WO2004010347A2

WO2004010347A2 - Method and installation for solving problems of adaptive chemistry

Info

Publication number: WO2004010347A2
Application number: PCT/EP2003/007895
Authority: WO
Inventors: Peter Zinn
Original assignee: Peter Zinn
Priority date: 2002-07-20
Filing date: 2003-07-18
Publication date: 2004-01-29
Also published as: AU2003257481A1; DE10233022B4; DE10233022A1; AU2003257481A8; WO2004010347A3

Abstract

The invention relates to a method and an installation for the adaptive analysis or synthesis of compounds and substance mixtures in the field of chemistry. According to the invention, random initial populations are produced and examined using measuring devices. The results so obtained are compared with a target and a fitness value is determined as a measure for the correspondence with the target. Genetic algorithms are used on the basis of said fitness values for determining the composition of the population of the next generation. This method is repeated throughout as many generations as required until the determined fitness values show for at least one individual of the population a sufficiently high correspondence with the target. The inventive adaptive chemistry can be used in the field of finding product formulations or for finding new compounds with desired properties.

Description

Process and system for solving tasks in adaptive chemistry

The present invention relates to a method and a plant for the analysis and production of mixtures and compounds. An iterative process is used to optimize the analysis and the manufacturing process.

The development of new products in the chemical and pharmaceutical industry is associated with a considerable research and development effort. In the cosmetics and perfume industry, too, one is often confronted with complicated questions when analyzing and recomposing substances and mixtures of substances. A typical example is the development of new color compositions or fragrances that are composed of a large number of individual components. The effort involved is often so great that high costs arise, the amortization of which is not always ensured. Since the usability of a new development can usually only be assessed after a test phase lasting several years, especially in the pharmaceutical industry, research work in this area is associated with a considerable financial risk.

Classical chemistry with its analytical approach, which is based on the fact that generally applicable principles are derived from experimental observations and used for the production of new molecules, seems to be reaching its limits. In particular, a typical organic chemist is only able to handle a very limited number of new compounds per day

BESTATIGUNGSKOPIE to manufacture, which must then be checked for their usability.

In order to overcome the problems described, various approaches to automation in the context of chemical synthesis have been taken in recent years. Today, combinatorial chemistry means that substance libraries are accessible, which include tens or hundreds of thousands of substances.

However, this method also has considerable disadvantages. The investigation of such a large number of substances is associated with immense effort. In addition, the diversity of such a combinatorial substance library is inherently limited, since as a rule only substituents are varied around a fixed basic structure or one has to restrict oneself to substance classes in which the molecules are made up of a large number of individual monomer units, such as z. B. peptides or oligonucleotides. Associated with the limited diversity is a reduced likelihood of finding substances with the desired activity compared to purely random substance libraries.

The aim of the invention is therefore to present a new method which overcomes both the limitations of classic chemistry and the problems of combinatorics. At the same time, an installation for carrying out the method is to be presented.

The object is achieved according to the invention by a method of adaptive chemistry, which combines combinatorial chemistry with principles of goal orientation and self-organization, and by the corresponding system. The invention is applicable to both chemical analysis and synthesis problems.

The invention is first described for chemical analysis. To analyze a mixture of substances, the components of which are part of a total of known, possible components that are in an unknown relationship to one another, a starting population of random substances is first mixed from at least two components of the entirety in different mixing ratios. Then at least one measured variable is determined for each of the substance mixtures produced and compared with the corresponding measured variable of the substance mixture to be analyzed. Typically, this is a spectroscopic method; however, other methods are also conceivable. From the comparisons of the measured variables with those of the substance mixture to be analyzed, a fitness value is derived for each member of the population as a measure of the agreement. As a rule, the mixtures of the starter generation are little adapted to the sample. However, since some blends have slightly better fitness levels than others, next generation blends can be determined using genetic algorithms. Now a new generation of substance mixtures is generated, with which the procedure is as before. This procedure is continued over so many generations until the fitness value for at least one member of the population demonstrates a sufficiently large correspondence with the substance mixture to be analyzed.

If necessary, the search space of the mixtures can be limited by boundary conditions, if these are known, during the production of the start generation.

