BACKGROUND OF THE INVENTION In the past, the discovery of therapeutic agents was conducted without fully understanding the molecular mechanism of action by which drugs influence biological processes in the organism. In contrast, modern drug discovery places increased emphasis on through understanding of molecular processes that cause disease ("molecular medicine") and also molecular interactions employed by successful drugs.

A large group of drugs currently on the market, and many other chemical compounds being investigated as drug candidates, exert their influence on the molecular level as enzyme inhibitors. An enzyme inhibitor is a chemical that, in a dose-dependent fashion, suppresses the catalytic action of a certain enzyme (a biological catalyst) in the living organism.

Inhibitors suppress the catalytic activity of an enzyme by employing a variety of distinct molecular pathways. For example, so-called competitive enzyme inhibitors can stop a particular biochemical reaction by mimicking a particular substrate-a chemical that is being transformed by the enzyme catalyzed reaction. In the case of competitive inhibition, either the substrate or the inhibitor, but not both, can be attached to the enzyme catalyst molecule. No ternary complex enzyme-substrate-inhibitor is formed.

In another example, so-called partial non-competitive inhibitors act by binding not only to the enzyme catalyst proper, but also to the enzyme-substrate complex. Thus, the reaction mixture contains the ternary enzyme-substrate-inhibitor complex as well as the binary complexes enzyme-inhibitor and enzyme-substrate. Importantly, in the case of partial non-competitive inhibitors the enzyme-substrate-inhibitor complex retains a partial catalytic activity. This means that no matter how high is the amount (concentration) of inhibitor, the enzyme reaction will never be completely stopped. For this reason, partial non-competitive inhibitors are unsuitable as drugs. This is true even for partial non-

competitive inhibitors characterized by very strong overall potency of binding to the enzyme (low inhibition constant or ICso).

According to the commonly established practice in the pharmaceutical industry, the kind of mechanistic investigation that could characterize a given inhibitor as competitive, non- competitive, or partial, is relegated to the later states of pre-clinical testing. The reason is that this type of mechanistic investigation requires a large amount of human effort for the design and interpretation of dose-response enzyme kinetic experiments.

Introduction High cost of pre-clinical R&D during pharmaceutical discovery A press release from the Tufts Center for the Study of Drug Development estimates the average cost to develop a new prescription drug at $802 million. According to the Pharmaceutical Research and Manufacturers of America, approximately 16% of the development cost are attributed to pre-clinical research. Thus the cost of pre-clinical testing, including the studies of molecular mechanism of action, can be estimated at $128 million for each new drug. Due to the high cost of pharmaceutical discovery and development, only 3 out of every 10 drug products (new chemical entities) introduced from 1980 to 1984 had returns higher than their average after-tax R&D costs.

Among the various phases of successful drug discovery, pre-clinical testing and discovery are two areas that are uniquely suited for increased efficiency. Indeed, it impossible to conserve expenses in the areas of testing on human subjects; clinical trials are governed by regulator requirements. Therefore the automation and miniaturization of pre-clinical phases of drug discovery offers unique areas for cost savings.

"The single largest challenge facing drug developers-both pharmaceutical and biotechnology companies-is to contain R&D costs and reduce development times without compromising clinical test design." Dr. Kenneth I. Kaitin Director, Tufts Center for the Study of Drug Development Many drugs act as enzyme inhibitors. Therapeutic enzyme inhibitors are among the most widely used drugs. The following are selected (20) enzyme inhibitor drugs that were among the 200 most widely prescribed drugs in 1999. This list of enzyme inhibitors includes the second most frequently prescribed drug, Lipitor, which is a specific inhibitor of the enzyme HMG-CoA (3-hydroxy-3-methylglutaryl-CoA) reductase (HMGR).

