accessing Drug Guru, options for sharing results inspection of this many hits and permit archiving for lat- with other scientists or exporting data are provided.
er study. We selected the commercially available RENEdata analysis tool as one that (1) understands chemin- 2.4. Special features of Drug Guru formatics and Daylight parameters, (2) permits searchand sort capabilities, and (3) uses Oracle-based informa- While the default usage of Drug Guru by the medicinal tion storage for data manipulation and archival.
chemist requires only input of a structure and pressingthe ‘Run' button, there are several user options that Many of the chemical structures that Drug Guru creates are provided on an ‘expert user' page. When an input are reasonable target structures that, admittedly, will be structure has more than one occurrence of the function- obvious to most experienced medicinal chemists. This is al group to which a rule applies, the question of how
K. D. Stewart et al. / Bioorg. Med. Chem. 14 (2006) 7011–7022 Figure 3. (a) Illustration of the ‘NC-switch' rule (rule #10 in (b) Speciﬁc example of the ‘NC-switch' rule in converting the antibacterialdrug ciproﬂoxacin to Figure 4. The structure entry web page of Drug Guru. In this example, the structure of sildenaﬁl is accessed by entering the corporate code,A-325043, for this drug.
many times Drug Guru applies the rule needs to be an- tion experiment is shown in In practice, the swered. The example of carrying out the hydroxy-to- combinatorial expansion leads to a very large number methoxy transformation on the two hydroxy groups of of structures (thousands) for a rule set selection of dopamine, see best illustrates this feature. In 100–200 rules; therefore, the number of generations is the default mode, only the two mono-methylated result- currently limited to four.
ing structures are obtained. Optionally, an exhaustiveapplication of the rule produces a third di-methylated Within Drug Guru, there is an option provided to study structure. This diﬀerentiation between single and the output list according to calculated physical property, exhaustive application of a rule becomes particularly such as log P, rule-of-5, polar surface area (PSA), or important for rules involving replacement of hydrogen rotatable bonds. In test studies with known drugs, no atoms. In this case, selection of exhaustive replacements obvious bias or uneven trend in the calculated physical can lead to a very large number of output structures.
properties of the output molecules was evident: struc-tures with a continuum of both increased and decreased Another user option is to allow multiple generations of physical properties were observed with approximately application of the set of Drug Guru rules to an input equal frequency. A typical distribution is shown in structure. In the default setting, a single round of trans- using Gleevec as input. Sorting the output list formations is applied. Optionally, the output structures according to physical property is provided as a user op- from the ﬁrst round can be automatically re-submitted tion and can facilitate identifying structures of high for a second round of transformations. This has the interest. Of particular value is the mathematical diﬀer- interesting eﬀect of generating quite novel molecular ence in property when comparing the input and output structures for evaluation. An example of a two-genera- structures. When the ‘D-property' column is selected
K. D. Stewart et al. / Bioorg. Med. Chem. 14 (2006) 7011–7022 Figure 5. Example output page from Drug Guru using rules 2 and 3 from The output page contains four columns, with each separate outputstructure as a new row. Column 1 is the name of the rule, column 2 is the generation number, and columns 3 and 4 are the input and outputstructures, respectively, shown as a 2D depiction. Atoms directly involved in the transformation are highlighted in color to facilitate visual inspection.
This example shows only two output structures, but in practice, lists of 50–150 structures are common.
for sorting, the user can quickly tabulate all of the struc- tion. In the application of 186 rules to the structure of tural transformations that give a desired result. For sildenaﬁl, 91 output structures were created by Drug example, users can see all of the changes that lead to Guru. While discussion of the entire output list is be- an increase in C log P or a decrease in PSA. In yond the scope of this publication, a few illustrative a D-PSA calculation is illustrated.
examples will be given.
As a critical aspect of Drug Guru, we note that the rule Five of the 10 rules listed in are applicable to sil- list is not static, and new rules can be added at any time.
denaﬁl and yield output structures represented in For example, new structural transformations reported in . Five rules do not apply because their functional the literature or proprietary research discoveries of poten- groups are not present in the input structure. The tial general applicability are excellent sources of new homologation-C rule (extend every substituent on a ring rules. There is also the capability of creating subsets of by one methylene unit) gives several non-redundant out- rules for special purposes. For example, in addition to put structures and only two, 36 and 39, are illustrated.
