Method for determining target synthesis specification

Abstract

The present invention relates to a method for determining a target molecule. Target characteristics and digital representations of potential target molecules are provided. The model is then utilized to determine characteristics of the potential target molecules. The model has been parameterized based on the explored data set and the unexplored data set. The explored data set includes characteristics of a plurality of explored molecules and molecular characterization parameter values of the plurality of explored molecules. The unexplored data set includes characterization parameter values for a plurality of unexplored molecules. The determined characteristic of the potential target molecule is compared to the target characteristic. Based on the comparison, i) a potential target molecule is determined as a target molecule, or ii) a new potential molecule is determined and the determination of properties is repeated. The determined molecule of interest is then provided.

Description

Technical Field

[0001] The present invention relates to a method, device and computer program product for determining the synthesis specifications and / or molecular structure of a target molecule including target technical application characteristics. Further, the present invention relates to a method, device and computer program product for generating a determination model based on machine learning, which can be used to determine the technical application characteristics of a molecule. In addition, the present invention relates to a method, device and computer program product for determining molecular characterization parameters for training a determination model based on machine learning, which can be used to determine the technical application characteristics of a molecule. In addition, the present invention relates to an interface method, an interface device and an interface computer program product for providing an interface for the above-mentioned method, device and computer program product. Background Art

[0002] In general, different molecules (in particular polymers) produced by the chemical industry are widely used in products for industrial and / or daily use with a wide range of application properties. In order to develop new molecules or to utilize known molecules in new application contexts, it is often useful to know in advance at least some of the technical application properties of the molecules, such as thermal insulation factor, hardness or reflectivity. For this purpose, in many cases, use is made of determination models which are, for example, based on machine learning principles and which are trained to determine the corresponding technical application properties of a given molecule. Summary of the invention

[0003] Developing and determining models so that they allow to realize high accuracy under specific application background is usually difficult, error-prone and time-consuming process. Especially, to allowing training to determine the model so that the provision of the training data set suitable for the predetermined application background is usually only based on expert knowledge and trial and error process, in this process, use potential training data set to train the model, and then based on the result of training, that is, based on the possibility of the trained determination model, repeatedly revise the training data set. Due to providing this training data set is always associated with multiple features and application characteristics of the molecule measured as a part of the training data set, so this process also can cause the high consumption of laboratory resources (for example, material resources, human resources, computing resources, etc.). In addition, the commonly known determination model training process is very consuming time and computing resources, so that avoiding non-productive training (that is, producing training of inappropriate models) will be very helpful. Therefore, if it is possible to train the determination model more efficiently (that is, with less resource consumption) and more objectively (that is, less based on expert knowledge) to provide training data for training the determination model, this will be advantageous. Further, typical training processes are based on the historical molecular data of specific applications. However, for new applications, the correspondingly trained models are often not very suitable, and a completely new model must be trained with the corresponding above-mentioned time- and computing-resource-intensive training process. It would therefore also be advantageous if the corresponding application possibilities of new applications could already be taken into account during the training process, thereby making the model more generally applicable.

[0004] The object of the present invention is to provide a method, a device and a computer program product for a) determining a target molecule comprising one or more target application characteristics, b) for training a determination model for determining the target molecule, and c) for providing a training data set for training the determination model, wherein the determination model used for training can be provided more efficiently and more objectively.