Table 1: Prominent enzyme inhibitor drugs RANK BRAND NAME MANUFACTURER GENERIC NAME ENZYME 2 Lipitor@ Parke-Davis Atorvastatin HMG-CoA reductase 21 Ibuprofen0 Various Ibuprofen COX 27 ZestrilS Zeneca Lisinopril ACE 31 ZocorE Merck Simvastatin HMG-CoA reductase 46 Naproxen@ Various Naproxen COX 47 Coumadin@ Dupont Warfarin Vitamin K-dependent coagulation factors 48 Accupril Parke-Davis Quinapril ACE 49 Pravachol0 B-M Squibb Pravastatin HMG-CoA reductase 50 Viagra@ Pfizer Sildenafil Citrate cGMP-specific phosphodiesterase type 5 (PDE5) 59 Prinivil@ Merck Lisinopril ACE 64 Vasotec0 Merck Enalapril ACE 82 Allopurinol0 Various Allopurinol Xanthine oxidase 93 Warfarin@ Various Warfarin Vitamin K-dependent coagulation factors 98 Monopril@ B-M Squibb Fosinopril ACE 136 Captopril Various Captopril ACE 140 Escot@ Novartis Fluvastatin HMG-CoA reductase 146 Zestoretice Zeneca Lisinopril/HCTZ ACE 159 BaycolE Bayer Cerivastatin HMG-CoA reductase 167 A ! tace@ Monarch Ramipril ACE 172 Naproxen@ Various Naproxen COX Sodium Sodium Not included in this list is a class of enzyme inhibitors currently developed by many major pharmaceutical companies, namely, inhibitors of protein kinases. Protein kinases are enzymes involved in many diseases, including cancer and inflammation. Protein kinase inhibitors are among the most promising therapeutic agents of the future.

Emphasis on Molecular Mechanisms of Action Modern drug discovery requires a good understanding of molecular mechanisms by which therapeutic compounds achieve their biological effects. During previous eras of pharmaceutical discovery, therapeutic agents were discovered largely by trial and error, without regard to molecular mechanism of action. This lack of knowledge regarding

molecular mechanisms of action resulted in the vast majority (more than 90%) of drug candidates eventually being rejected due to toxicity and side effects.

In contrast, proper understanding of molecular mechanisms of action allows precise targeting of particular biochemical molecules, which then selectively interact with the drug.

Thus, the increased selectivity of bio-molecular targeting ("designer drugs") is directly related to the depth of knowledge regarding molecular mechanisms.

Regulatory agencies, such as the U. S. Food and Drug Administration, explicitly require that certain aspects of the molecular mechanism by which a drug acts are documented in an application to market a new drug. Thus, for example, the FDA mandates in Title 21 of U. S.

Code, Part 314:"If the drug is an anti-infective drug, [the application must contain] a section describing the microbiology data, including [a] description of the biochemical basis of the drug's action on microbial physiology." Although new medications other than anti-infective agents are not specifically mentioned in the law, it is understood that all reasonable efforts have been made to understand the molecular basis of drug action for every new medication being introduced on the market.

Mechanisms of Enzyme Inhibition Effective discovery of therapeutic enzyme inhibitors relies on detailed knowledge of molecular mechanisms by which the inhibitors exert their therapeutic effect.

SUMMARY OF THE INVENTION Here we describe a method and a system for disclosing enzyme inhibitors that are characterized by a particular mechanism of action. In the context of this document, the term"mechanism of action"is defined on the biochemical (molecular) level. In particular, the mechanism of action for an enzyme inhibitor is determined by the particular type and order of molecular interactions between the enzyme, the substrate, and the inhibitor, and possibly other molecular components such as activators.

The purpose of the present invention is to perform automatic biochemical analyses of a number of chemical compounds that are considered as drug candidates and that act as enzyme inhibitors.

This purpose is obtained by a method comprising the combined usage of a number of hardware components and a number of software components, said usage consisting in the number of hardware components combining an enzyme and a substrate molecule, and the number of software components monitoring a relationship between biochemical reaction

rates and molecular reaction mechanisms, when the substrate converts into a product and when the enzyme regenerates.

According to the present invention, we thus propose a method that makes fully automatic the process of disclosing inhibitors that strictly adhere to a particular molecular mechanism of action, for example, inhibitors that are competitive with respect to a particular substrate. In this fashion, the effectiveness of pre-clinical testing dramatically increases, which results in cost savings and in accelerated drug discovery.