By using genetic algorithms, the number of combinations of mixture spectra to be produced can be significantly reduced compared to purely combinatorial analysis. For example, if you carry out a 5-component analysis with a spectrometer that allows the differentiation of 128 concentration levels, you get 128 ⁵ = 34.4 billion combinations of mixture spectra that would have to be measured in order to ensure the sample spectrum with 100% certainty hold true. In the case of the adaptive analysis according to the invention, for a typical example in which more than 15 generations of substance mixture populations with 16 members each are generated, only 240 measurements would have to be carried out.

A wide field of applications opens up for the described adaptive analysis, particularly in finding product formulations. For example, the analysis of a perfume by adaptive adjustment is one Library of fragrances possible using odor sensors or the like.

In addition to the possibility of adaptive analysis described above, a further feature of the invention is that adaptive chemistry can also be used for synthesis purposes. The main difference to the analysis method described above is that the synthesis cannot be adapted to a size of the test substance, since such a quantity is not available. Accordingly, the goal orientation must be done in a different way.

In the case of adaptive synthesis, however, a starting population of compounds or mixtures of substances is first generated before certain measured variables are determined for each member of the population. For each member of the population, a fitness value is now determined by comparison with a target as a measure of the agreement with the target. By using genetic algorithms, as a rule some members of the population happen to have a slightly better fitness than others, the mixing ratios for producing the next generation of compounds or mixtures of substances can be derived. After a new generation of substance mixtures or compounds has been produced, the procedure is the same as before. This process is repeated over so many generations until a sufficiently large correspondence with the target value via the fitness value is determined for at least one compound or substance mixture.

As with adaptive analysis, the search space for the compounds or mixtures of substances can also be limited here by known boundary conditions.

An example of adaptive synthesis is the development of dyes in which certain color coordinates serve as the target. Starting from a basic framework, e.g. B. different side chains can be coupled. The properties of the synthesized dyes are better adapted to the given color coordinates from generation to generation. In the field of drug research, antibody-antigen interactions are possible as targets, which can be detected via immunoassays. The use of genetic algorithms makes it possible to "breed" new active substances.

Both in the adaptive analysis and in the adaptive synthesis, when comparing the measured variables determined with those of the substance mixture to be analyzed or the target and when determining the composition of the substance mixtures or the conditions for producing the compounds or substance mixtures of the next generation, further, characteristic parameters are taken into account for the components or starting materials used. As such z. B. the costs, MAK values, ecological data, the manageability and / or the availability of the components or starting materials. The additional consideration of such parameters or properties of the components or starting materials allows the particularly advantageous production of compounds or mixtures of substances which meet the required conditions, since the compounds or mixtures of substances can also be optimized in this way with regard to the additional parameters.

By taking additional parameters into account, the properties of given substance mixtures or compounds can be imitated without automatically having to be identical with regard to the starting materials used. The mixtures or compounds newly produced by means of imitation can therefore u. U. be advantageous compared to the educts for the original mixture or the original compound, for example with regard to the educt costs or environmental friendliness.

An example of such an imitation of the target is the adaptive adaptation of a library of fragrances to a given perfume. By taking into account further parameters for the starting materials used, the process according to the invention makes it possible to find a mixture of substances which largely resembles the specified product in terms of smell, but which contains less expensive or more skin-compatible components. A color mixture can also be produced in a similar way, although in with regard to the color itself is practically identical to the target, but uses different starting materials.

According to the invention, genetic algorithms are used to solve the problems. In contrast to conventional algorithms, which are designed to solve a given problem exactly, the genetic algorithms imitate nature in order to find the "best" solution. They work according to the principles of selection known from natural evolution , Crossing and mutation: In this way, better and better solutions are generated over time, ie from generation to generation.

In nature, when two animals are mated, their genetic information is combined and reassembled. As a result of this crossbreeding process, the offspring are similar to their parents, but not identical. In addition, spontaneous changes (so-called mutations) can occur, which also change the genetic information. This can e.g. B. happen through external environmental influences.