the set of general transformation rules described above, The ethano-to-S and ether-to-thioether rules give struc- we have included separate sets of rules for increasing sol- tures 62 and 69, respectively. In this example, the more ubility, decreasing albumin binding, or decreasing meta- interesting and diverse structures result from molecular bolic liability of the input structure. These rules are less framework changes. The NC-switch rule described well documented and more anecdotal in character, but above gives rise to two structures possessing new [6.5] nonetheless, still very useful in practice. As an example heterocyclic core rings, output structures 64 and 65.
of this kind of additional Drug Guru rule, one of the Gratifyingly, structure 64 possesses the heterocycle metabolism rules is a ‘pyridine-block' rule which adds a found in vardenaﬁl (LevitraÒ), a drug in the same phar- methyl group to the ortho-position of a pyridine. This rule maceutical class as sildenaﬁl and shown in .
is based on the strategy to sterically block the facile oxida- Another NC-switch output, structure 65, possesses a tion of pyridines with an ortho substituent. An example heterocycle not previously encompassed within sildenaﬁl from the Roche group of an application of this design rule or vardenaﬁl patents.In another example of a frame- has recently been publis work change, the ring-break rule results in output struc-tures 60 and 78 (only 78 is illustrated) in which thearomatic 5-membered ring of sildenaﬁl is opened to give an amino-ketone structure. An intramolecular hydrogenbond between the amino and carbonyl groups of 78 has 3.1. Example of Drug Guru applied to known drug system the potential of forming a intramolecular H-bond, andthus mimicking the aromatic ring of sildenaﬁl. As dis- Sildenaﬁl (ViagraÒ) is a phosphodiesterase-5 inhibitor cussed above, this ‘virtual' ring strategy has precedent approved in 1998 for treatment of male erectile dysfunc- in other studies. No literature reports of studies of chem- K. D. Stewart et al. / Bioorg. Med. Chem. 14 (2006) 7011–7022 Figure 6. Example of Rule History Page.
ical systems related to structures 60, 65, and 78 in inhi- ed drug design software: docking and scoring programs, bition of phosphodiesterase-5 or utility in treatment of such as DOCK, FlexX, AutoDock, GLIDE, Chem- male erectile dysfunction could be found. Therefore, Score, and GOLD, de novo ligand construction pro- compounds based on these structures may represent grams, such as LigBuilder, SkelGen, Ludi, GrowMol, interesting targets for further analoging around sildena- and SPROUT, scaﬀold-hopping programs such as ﬁl and vardenaﬁl. The exact structure of vardenaﬁl, LeapFrog, EA-Inventor, and FEPOPS, and pharmaco- shown in is produced from sildenaﬁl by Drug phore analysis programs, such as COMFA, Disco, Cat- Guru in a two-generation run when both the homologa- alyst, and GASP.However, these other programs have tion-C and NC-switch rules are applied. It should be operational strategies that are fundamentally diﬀerent emphasized that information from sildenaﬁl structure– from Drug Guru. Drug Guru does not rely on a pre-ex- activity studies was not used in the creation of the rules isting database of ligand structures or utilize energy- of Drug Guru; therefore, these examples shown here are based ﬁtness functions to score or assemble ligands.
truly derived from the history of medicinal chemistry, No prior structure–activity data are required. Drug implemented as computer-coded rules-of-thumb for Guru uses a set of pre-selected and well-deﬁned medici- nal chemistry rules to construct new structures. Rankingis optional and currently is based on calculated physical 3.2. Comparison with other computational chemistry properties. ‘Comparison' of Drug Guru with the com- puter programs listed above is useful mainly in thinkingabout synergy, rather than ranking performance capa- The overall objective of Drug Guru is to facilitate drug bilities. Many of these programs could conceivably be discovery, a goal that is shared with other computer-aid- used to help analyze the output of Drug Guru. We have
K. D. Stewart et al. / Bioorg. Med. Chem. 14 (2006) 7011–7022 Figure 7. Example of an ‘exhaustive' application of a Drug Guru rule. The ﬁrst two output structures result from a single application of the rule. Thelast structure results from application of the rule at all possible sites exhaustively.
Figure 8. Example of a ‘multi-generation' use of Drug Guru. The generations are indicated in column 2. In generation 1, phenol is transformed toanisole by the hydroxy-to-methoxy rule. In generation 2, the anisole formed in generation 1 is transformed to thioanisole by the ether-to-thioetherrule.
K. D. Stewart et al. / Bioorg. Med. Chem. 14 (2006) 7011–7022 Figure 9. Distribution of the C log P values that are calculated for the 116 output structures resulting from using Gleevec as input structure. Gleevechas a C log P of 4.5 and would be located at the center of the distribution.