As a result of the automated analysis, the pool of drug candidate compounds is evaluated not only for overall inhibitory activity, which is the commonly accepted practice in the drug discovery industry, but simultaneously the compounds are also categorized according to their most likely molecular mechanism of action.

The advantage is that only compounds that satisfy a pre-determined requirement on molecular mechanism of action can be taken to further states of drug discovery and development. Thus, the knowledge of molecular mechanism of inhibition saves valuable resources by eliminating from the discovery stream those chemical compounds that might exhibit toxic side effects in later stages of drug development.

The invention provides a method for the discrimination among alternate mechanistic models, and for identifying in the collection of enzyme inhibitors those that conform to a particular, preferred molecular mechanism of action ("mechanistic screening"). It is essential to this invention that all steps in the entire screening protocol are fully automatic, directed by a dedicated software program and executed by automated laboratory equipment (either a conventional laboratory robot or a specialized micro-fluidic device).

The method involves measuring the initial reaction rates of enzyme reactions in dependence on the varied concentrations of inhibitors and substrates. Results of these enzyme kinetic experiments are analyzed by using statistical methods for discrimination among mathematical regression models. The stepwise process begins by initially selecting a certain starting experimental design (i. e. , a set of concentrations for the substrates and inhibitors). The choice of the initial design is made empirically based on physical properties of substrates and inhibitors, such as solubility.

If and when an unequivocal decision among several molecular mechanisms can be made immediately after the first round of experiments, the analysis is terminated and the appropriate molecular mechanism for the given inhibitor is reported, along with the quantitative characteristic of inhibitory potency (namely, inhibition constants).

In a typical case, unequivocal decision among possible mechanistic models (e. g., competitive inhibition versus partial non-competitive inhibition) is not possible immediately after the first round of dose-response experiments. In such case, the invention employs statistical methods, based on the theory of Optimal Experiment Design, to select for the next round of experiments such concentrations of substrates and inhibitors that are optimal for model discrimination.

This three-part process of (1) optimal experimental design, (2) experimentation, and (3) data analysis is repeated cyclically until a decision can be made among candidate mechanisms, or until a certain maximum number of experiments has been reached without reaching conclusion on mechanism. In such case the device embodied by this invention reports to the investigator that the given inhibitor follows an unknown mechanism, not on the list of specified alternatives.

In a first embodiment, this invention provides a method of mechanistic screening by using a multi-purpose laboratory robotic equipment that has been modified to allow automatic, "on the fly"re-configuration in each iterative cycle of the screening protocol. In particular, the laboratory robot delivers optimal quantities (concentrations) of the enzyme, the substrates, and the inhibitor under consideration based on the Optimum Experimental Design software module.

In a second embodiment, this invention utilizes a dedicated micro-fluidic device that performs the same functions as described above for the laboratory robotic system.

The essence of the method is in tight integration between two separate classes of components, namely, the hardware component and the software component. The hardware component consists of a device for the precisely controlled mixing of biochemical reactants (enzymes, substrates, and inhibitors), and a device for the monitoring the course and extent of enzyme reactions. These hardware components may utilize either classical laboratory equipment, or they may utilize special purpose micro-fluidic equipment. The software component consists of a data analysis module and an experiment design module.

The essence of the system is to automate and miniaturize the model discrimination analysis described below in steps 1 through 5 of the experimental method. Thus, we can ascertain in a very short time for many chemical compounds their molecular mechanism of action, and (using pre-determined mechanistic criteria).

Thus, the main benefit of the method and of the system is that it allows the screening of candidate enzyme inhibitors according to a chosen mechanism. From a collection of synthetic or naturally occurring inhibitors, the system can select those that conform to a particular mechanism of action. For example, the system can select from thousands of 3- HMG-CoA reductase inhibitors those chemical entities that are competitive with 3-HMG- CoA. Similarly, the system can select from many candidate compounds those inhibitors of protein kinases that are competitive with ATP.

DETAILED DESCRIPTION OF THE INVENTION In this section we provide a detailed description of the invention and its application to drug discovery. Fig. 1 schematically shows a system consisting of an automated experimental system consisting of two sub-systems: (1) a hardware sub-system and a (1) software sub- system. The communication between these two sub-systems is accomplished by way of a communication interface.