Finally, a natural selection takes place as a further step. The survivability of living beings with their genetic information, which due to crossings and mutations have a random composition, depends heavily on the adaptation to the prevailing environmental conditions. To put it simply, it can be said that the most viable specimens have the greatest number of offspring and will therefore have the greatest influence on the development of the genus. Over the course of a multitude of generations, the combination of the characteristics of crossing, mutation and selection finally leads to living beings that are quite well adapted to their environmental conditions. By using genetic algorithms, one tries to transfer natural evolution to the computer. They are used to solve optimization problems in which an exact solution cannot be ascertained with reasonable effort and the best possible solution is to be found out of an initially very large number of possible solutions. To do this, the computer first creates a number of possible solutions. This set becomes population called, their components are called individuals. Now the evolutionary principles described above are applied:

Selection: natural selection is simulated by removing the worse individuals from the population.

Intersection: new individuals are made up of two individuals each.

Mutation: Mutations are caused by selective changes in individuals.

The selection runs in two steps. In the first step, the quality of adaptation of each individual is represented by a numerical value, the fitness. The larger this number, the better the corresponding individual is adapted to the requirements. The actual selection takes place in the second step. Each individual survives with a certain probability, the higher the fitness, the higher.

Since a large number of process steps are frequently repeated in the context of adaptive chemistry, it is advisable to automate the procedure.

For both adaptive analysis and adaptive synthesis, it makes sense to produce the individual substance mixtures or compounds in separate compartments. In this way it is possible to carry out the measurements used to determine the fitness values without having to advance a separate purification or separation step.

A sample robot is particularly advantageously used to produce the substance mixtures or compounds. This sample robot leads the

Mixing or synthesis steps for the individual components and transports the products and comparative samples. Such sample robots are known from the prior art and find z. B. in the pharmaceutical industry in combinatorial or parallel synthesis of a variety of compounds use.

Such a sample robot used in the context of adaptive chemistry should have devices for performing standard chemical operations. These typical chemical standard operations include steps such as the metering of starting materials, reagents or catalysts and the effective mixing of the substance or reaction mixtures. In the case of adaptive synthesis, further features are useful. The sample robot should also have a temperature control option to keep a reaction mixture at a certain temperature. Since, as a rule, further processing steps are required after a synthesis has been successfully carried out, automation is also desirable here. So z. B. advantageously simple processing steps such as extraction can be implemented in the sample robot.

In addition to the use of a sample robot described above, the system for carrying out the method expediently also includes the use of an analyzer system with which the measured variables can be determined, on the basis of which the fitness values are determined as a measure for the agreement of the measured variables with a target.

Such an analyzer system can e.g. B. include a spectrometer of various types, a chromatograph or an electro- or biochemical detector or sensor. A wide variety of measurement methods are conceivable, particularly in the field of spectroscopy. The method according to the invention can include the use of a UV, NMR, fluorescence or mass spectrometer. The exact choice of the device depends on the exact task and the composition of the substance mixtures or compounds. Of course, several analysis methods can also be coupled. An example of this is the LC-MS coupling. Advantageously, the sample robot used is also able to transport samples and comparison samples to the analyzer system. While conventional analyzer systems often use specially adapted samplers for routine analyzes, the integration of a freely movable sample robot in a modular overall system achieves maximum flexibility to enable the most varied of adaptive chemistry tasks to be processed.

Common to both sample robots and analyzer systems from the prior art is that the sequence of the steps to be carried out has to be determined by the user from the outset. In principle, it is also possible with such systems to stop the system after completing one or more steps and to redefine the sequence of the next tasks. Ultimately, however, the fully automatic implementation of the entire process is abandoned and the further determination of the steps is left to the knowledge and experience of the user. Systems from sample robots and analyzer systems which, based on implemented strategies, autonomously determine the sequence of the steps to be carried out by the user are not yet known.

In order to overcome this problem from the prior art, the use of a higher-level sequence control is proposed. This should control both the work of the sample robot and that of the analyzer system and determine the progress of the adjustment accordingly. The sequence control is intended to synchronize the various processes running and to control the entire process autonomously.

The sample robots and analyzer systems used are expediently used as modules within a feedback system. To enable the individual devices to be put together as modules, the devices should be macro-capable. Mechanisms of program interaction send macro commands directly from the higher-level process control to the device-side control software. The process control in turn receives data from the devices used via macro interfaces. Part of this higher-level process control is the software core, which works with genetic algorithms and thus determines the composition of the members of the population. The process control should receive information and data from the sample robot or analyzer system via interfaces, so that it is able to make autonomous decisions and to forward control commands to the other system components. The necessary analyzer and robot control programs are commercially available.