Figure 10. Example of calculation of physical property of a Drug Guru output structure. The polar surface areas, PSA, of the input and outputstructures are 49.33 and 20.23 A ˚ 2, respectively. The web page reports the PSA of the output along with the mathematical diﬀerence, the D-PSA value, of 29.1. In this case, the decrease in PSA is indicated with a minus sign. This ‘D' column is available for all calculated properties and can be sorted toﬁnd all structural changes that give a consistent change, for example, all transformations that lead to a decrease in polar surface area.
already experimented with taking Drug Guru output in Drug Guru, BIOSTER compiles a large number of directly into docking and pharmacophore programs speciﬁc examples (14,300 bioanalogous pairs in the for further evaluation (work not reported here).
2005.1 release). Drug Guru diﬀers from BIOSTER inshowing the structural changes within the context of a We are aware of four commercial products, BIOSTER, single input molecule, rather than listing literature EMIL, WABE, and BIOISOSTER, developed indepen- examples. The second commercial package, EMIL dently, and possessing some features in common with (example mediated innovation for lead evolution), com- Drug Guru. Unfortunately, published scientiﬁc descrip- bines a database analogous to the BIOSTER database tions of each are limited or unavailable, precluding a de- (3500 ‘optimization schemes' in the 2003 v.2.4 release) tailed comparison. The following brief description is with structure input and viewing that is analogous to intended to draw attention to this area and to stimulate Drug Guru.Unlike Drug Guru, the transformations further research. The ﬁrst commercial package, BIO- are not collectively organized into rules-of-thumb writ- STER (bioisosterism), is a database of pairs of mole- ten in SMIRKS language, but rather each literature cules diﬀering in one structural element—termed example of molecule pairs serves as a separate rule. Like bioanalogous pairs in the original literature Drug Guru, new transformations can be added to EMIL A search of this database will generate literature exam- via programming within the EMIL software. EMIL is ples of pairs of compounds that illustrate many of the installed as a stand-alone application (more recent ver- same transformations encoded within Drug Guru. Rath- sions are optionally web-based) in contrast to both BIO- er than represent the changes as a list of general rules, as STER (requires ISIS environment, Molecular Design K. D. Stewart et al. / Bioorg. Med. Chem. 14 (2006) 7011–7022 Figure 11. Examples of Drug Guru output for the input structure of sildenaﬁl. The location of structural alteration is denoted with an arrow. Thetotal list contained 91 structures, and the arbitrary ranking within the list is designated here.
based scoring.The fourth commercial package, BIO- ISOSTER, transforms an input structure according to approximately 300 ‘target-speciﬁc' changes derived from literature reports of kinase, protease, ion channel, phos- phodiesterase, and nuclear receptor ligands.In addi- tion to these commercial products, publications from the Merck, Novartis, GlaxoSmithKline, Celera Genom- ics, and Organon computational groups mention pro-prietary databases of functional group replacements Figure 12. Chemical structure of vardenaﬁl. This structure is created and implementation within compound design pro- by Drug Guru in a two-generation run using sildenaﬁl as input. The two generations use (1) the homologation-C rule, and (2)the NC-switch rule.
3.3. Limitations of this approach Limited) and Drug Guru (requires Daylight toolkit and From the listing of commercially available packages and Oracle environments). A third commercial package, publications discussed above, it is evident that molecular WABE, generates isosteres of an input molecule accord- replacement rule-based software for drug design is an ing to an atom-replacement algorithm with an output emerging and exciting area of research. In the case of ranking based on electrostatic similarity or receptor- Drug Guru, we feel that there are two main limitations K. D. Stewart et al. / Bioorg. Med. Chem. 14 (2006) 7011–7022 that happen to also be limitations common to other computer-aided drug design programs: (1) imperfectranking scheme and (2) lack of synthetic chemistry In this report, we describe a new web-based computer knowledge. The ﬁrst limitation is related to the age-old application that applies a library of medicinal chemical question of ‘how do you recognize a drug when you transformation rules to an input structure and then per- see it?' The example of the sildenaﬁl analogs illustrated mits evaluation of the resulting output structures. The above is telling. As a chemist confronted with a list of name of the new software captures this intellectual pro- 91 suggestions, how does one know, a priori, that num- cess of drug generation using rules with the acronym ber 64 (the vardenaﬁl analog) is the structure on which ‘Drug Guru.' It is hoped that Drug Guru will provide to focus attention? Also important is the question of an intellectual guide to medicinal chemists in the how good the remaining 90 structures are? Unequivocal increasingly diﬃcult task of discovering new compounds answers to these questions are not possible. The main in- as potential drug candidates.