Fig. 2 is a flow-chart of the experimental method used by the system to distinguish between possible molecular mechanisms of inhibition proceeds, which is as follows : 1. Measure the reaction velocity v, defined as the rate with which enzymatic reaction product P is formed, at various suitably chosen concentrations of the substrates and the inhibitor.

2. Perform a regression analysis (a"fit") of the experimental reaction velocities using suitable mathematical models.

3. Assess the goodness of fit of the experimental data to candidate theoretical models, i. e. , kinetic equations, by using established statistical procedures.

- If the concentrations were sufficiently well chosen, only one of the theoretical equations will fit the experimental data, whereas the other equations will lead to a pronounced lack of fit. The particular molecular mechanism that corresponds to a well- fitting rate equation is considered as the true model, and the investigation can be terminated.

- In many practical situations the concentrations of the substrates and the inhibitor that were initially chosen do not allow conclusive discrimination between mechanisms, because the corresponding rate equations fit the experimental data almost equally

well. In such a case, a new set of concentrations is chosen for the next series of velocity measurements as is described in Step 4.

4. Using the Optimal Experiment Design theory of statistics, choose concentrations of substrates and the inhibitor in such a way that any possible differences between candidate reaction mechanisms is amplified.

5. Go to Step 1 above, and repeat the procedure as many times as necessary.

It is an essential feature of the method that all functions described above, including Optimal Experimental Design (i. e. , making a decision on how to perform the next series of experiments), are fully automated.

Example : HMGR Inhibitors Fig. 3 shows the structure of Lipitor. Lipitor is the second most frequently prescribed drug world-wide. It suppresses the synthesis of cholesterol in the body, by inhibiting the enzyme 3-hydroxy-3-methylglutaryl-CoA reductase (HMGR). HMGR catalyzed the reduction of 3-hydroxy-3-methylglutaryl-CoA to mevalonate according to the scheme shown in Fig.

4. Thus, fig. 4 schematically shows an enzyme reaction inhibited by Lipitor.

Fig. 5 shows schematically inhibition of the HMGR enzyme by a hypothetical inhibitor competitive with NADPH. It is important to note that the enzyme to be inhibited, HMGR, utilizes two different substrates: (1) 3-hydroxymethyl-glutaryl-CoA, and (2) NADPH (see Fig. 4 and 5). The second of these two substrates, NADPH, is a ubiquitous cofactor, which is utilized by many different enzymes in the body. Assume for the sake of discussion that a hypothetical inhibitor of MHGR would exert its inhibitory effect by way mimicking the NADPH substrate.

The problem with a hypothetical inhibitor such as the one shown schematically in Figure 5 is that NADPH is needed by many different enzymes in the body. Therefore, it is very likely that any such inhibitor of HMGR would also inhibit various different enzymes. For this reason, it is likely that the inhibitor would show undesirable side effects when applied as a drug.

In contrast, consider the scenario shown in Fig. 6. Fig. 6 shows schematically Lipitor being competitive with 3-HMG-CoA, instead with NADPH substrate. No other enzyme is capable of processing 3-HMG-CoA as shown in Fig. 3. Thus, Lipitor is very selectively inhibiting only the HMGR enzyme and therefore has few side effects. The 3-HMG-CoA-competitive mode of inhibition was directly demonstrated by Istvan et. al.

The above example system demonstrates the general importance of understanding the molecular mechanisms of inhibition in the disclosing of enzyme inhibitor drugs. It should be emphasized that each enzyme as a therapeutic target is unique, and therefore the ideas presented here for Lipitor may not be applicable directly.

For example, a large class of enzymes called protein kinases also utilizes two different substrates, the first of which is very specific for each given kinase, while the second substrate (namely, ATP) is a ubiquitous cofactor used by many different enzymes.

Regardless of this fact, one school of kinase inhibitor design emphasizes ATP competition.