The sequence control should expediently be designed in such a way that, in the case of adaptive analysis, it evaluates the fitness values serving as a measure for the correspondence of the measured quantities of the substance mixtures produced with those of the substance mixture to be analyzed with the aid of genetic algorithms and then the compositions of the substance mixtures of the next generation sets. Similarly, in the case of adaptive synthesis, the sequential control system should use genetic algorithms to evaluate the fitness values for the substance mixtures or compounds produced, which form a measure of the agreement with the target, in order to determine the conditions for the production of the compounds of the next generation. In this way, the system according to the invention enables the autonomous implementation of adaptive analyzes or syntheses without the user having to intervene in the meantime. In addition, when determining the compositions of the substance mixtures or the conditions for producing the compounds or substance mixtures of the next generation, the sequence control can take into account further parameters that are characteristic of the components or starting materials used, such as costs, MAK values, ecological data, usability or availability.

The accompanying drawing is intended to illustrate the invention. It shows:

Figure 1 is a feedback loop showing the general approach to adaptive chemistry. The method of adaptive analysis or synthesis on which the invention is based is illustrated once again in FIG. After the production of a starting population, one or more parameters are determined for each member of this population. These measured variables are compared with a target value which, in the case of analysis, results from the corresponding measured variable for the mixture to be analyzed and, in the case of synthesis, must be specified on the basis of the synthesis target. By comparison with the target value, a measure of the agreement with this fitness value is determined. Finally, on the basis of these fitness values, genetic algorithms are used to determine the conditions for the production of the next generation population. This cycle can be repeated until after X generations at least one member of the population has a sufficiently good fitness level.

The spectrophotometric multi-component analysis is described as an example of the adaptive analysis. This is a common method for the analysis of mixtures, in which the additivity of the extinctions of the components is exploited without being separated into individual components.

As part of the adaptive analysis, the spectra of newly produced mixtures are now gradually adapted to the sample spectrum. For this purpose, a start generation of mixtures is randomly produced. A spectrum is measured for each mixture of substances and compared with the spectrum of the sample of unknown composition. For each sample, fitness values are derived from the comparison, which provide information about the agreement with the sample spectrum. In the starting population, these fitness values will generally not show good agreement with the sample to be analyzed.

For some of the mixtures examined, however, the measured fitness will be better than for others. Because of these differences, the mixtures of the next generation are then determined using genetic algorithms. This process is repeated over so many generations until the adaptation to the sample spectrum is achieved with a desired fitness. Various parameters can be adapted when using genetic algorithms. In the context of the crossing in particular, this is the so-called crossover rate, which makes a statement about the probability with which two arbitrarily removed individuals are "paired" or put back unchanged back into the population.

In a similar way, a further parameter can be set for the mutation. This mutation rate provides information about the probability of a change occurring at any point within the individual.

Claims

claims

1. Method for analyzing a mixture of substances, the components of which are part of a total of known, possible components in an unknown ratio, characterized by

a) production of a starting generation of random substance mixtures from at least two components of the total in different mixing ratios, the search space of the mixtures being able to be limited by known boundary conditions,

b) determination of one or more measured variables for each of the substance mixtures produced,

c) comparison of the measured variables determined for the substance mixtures produced with the measured variable (s) of the substance mixture to be analyzed,

d) Creation of a measure for each of the substance mixtures produced for the correspondence of the measured variable (s) with that of the analyte

Mixture of substances,

e) use of genetic algorithms taking into account the dimensions from step d) to determine the compositions of the substance mixtures of the next generation,

f) Production of a new generation of mixtures

Basis of step e), g) repetition of steps b) to f) until the measured variable (s) for at least one component of the generation of substance mixtures produced has a deviation from the measured variable (s) for the substance mixture to be analyzed which is smaller than a predetermined value.

2. The method according to claim 1, characterized in that the

Comparison of the measured quantities determined for the substance mixtures produced with those of the substance mixture to be analyzed and further parameters characteristic of the components used are taken into account when determining the composition of the substance mixtures of the next generation.