tent of Drug Guru is to generate ideas. We submit thatthe idea-list that a medicinal chemist has in mind at anyone time is by its nature incomplete, and Drug Guru is designed to help ﬁll that gap. However, idea generationis not usually cited as the rate-determining step in drug Funding for the Drug Guru project was provided by the discovery. The ranking that Drug Guru currently pro- vides is based on calculated physical properties and Acknowledgment is also made to the Abbott Laborato- has acknowledged weaknesses. We speculate that assess- ries Medicinal Chemists who contributed many ideas for ment of ‘success frequency' for individual rules may pro- structural transformations (rules). Cheminformatics and vide a novel way of ranking compounds for future web programming was carried out by Moonview Con- evaluation. Chemists could conceivably prioritize trans- sulting, LLC.
formations that have historically performed well in cre-ating drug candidates, that is, transformations with thebest ‘yield.' Confounding any single prioritization References and notes scheme is the fact that analog design strategies at theoutset of a research project diﬀer from those in a mature 1. Wermuth, C. G., Ed., The Practice of Medicinal Chemis- program close to a clinical candidate. Currently in Drug try; 2nd Ed.; Academic: New York, 2003.
Guru, all design rules are treated equally. We welcome 2. Sneader, W. Drug Prototypes and their Exploitation; John suggestions on the optimal computational protocol for Wiley & Sons: New York, 1996.
diﬀerentially weighting the rules.
3. Burger, A. Prog. Drug Res. 1991, 37, 287–371.
4. Chen, X.; Wang, W. Annu. Rep. Med. Chem. 2003, 38, Assessing synthetic feasibility is an important part of target evaluation. In an ideal setting within any comput- 5. Patani, G. A.; LaVoie, E. J. Chem. Rev. 1996, 96, 3147– er-aided drug design software, the user would receive quick feedback whether suggested output structures 6. Thornber, C. W. Chem. Soc. Rev. 1979, 8, 563–580.
7. Lipinski, C. A. Annu. Rep. Med. Chem. 1986, 21, 283–291.
were conveniently accessible from available starting 8. Spatola, A. Ann. Rep. Med. Chem. 1981, 16, 199–209.
materials, or would require multi-step syntheses with 9. Rudinger, J. In Drug Design; Ariens, J., Ed.; Academic: varying degrees of diﬃculty and precedent. In fact, when New York, 1971; Vol. 2, pp 319–419.
Drug Guru was evaluated with a test audience of 25 10. Drug Guru is a trademark of Abbott Laboratories. The medicinal chemists, incorporation of synthetic chemistry software is proprietary to Abbott Laboratories and knowledge into the evaluation process was the single Moonview Consulting, LLC. Readers desiring more most requested improvement. Unfortunately, a fully sat- information are encouraged to contact the ﬁrst author.
isfactory computational protocol to conveniently, and 11. James, C. A.; Weininger, D.; Delaney, J., Daylight Theory reliably, assess laboratory access to computer-generated Manual, Daylight Chemical Information Systems, Inc., ideas is currently not possible. One simple, yet specula- tive, possibility for Drug Guru would be to assign a ‘dif- 12. Jung, M. E.; Piizzi, G. Chem. Rev. 2005, 105, 1735–1766.
ﬁculty ranking' to each rule. Rules that correspond to a 13. Furet, P.; Bold, G.; Hofmann, F.; Manley, P.; Meyer, T.; common laboratory operation, such as a methylation (the Altmann, K.-H. Bioorg. Med. Chem. Lett. 2003, 13, 2967– hydroxy-to-methoxy rule, rule 3, ), would be ranked ‘high priority.' Rules that correspond to potential- ly laborious multi-step procedures if interpreted literally, 15. Trejo, A.; Arzeno, H.; Browner, M.; Chanda, S.; Cheng, such as ring transformations (for example, NC-switch, S.; Comer, D. D.; Dalrymple, S. A.; Dunten, P.; Lafargue, rule 10, would be ranked ‘low priority.' Subse- J.; Lovejoy, B.; Freire-Moar, J.; Lim, J.; Mcintosh, J.; quent sorting of the output list by the medicinal chemist Miller, J.; Papp, E.; Reuter, D.; Roberts, R.; Sanpablo, F.; would yield a list of the more synthetically accessible sug- Saunders, J.; Song, K.; Villasenor, A.; Warren, S. D.;Welch, M.; Weller, P.; Whiteley, P. E.; Zeng, L.; Gold- gestions high on the list. More extensive software links be- stein, D. M. J. Med. Chem. 2003, 46, 4702–4713.