For a thorough review of structure-based design of ATP-site directed protein kinase inhibitors. However, what all strategies for disclosing of enzyme inhibitors have in common is that understanding of molecular mechanisms of action is a very important requirement.

Methods As mentioned, the system consists of both hardware components (i. e. , instrumentation) and software components (numerical algorithms as shown in fig. 1.

Enzyme Kinetics The scientific foundation of the system relies on enzyme kinetic analysis. Enzyme kinetics is a branch of quantitative enzymology, which studies the relationship between (macroscopic) biochemical reaction rates and (microscopic) or molecular reaction mechanisms. Thus, enzyme kinetic methods provide a unique opportunity to investigate the fine details of enzyme-inhibitor interactions on the molecular level, while observing such macroscopic quantities as the bulk change in absorbance (optical density) of the reactant solution.

This section illustrates basic principles of enzyme kinetic analysis on very simple examples of enzyme inhibition. An enzyme (E) is a biochemical catalyst that combines with a substrate molecule (S) and converts it into a product molecule (P). Scheme 1 shows a simple molecular mechanism of enzyme catalysis, involving a single intermediate complex ES.

E+S = ES ~ E+P Scheme 1 This molecular mechanism can be translated from its graphical representation into plain language as follows : "In the first step of the mechanism, an enzyme molecule combines

with a substrate molecule, reversibly forming the intermediate complex ES. In the second step, the intermediate complex ES breaks down to form the product and regenerated enzyme catalyst." Competitive vs. Partial Non-competitive Enzyme Inhibition Enzyme inhibitors can act through a variety of molecular mechanisms. Some mechanisms are very desirable for a practicaily useful drug, while other mechanisms are undesirable. In the most general sense, the present invention is concerned with disclosing those enzyme inhibitors that follow a desirable molecular mechanism.

One generally desirable molecular mechanism of enzyme inhibition is the competitive inhibition mechanism shown in Scheme 2. According to this mechanism, the inhibitor (I) creates a bond with the free enzyme (E), but not with the enzyme-substrate complex (ES).

The desirable feature of this mechanism is in that at sufficiently high concentration of the inhibitor, the enzyme will be completely inhibited.

E+S <='i E-S--E+P + <BR> <BR> <BR> <BR> I<BR> <BR> <BR> <BR> <BR> <BR> il<BR> <BR> <BR> <BR> <BR> <BR> <BR> E-l Scheme 2 On the other hand, the partial non-competitive inhibition mechanism (Scheme 3) is undesirable for a therapeutic inhibitor. According to this mechanism, the inhibitor combines not only with the free enzyme (E) but also with the enzyme-substrate complex (ES). Importantly, the ternary complex"enzyme-substrate-inhibitor" (ESI) retains a partial catalytic activity, thus leading to the formation of product P even at very high (in principle, infinitely high) concentrations of the inhibitor.

E + S E-S-E+P + + <BR> <BR> <BR> <BR> # #<BR> <BR> <BR> <BR> <BR> <BR> # #<BR> <BR> <BR> <BR> <BR> <BR> <BR> E#I + S # E#S#I # E#I + P Scheme 3 As a consequence of this partial inhibition mechanism, if the inhibitor (I) were used as a drug with the intent to inhibit an enzyme involved in a particular disease process, the

dosage would have to be extremely high and, even more importantly, the inhibition would still be incomplete. This drug candidate would mostly likely become one of a great majority of drug candidates that are never developed into a successful commercial product.

Experimental Investigation of Enzyme Inhibition Mechanisms A given inhibition mechanism is an intrinsic microscopic property (at the molecular level) of an enzyme inhibitor. However, we can study this molecular-level behavior of enzyme inhibitors in macroscopic experiments, by using the methods of experimental enzyme kinetics.

Each microscopic enzyme mechanism corresponds to a specific mathematical relationship between the concentrations of reactants (namely, the substrate S and the inhibitor I) and the macroscopic reaction velocity. Reaction velocity (v) is the most basic concept in chemical and biochemical kinetics. It is defined as the concentration of product P formed over a unit of time, or, equivalently, as the concentration of substrate S that is consumed over a unit of time.