[0005] In a first aspect of the present invention, a computer-implemented method for determining the synthesis specifications and / or molecular structure of a target molecule, in particular a target polymer, including target technical application characteristics is proposed, wherein the method comprises: i) providing the target technical application characteristics, ii) providing a digital representation of the potential target molecule, the digital representation indicating the characterization parameters or being associated with these characterization parameters, iii) determining the technical application characteristics of the potential target molecule using a trained determination model, wherein the trained determination model has been parameterized based on an explored data set and an unexplored data set, so that the trained determination model is suitable for determining the technical application characteristics of the molecule based on one or more molecular characterization parameters of the molecule, wherein the explored data set comprises a) a plurality of explored molecules; The invention further comprises the steps of: providing a method for determining a target molecule of a plurality of molecules, wherein the method comprises: ... In particular, unexplored data sets can be easily generated with very few computing resources and used to define possible technical applications of a determined model, for example, can be used to determine characterization parameters of technical application characteristics and / or molecules to which a trained determined model can be applicable. Then, the unexplored data set generated in this way can be used to customize a training data set for the technical application background determined in this way based on the explored data set. This allows reducing the amount of trial and error runs used to train the determined model and also the amount of additional measurements used to determine further exploration data. In addition, the applicable field of the trained model can be increased based on the unexplored data set, and therefore this trained model can be more generally applicable. In addition, the influence of expert knowledge on the customization of the training data set can be greatly reduced, thereby producing more objective training.

[0006] The method refers to a computer-implemented method, and therefore can be performed by a general or special computer or a network of computers, which is suitable for performing the method, for example, by executing a corresponding computer program. The method is configured to determine a target molecule including a target technology application characteristic. Generally, the term "molecule" refers to a group of two or more atoms bonded together by chemical bonds, wherein all bonds between non-hydrogen atoms are defined. For example, these bonds can have random characteristics, like in random copolymers. In addition, each molecule as used herein is associated with a specific molecular structure and / or synthesis specification, so that the molecular structure and / or synthesis specification define the molecule. However, a molecule can be associated with different molecular structures. For example, this will be the case for a polymer with a molar mass distribution. Accordingly, providing a molecule generally also includes providing the corresponding molecular structure and / or synthesis specification of the molecule. In an embodiment, a molecule is a small molecule, wherein a small molecule is defined herein as a molecule present in an environment in a form that allows the molecule to be fully described using a simple structural formula containing relevant information. A simple molecular formula refers to a molecule that can be described by a covalent bond between the atoms of a molecule. However, a molecule can be composed of isomers and can be partially protonated. Examples where this is not the case are systems with dynamic equilibrium, for example, between several forms, such as monomers and oligomers, as in the case of several inorganic acids, or ionic species with very localized charges that strongly interact with solvents, such as by hydrogen bonds. Preferably, small molecules are defined as molecular weights less than 600 g / mol, more preferably less than 300 g / mol. Preferably, the target molecule is a target polymer, and the method is configured to determine the target synthesis specification of the target polymer. Typically, synthesis specifications are defined as instructions about how to synthesize molecules (particularly polymers). In particular, for polymers, synthesis specifications indicate starting materials and the corresponding parameters for polymerization by the starting materials.