3. A process for the discovery and production of new compounds or mixtures of substances from components, associated with the establishment of a total of suitable starting materials and reaction conditions, characterized by

a) Production of a starting generation of different compounds or mixtures of substances by varying the starting materials, the ratios of the starting materials and / or the reaction conditions from the set as a whole, it being possible for the search space for the compounds or mixtures of substances to be limited by known boundary conditions.

b) determination of one or more parameters for each of the compounds or mixtures of substances produced,

c) comparison of the measured variables determined for the compounds or substance mixtures produced with a target,

d) preparation of a measure for each of the compounds or mixtures of substances produced for the conformity of the measurement variable / s with the target, e) use of genetic algorithms taking into account the degree of agreement from step d) to determine the conditions for the production of the compounds or mixtures of substances of the next generation,

f) production of a new generation of compounds or substance mixtures based on step e),

g) repetition of steps b) to f) until the measured variable (s) for at least one component of the generation of compounds or substance mixtures produced has a deviation from the target which is smaller than a predetermined value.

4. The method according to claim 3, characterized in that the

Comparison of the measured quantities determined for the compounds or mixtures of substances produced with the target and when defining the conditions for the preparation of the compounds or mixtures of substances of the next generation, further parameters which are characteristic of the educts used are taken into account.

5. The method according to claim 2 or 4, characterized in that costs, MAK values, ecological data, manageability and / or the availability of the components or starting materials are taken into account as further parameters characteristic of the components or starting materials used.

6. The method according to any one of claims 1 to 5, characterized in that the compounds or mixtures of substances are prepared in spatially separate compartments.

7. The method according to any one of claims 1 to 6, characterized in that at least one sample robot is used to produce the substance mixtures or compounds.

8. The method according to claim 7, characterized in that the sample robot has devices for carrying out standard chemical operations such as dosing, mixing, tempering or extracting.

9. The method according to any one of claims 1 to 8, characterized in that at least one analyzer system is used to determine the measured variables for the individual substance mixtures or compounds.

10. The method according to claim 9, characterized in that the analyzer system comprises a spectrometer, a chromatograph or an electro- or biochemical detector or sensor.

11. The method according to any one of claims 1 to 10, characterized in that a higher-level sequence control is used.

12. The method according to claim 11, characterized in that the higher-level sequence control controls the work of the sample robot and the work of the analyzer system in accordance with the progress of the adaptation and controls the time sequence.

13. System for carrying out a method according to claim 1 or 3, which comprises at least one sample robot for the production of mixtures or compounds and at least one analyzer system for determining measured variables for the individual mixtures or compounds, characterized in that the system has a feedback sequence control that autonomously controls the work of the sample robot and the work of the analyzer system and controls the chronological sequence.

14. Plant according to claim 13, characterized in that the sequence control uses genetic algorithms to based on the

Correspondence of the measured quantities for the substance mixtures or the compounds produced with those of the substance mixture to be analyzed or the target, the compositions of the substance mixtures or the Establish conditions for making next generation connections.

15. Plant according to claim 13, characterized in that the sequence control takes into account further parameters characteristic of the components or starting materials used when determining the compositions of the substance mixtures or the conditions for producing the compounds or substance mixtures of the next generation.

16. System according to claim 15, characterized in that the sequence control takes into account costs, MAK values, ecological data, manageability and / or the availability of the components or starting materials as further parameters which are characteristic of the components or starting materials used.

17. Plant according to one of claims 13 to 16, characterized in that the compounds or mixtures of substances are produced in spatially separate compartments.

18. Plant according to one of claims 13 to 17, characterized in that the sample robot has devices for carrying out standard chemical operations such as dosing, mixing, tempering or extracting.

19. Plant according to one of claims 13 to 18, characterized in that the analyzer system comprises a spectrometer, a chromatograph or an electro- or biochemical detector or sensor.

20. Plant according to one of claims 13 to 19, characterized in that the sequence control has interfaces via which

Data can be exchanged with the sample robot and the analyzer system.

21. Plant according to one of claims 13 to 20, characterized in that the sequence control has interfaces via which the sequence control gives control commands to the sample robot and the analyzer system.