Traditional formal enzyme kinetics provides several methods that can be used to derive a mathematical relationship between the reaction rate (v) and the concentration of substrates or inhibitors. Additionally, modern methods of enzyme kinetic analysis have been developed that do not rely on traditional algebraic formulas.

Using the traditional mathematical formalism, competitive inhibition in Scheme 2 corresponds to equation (1), whereas mixed-type non-competitive inhibition in Scheme 3 leads to equation (2). In these equations, kcat, Km, ß, and Ki are adjustable kinetic constants, and the terms in square brackets are macroscopic concentrations usually expressed in moles per liter.

The essential difference between these two mathematical models is that in equation (1) the reaction velocity v tends to zero as [I] increases to very high values, whereas in equation (2) the reaction velocity v tends to a non-zero value at very high [I]. In other words, two different microscopic reaction mechanisms (Scheme 2 and 3) lead to distinctly different macroscopic mathematical models for observable reaction velocity, namely,

equations (1) and (2). Thus, an experimental method involving the above two equations in the formerly listed steps of the method, the method can be used to distinguish between two molecular mechanisms of inhibition, which proceeds as follows : 1. Measure the reaction velocity v, defined as the rate with which product P is formed, at various suitably chosen concentrations of the substrate, [S], and the inhibitor, [I].

Perform a regression analysis (a"fit") of the experimental reaction velocities using in equation (1) and equation (2).

2. Assess the goodness of fit of the experimental data to candidate theoretical models, i. e., equations (1) and (2), by using established statistical procedures.

- If the concentrations were sufficiently well chosen, only one of the equations (1) or (2) will fit the experimental data, whereas the other equation will lead to a pronounced lack of fit. The particular molecular mechanism that corresponds to a well-fitting rate equation is considered as the true microscopic model, and the investigation can be terminated.

- In many practical situations the concentrations [S] and [I] that were initially chosen do not allow conclusive discrimination between mechanisms, because the corresponding rate equations (in this case (1) and (2) ) fit the experimental data almost equally well. In such a case, a new set of concentrations is chosen for the next series of velocity measurements as is described in Step 4.

4. Using the Optimal Experiment Design theory of statistics, choose concentrations or reactants [S] and [I] in such a way that any possible differences between candidate reaction mechanisms is amplified.

5. Go to Step 1 above, and repeat the procedure as many times as necessary.

This procedure for model discrimination is applicable if the investigation were limited to only two mechanisms specifically mentioned above (i. e. , competitive inhibition vs. mixed type non-competitive inhibition). In the case of a greater number of possible molecular mechanisms, the same experimental procedure is used, while other appropriate rate equations are added to the list of candidate mathematical models.

Example : Factor Xa Inhibitor with complex mechanism of action We have recently disclosed an unusual inhibitor of the blood coagulation Factor Xa. Factor Xa is an enzyme involved in the formation of blood clots upon injury of blood vessels. The

blood-feeding hookworm Ancylostoma Ceylanicum secretes a small protein inhibitor of Factor Xa which, we hypothesize, enables the hookworm to feed.

Figure 7 shows the graph of reaction velocity (v) for an enzyme reaction catalyzed by the blood coagulation Factor Xa, in dependence on the concentration of the newly disclosed inhibitor ([rAceAP1]). Initial velocities of an enzyme reaction catalyzed by the blood coagulation Factor Xa (v) plotted against the concentration of the inhibitor rAceAP1. Thin dashed curve is the best fit of initial velocity data to Equation (1), corresponding to the reaction mechanism in Scheme 2. Thick solid curve is the best least-squares fit to Equation (2), corresponding to the reaction mechanism in Scheme 3. The experimental data (open circles) were initially fit to Equation (1) above, under the assumption that the inhibitor might be competitive with the substrate. The results of the non-linear least squares fit are shown as the thin dashed curve in Figure 7. We see that Equation (1) does not fit the experimental data, because the open circles and the thin dashed line are not overlapping.

We have used specialized software, DYNAFIT, to disclose that the proper mechanism, which does fit the experimental data very well (see the thick solid curve in Figure 7), is the partial mixed-type non-competitive mechanism shown in Scheme 3 and represented mathematically by Equation (2).