[0007] Technical application characteristics can generally refer to any characteristics of a molecule and / or a substance at least partially composed of the molecule (e.g., a formulation or mixture containing the molecule), which allows the assessment of the technical suitability of the corresponding molecule provided after synthesis. Preferably, the technical application characteristics are characteristics associated with a substance containing or composed of a molecule. In particular, the technical application characteristics can be defined by the expected technical application of the molecule and / or the substance at least partially composed of the molecule. In particular, the difference between technical application characteristics and non-technical application characteristics can be that the technical application characteristics directly allow the determination of whether the molecule is suitable for a technical application, while non-technical application characteristics (such as the bond strength between two atoms of a molecule) do not directly allow the determination of technical application characteristics. Preferably, the technical application characteristics include at least one of mechanical characteristics, optical characteristics, physicochemical characteristics, chemical characteristics and biological characteristics. Generally, mechanical characteristics can refer to any one of the following: adhesion, tensile strength, stiffness, hardness, shrinkage, elongation, crack tearing, tear strength, rebound, compressibility, wear, overflow, morphology, tactile characteristics, breaking stress, elongation at break, particle size and filling degree. Optical properties may generally include any of the following: coloration, turbidity, opacity, transparency, reflection, appearance, absorption, scattering, color strength, hue, color saturation, color intensity, cloud point, extinction, optical density, spectrum, refractive index. In addition, physicochemical properties may refer to any of the following: density, viscosity, K value, molar mass, dispersibility, molar mass distribution, particle size distribution, solubility, partition coefficient, interface properties, surface tension, dispersibility, storage stability, odor, segregation, condensation, conductivity, capacitance, surface area, flow time, vapor pressure, VOC, solid content, hygroscopicity, magnetism, miscibility, thixotropy, phase change properties, glass transition temperature, corrosion inhibition, solvent separation, aggregation, self-heating ability, impact sensitivity, drying loss, response angle, static charge, minimum film forming temperature and charge density. Chemical properties may include any of the following: chemical resistance, reaction time, demolding time, growth, hard / soft segment content, crystallinity, reaction temperature, reaction pressure, decomposition, thermal decomposition, photodegradation, acidity, pKa, pH, moisture / water content, flammability, burning rate, autoignition, flash point, formation of flammable gases, reaction to fire, deflagration rate, residual monomer count, by-product formation, degree of polymerization, salt content, temperature resistance, oxidizing properties, reducing properties, reactivity, ash content, non-volatile matter content, stability, chelating ability, calorific value, saponification value.Furthermore, the biological properties may include any of the following: biodegradability, bioresistance, in particular resistance to pathogenic viruses, bacteria, fungi, plants or animals or developmental stages of said pathogens, tolerance to environmental parameters, such as drought tolerance, tolerance to enzymatic degradation, such as protease resistance, lipase resistance, amylase resistance, hydrolase resistance, pesticide resistance, toxicity, biotransformation, ecotoxicology, sensitization, in particular allergenicity, bacterial count, enzyme activity, substrate specificity, cofactor dependence, product specificity, substrate and / or product inhibition, dissociation constant, Michaelis-Menten kinetic value, activity / stability under or in different terms of pH, temperature, pressure, organic solvent concentration, carrier formulation, encapsulation formulation; distribution in the environment, compartmentalization, bioaccumulation, biological exposure LD50, mutagenicity.

[0008] In the first step, the method includes providing a target technical application characteristic. Generally, in the present invention, if not otherwise specified, providing an amount also refers to providing a quantitative amount of the amount, such as a value or a value range. Therefore, providing a target technical application characteristic also includes providing a quantitative amount of the target technical application characteristic, such as a value or a value range of the target technical application characteristic. In particular, the provision may refer to applying, for example, a corresponding input unit to receive the target application characteristic from the input of a user. In addition, the provision may also refer to accessing a storage unit in which the target application characteristic has been stored. Further, the provision may also include, for example, receiving the target application characteristic from other sources via a network connection, and providing the received target application characteristic. Generally, the target application characteristic may refer to a target value, such as a specific hardness of a molecule, or may refer to a value range that the molecule should conform to. In addition, the target application characteristic may also refer to any type of objective function, for example, a time series of characteristics under varying environmental conditions, such as hardness under varying temperature conditions. This more complex target application characteristic may be advantageous in the case where the application of the molecule includes different environmental conditions (e.g., different temperatures). Then, the target molecule refers to a molecule that provides a corresponding target technical application characteristic when provided in a corresponding form (e.g., as a pure substance or a mixture). In particular, the target molecule provides corresponding target technical application properties when synthesized according to corresponding synthesis specifications and / or molecular structures.