tween Drug Guru and organic chemistry synthesis design 16. Sildenaﬁl patents US 5250534, US 5346901, US 5719283; programs, such as Lhasa, Chiron, CAESA, etc, or reac- vardenaﬁl patents US 6362178, US 6566360.
tion databases, such as REACCS, can be envisioned. Re- 17. Representative references: docking/scoring programs: (a) search into incorporating synthetic chemistry input into Perola, E.; Walters, W. P.; Charifson, P. S. Proteins: Drug Guru is in progress.
Struct. Funct. Bioinf. 2004, 56, 235–249; De novo pro- K. D. Stewart et al. / Bioorg. Med. Chem. 14 (2006) 7011–7022 grams: (b) Stahl, M.; Todorov, N. P.; James, T.; Mauser, 21. Balakin, K. V.; Tkachenko, S. E.; Okun, I.; Skorenko, A.
H.; Boehm, H. -J.; Dean, P. M. J. Comput. Aided Mol.
V.; Ivanenkov, Y. A.; Savchuk, N. P.; Ivashchenko, A. A.; Des. 2002, 16, 459–478; Scaﬀold hopping programs: (c) Nikolsky, Y. Chem. Oggi 2004, 22, 15–18, The BIOISO- Jenkins, J. L.; Glick, M.; Davies, J. W. J. Med. Chem.
STER program is commercially available from Chemical 2004, 47, 6144–6159; Pharmacophore programs: (d) Patel, Diversity Labs, Inc. Y.; Gillet, V. J.; Bravi, G.; Leach, A. R. J. Comput. Aided 22. Sheridan, R. P. J. Chem. Inf. Comp. Sci. 2002, 42, 103–108.
Mol. Des. 2002, 16, 653–681.
23. Ertl, P. J. Chem. Inf. Comp. Sci. 2003, 43, 374–380.
18. Ujvary, I. Pestic. Sci. 1997, 51, 92–95, The BIOSTER 24. Lewell, X. Q.; Jones, A. C.; Bruce, C. L.; Harper, G.; program is commercially available from Accelry Inc., Jones, M. M.; Mclay, I. M.; Bradshaw, J. J. Med. Chem.
2003, 46, 3257–3274.
19. Fujita, T. In Trends in QSAR of Molecular Modeling 92; 25. Southall, N. T.; Ajay J. Med. Chem. 2006, 49, 2103–2109.
Wermuth, C. G., Ed.; ESCOM: Leiden, 1993; pp 143–159, 26. Wagener, M.; Lommerse, J. P. M. J. Chem. Inf. Model.
The EMIL program is commercially available from 2006, 46, 677–685.
CompuDrug International, Inc, .
27. Li, Q.; Chu, D. T. W.; Claiborne, A.; Cooper, C. S.; Lee, C.
20. Sayle, R., Skillman, G.; Nicholls, A., Abstracts, 227th M.; Raye, K.; Berst, K. B.; Donner, P.; Wang, W.; Hasvold, National Meeting of the American Chemical Society, L.; Fung, A.; Ma, Z.; Tufano, M.; Flamm, R.; Shen, L. L.; Anaheim, CA, March 28–April 1, 2004, American Chemical Baranowski, J.; Nillius, A.; Alder, J.; Meulbroek, J.; Marsh, Society: Washington, DC. The WABE program is com- K.; Crowell, D.; Hui, Y.; Seif, L.; Melcher, L. M.; Henry, R.; mercially available from OpenEye Scientiﬁc Software, Spanton, S.; Faghih, R.; Klein, L. L.; Tanaka, S. K.; Plattner, J. J. J. Med. Chem. 1996, 39, 3070–3088.
JULY 2006 ISSUE Save The Date DRUGS USED TO TREAT BPH MAY September 14 ALSO PREVENT PROSTATE CANCER The KnowledgeNet "In TheKnow" Awards Luncheon by Diane Johnson New York, NYDetails at: ew evidence shows that doxazosin and terazosin (alpha-blockers), currently being used for the treatment of BPH (Benign Prostatic Hyperplasia) and hypertension, may alsodecrease the risk of developing prostate cancer. In addition, they may prevent the