In contrast, we have found that a related inhibitor of Factor Xa (code name rAcAPS) does follow the competitive inhibition mechanism, because the initial reaction velocities fit very well to Equation (1). This is shown in Fig. 8, which is the graph of initial velocities of an enzyme reaction catalyzed by the blood coagulation Factor Xa (v) plotted against the concentration of the inhibitor rAcAPS. The curve is the best fit of initial velocity data to Equation (1), corresponding to the competitive inhibition mechanism in Scheme 2.

Let us assume for the sake of discussion that both inhibitors (rAcAPS and rAceAPl) were included in a large set of Factor Xa inhibitors. The purpose of a hypothetical screening program could be to select from the pool of candidate compounds those that are suitable for the development into drugs.

A basic requirement for a potent enzyme inhibitor is that it can inhibit the target enzyme completely (as does rAcAPS), not only partially (as is the case for rAceAPl). The system would flag the inhibitor rAcAP5 as suitable for further study, whereas the partial inhibitor rAceAP1 would be flagged as unsuitable for further development. Thus, the two enzyme inhibitors would be ranked and clearly distinguished on the basis of their molecular mechanism, even though their overall potency is comparable.

Type of experimental data The experimental data that are needed for the system are initial reaction velocities of enzyme reactions in dependence on the concentrations of reagents. In other words, the experimental data are dose-response curves. For model discrimination analyses, the reproducibility of the experimental data (i. e. , the measured reaction velocities) is required to be better than 5%. Most typically the varied reagent is be the inhibitor and the substrate (s).

Hardware Sub-system The hardware subsystem of system is embodied using instrumentation for experimental enzyme kinetics. In particular, it requires the following components.

Reactant Storage. The enzyme, the substrate (or substrates, if more than one is involved), and the inhibitors are stored in solution of biochemical buffers. It is important that the enzyme is stored so as to minimize protein denaturation, which typically means storing the enzyme solution at low temperature (4°C) and avoiding contact with metal surfaces.

Liquid delivery. Technical means of delivering precise quantities of liquid for subsequent mixing. It is important to be able to deliver variable amounts of each component (especially inhibitors) within approximately 5% accuracy/reproducibility.

Reactant Mixing. All reactants must be thoroughly and rapidly mixed in order to initiate the enzyme reaction.

Reaction Monitoring. The enzyme reaction must manifest itself in some suitable way, such as by a change of absorbance (optical density) or fluorescence upon the conversion of the substrate into product.

In the final implementation of the system, these hardware functions will be embodied in a microfluidic device. However, for a proof of concept the hardware sub-system can be implemented more simply by utilizing a typical programmable laboratory robot. It is essential that the robot can be programmed with sufficient flexibility, so that the amounts of reaction components can be changed on the fly in each cycle of the iterative experiment design (see below).

Software Sub-system "In the 100 years to 1995, the pharmaceutical industry worked on about 500 targets with a limited number of compounds, whereas now, using new technologies like genomics, high

throughput screening and combinatorial chemistry, drug companies will see an explosion in the number of targets and leads [they] can explore. To improve the transition from research to development it is necessary to automate the research and development process [...] using information technologies to make better use of existing data." J. Kuhlman Bayer AG Pharma-Research-Center Kinetic Simulation Module In order to generate the theoretical curves for the modeling of experimental data, such as the smooth curves shown in Figures 7 and 8, the software sub-system contains a chemical kinetic simulation engine. Here we utilize the experience in biochemical kinetic modeling, embodied in the software program DYNAFIT [Kuzm9660] developed by BioKin Ltd.

Non-linear Regression Module The data analysis module in consists of an algorithm for non-linear least-squares regression. This system is built on the standard theory of non-linear regression analysis [Marq6331].

However, the software system for non-linear regression analysis contains several fundamental innovations, such as the estimation of full confidence intervals on model parameters [Bate88, p. 302] rather then merely the formal standard errors.

Experiment Design Module This module is based on the statistical theory of optimal experiment design [Box6523, Box6757, Hunt6507. Atki92]. The basic idea is that there exists a unique, optimum experimental design for the discrimination among mechanistic models.