[0009] Further, the method includes providing a digital representation of a potential target molecule, which indicates or is associated with a characterization parameter of the potential target molecule. Generally, a digital representation is defined as data provided in a form and structure that allows processing in a digital system, particularly a computing system, for example, a digital representation can be a text string, as well as an ID, an image, a chemical formula, etc. Preferably, the digital representation includes a characterization parameter of a potential target molecule. Generally, the characterization parameter of a molecule quantifies one or more characteristics inherent to the molecule itself, particularly the physicochemical characteristics of the molecule itself. Therefore, the characterization parameter refers to the molecule itself, and the technical application characteristic refers to the characteristic of the molecule (for example, as a substance containing or composed of molecules) in its technical application form. Preferably, the characterization parameter is associated with synthesis specifications and / or molecular structure. Here, "associated with..." means that the characterization parameter can directly refer to synthesis specifications and / or molecular structure, for example, as a structural or process parameter, or can be obtained from synthesis specifications and / or molecular structure, for example, as a calculated physicochemical characteristic that can be calculated from synthesis specifications and / or molecular structure. Preferably, for a molecule that can be defined by a simple structural formula, the characterization parameter is associated with the molecular structure because the molecular structure clearly defines the molecule. However, in this case, the synthesis specification also allows to obtain suitable characterization parameters. In the case of more complex molecules, which can usually exist in a variety of different structures or for which it is usually difficult to determine the exact molecular structure, especially for polymers, it is preferred that these characterization parameters are associated with the synthesis specification. However, in these cases, at least some parts of the molecular structure can also be defined and then used to determine the characterization parameters. Generally, the characterization parameters can include any of the following: process parameters, recipe parameters, synthesis routes or physicochemical parameters of the molecule. In this context, the process parameters quantify the production process of the corresponding molecule and can be obtained from the synthesis specification of the molecule. Since specific process parameters lead to the production of specific molecules, the process parameters also characterize the molecule itself. For example, process parameters can refer to any of the following: temperature profile, pressure, stirring power, synthesis steps, etc. Recipe parameters usually quantify the production of the molecule itself, especially the substances used to produce the molecule. Such substances can refer to starting substances, such as corresponding prepolymers in the case where the molecule is a polymer, from which the molecule is produced. However, these substances can also refer to auxiliary substances, such as catalysts. In this context, the recipe parameters quantify the effects of these substances on the produced molecules and therefore also characterize the molecule itself. For example, a formulation parameter may refer to any of the following: the amount of a specific chemical substance, the amount of a specific additive, the mixing ratio between different substances, or the solvent used. Preferably, the digital representation comprises molecular physicochemical parameters, preferably, refers to molecular descriptors, i.e. calculated physicochemical molecular properties, wherein the molecular physicochemical parameters indicate the physicochemical characteristics of the molecule. In particular, the molecular physicochemical parameters indicate parameters that quantify the physicochemical characteristics of the molecule.In this context, the term "physicochemical characteristics" refers to the physical and / or chemical characteristics of a molecule. For example, a physicochemical parameter may refer to any of the following: molar mass distribution, viscosity, degree of protonation, etc. However, a digital representation may also be provided so that it allows the molecular physicochemical parameters to be obtained, for example by providing a representation of the molecule, wherein the corresponding molecular physicochemical parameters have been stored or can be determined, for example, by calculation of the corresponding molecular descriptors. Preferably, the digital representation includes at least one of the formula, structural formula, brand name, IUPAC name, chemical identifier, and CAS number of the molecule.