In this case, the mechanistic models of enzyme kinetics represent the dependence of initial reaction velocity of enzyme reactions upon the concentration of reactants (enzymes, substrates, and inhibitors). Therefore, the Experiment Design software module within the system chooses the optimal concentrations of reactants.

The integration of the Experiment Design module with the remaining components of the system are shown in Figure 2.

Communication Module

The communications module is responsible for two-way transfer of information between the software sub-system and the hardware platform, either a conventional laboratory robot or a dedicated microfluidic device.

In each cycle of the iterative experimental design (see Figure 2), the communications module instructs the hardware to deliver optimal amounts of reactants (enzyme, substrates, and inhibitors) into a series of enzyme assays. After the assays are complete, the communication module is responsible for transferring the experimental data into the data analysis module.

Glossary - Activator : A chemical compound that causes the increase in biochemical activity of a particular enzyme.

- Binding : (Molecular Binding) The formation of chemical bonds, possibly non-covalent, between molecules. A typical example is the intermolecular bond between an enzyme and an inhibitor.

- Catalysis : The acceleration of chemical or biochemical reactions which is caused a catalyst. Chemical reactions in living systems would proceed extremely slowly without catalysis.

- Catalyst : A special chemical which causes the acceleration of chemical or biochemical reactions. In biochemical systems the catalyst is often an enzyme.

- Concentration : The amount of a particular chemical compound (typically measured in moles) present in a unit volume.

- Competitive Inhibition: A molecular mechanism in which an inhibitor (I) binds to a particular enzyme (E) in such a way that it strictly"competes"for binding with a particular substrate (S). As a result, the biochemical mixture contains only the binary complexes EeS and E. I, but not the ternary complex E. S. I. The definition is more complex in the case of multi-substrate enzymes.

- Enzyme : A protein which acts as a biological catalyst. All living organisms contain thousands of enzymes, many of which are involved in disease processes. A disease is often caused by overproduction, or by increased activity, of a particular enzyme. In this case an enzyme inhibitor can be used as a therapeutic agent.

- Inhibitor : A chemical that causes a decrease in the catalytic activity of a particular enzyme. Inhibitors cause this decrease in catalytic activity by way of molecular binding to enzymes.

- Mechanism : (Molecular Mechanism) The particular type and ordering of inter-molecular interactions occurring in a biochemical system. For example, the competitive inhibition mechanism is characterized by the following scheme, where E = enzyme, S = substrate, P = product, and I = inhibitor.

E+S = ES E+P + <BR> <BR> <BR> <BR> #<BR> <BR> <BR> <BR> <BR> <BR> <BR> # E I - Multi-substrate Enzyme: A biochemical catalyst which causes the simultaneous chemical transformation of at least two substrates. Typically, more then one product is also formed.

- Non-competitive Inhibition: A particular molecular mechanism of enzyme inhibition whereby the inhibitor (I) binds both with the free enzyme molecule (E) and with the enzyme-substrate complex (E. S, see scheme below). The ternary molecular complex E#S#I has no catalytic activity and thus does not give rise to the product P.

E+S ES-E+P + + <BR> <BR> <BR> <BR> # #<BR> <BR> <BR> <BR> <BR> <BR> # #<BR> <BR> <BR> <BR> <BR> <BR> <BR> E I + S # E S I - Partial Inhibition: A particular molecular mechanism of enzyme inhibition which is very similar to the non-competitive inhibition mechanism, except for the fact that the ternary molecular complex E. S. I retains some catalytic activity (see scheme below). The product (P) is formed even at extremely high inhibitor concentration. Thus, any partial inhibitor would be very ineffective for therapeutic purposes.

E + S ES E+P + + <BR> <BR> <BR> <BR> # #<BR> <BR> <BR> <BR> <BR> <BR> # #<BR> <BR> <BR> <BR> <BR> <BR> <BR> E I + S # E S I # E I + P - Product : See substrate.

- Substrate : A chemical compound that is being transformed into a different compound called a product, by the action of a biochemical catalyst (i. e. , enzyme).