[0010] In a preferred embodiment, the target molecule is a polymer, and the characterization parameter indicates a parameter that quantifies the physicochemical characteristics of a subgroup of the polymer. In this embodiment, a digital representation can also be provided so that it allows the characterization parameter to be obtained by determining the subgroup of the polymer, and the characterization parameter is determined based on the physicochemical characteristics of the determined subgroup. Generally, a subgroup refers to a part of a polymer, wherein all subgroups of the polymer form a polymer together. For example, a subgroup can refer to a part of a polymer, wherein the subgroups are continuously linked together along a chain or network to form a polymer. Preferably, a subgroup of a polymer refers to a repeating unit describing a part of a polymer, which produces a complete polymer chain when repeated. However, in some cases, a subgroup can also refer to a single part that is not repeated in a polymer. In addition, it is preferred that a subgroup contains a repeated part, for example, a subgroup of a polymer can contain a repeating core that is also present in other subgroups and other additional parts that are not present in other subgroups. Preferably, a subgroup refers to at least one of a polymerized monomer or an oligomer fragment. More preferably, a subgroup refers to a polymerized monomer. In this context, a polymerized monomer refers to a monomer after polymerization, sometimes also referred to as a "monomer unit (mer unit or mer)". In particular, polymerized monomers do not refer to monomers present in the reaction mixture before polymerization, i.e., raw materials, but to repeating units derived from monomers that have changed during or after polymerization. Therefore, the subgroup descriptors determined for polymerized monomers are different from the subgroup descriptors determined for unreacted monomers before polymerization. The inventors have found that, in particular, polymerized monomers allow polymer descriptors to be determined from the subgroup descriptors of polymerized monomers, thereby allowing a model to be determined to accurately determine the technical application characteristics of the polymer. In a preferred embodiment, the digital representation of the polymer includes a subgroup provided as a molecular model, which indicates the chemical structure of the subgroup after polymerization. Even more preferably, the molecular model of the subgroup is determined in such a way that it is suitable for quantum chemical calculations about many atoms and their connectivity, and its connectivity represents the characteristics of the subgroup in the polymer. In addition, in addition to or as a substitute for the molecular model of the subgroup that considers the subgroup as a monomer structure, a molecular model involving an oligomer model can also be used, which takes into account the influence of the adjacent molecular structure of the subgroup in the polymer.

[0011] In an embodiment, if the digital representation of a molecule does not directly include a characterization parameter, it is preferred that the characterization parameter is determined by a specified molecular structure and / or synthesis specification of the provided molecule. If the molecular structure and / or synthesis specification are not provided as a part of the digital representation, the molecular structure and / or synthesis specification can be obtained from the information of the molecule itself, for example, by utilizing a corresponding molecular database or a known method for determining the synthesis specification, like a data-driven reverse synthesis path planner. In addition, in this case, it is also possible to request that the molecular structure and / or synthesis specification be provided by the user. If the molecule is a polymer, it is preferred that the characterization parameter is determined based on the synthesis specification provided rather than the molecular structure that cannot be determined for the polymer usually. In addition, in this case, if the characterization parameter is a characterization parameter based on a subgroup, it is preferred that the characterization parameter is determined by first determining the subgroup of the polymer. For example, known methods can be used to determine each subgroup of a polymer. In particular, it is preferred that a subgroup is determined so that the bonds between atoms of different subgroups in the polymer are polarized as little as possible, and preferably, the bond order is as small as possible (for example, CC single bonds). In addition, it is preferred that the subgroup representing the polymer comprises the same number of active non-hydrogen atoms as the polymer. In addition to active atoms, subgroups can also contain other atoms, which can be ignored when calculating the physicochemical parameters of subgroups. In addition, preferably, subgroups are determined in such a way that the polymer comprising the part constructed with different polymerization techniques is fully covered and meets the above conditions. Example is a polyether used as a polyurethane component. Usually, a database or archive comprising multiple reactions between polymer parts can be generated, and subgroups can be obtained from the corresponding structure of the reaction. For example, specific chemical languages ​​like SMILES and SMARTS can be used to easily obtain the subgroup of polymer. For example, a database of reaction SMARTS can be generated, and then corresponding reaction SMARTS can be selected based on the polymerization of corresponding polymers. From the selected reaction SMARTS, the SMILES of polymer monomers can be directly obtained, and for example, RDkit can be used to determine the SMILES of subgroups from the SMILES of monomers, i.e. the number and connectivity of atoms.

[0012] The determined subgroups of polymers are associated with subgroup characterization parameters, which indicate parameters quantifying physicochemical characteristics of subgroups, in particular in polymers. In particular, it is preferred that, if the characterization parameters are not directly provided by the digital representation, the characterization parameters are determined by determining a corresponding subgroup characterization parameter for each of the subgroups, and determining the characterization parameters based on the subgroup characterization parameters of the subgroups, for example by averaging. Therefore, the method in this embodiment preferably comprises: first providing or determining subgroups for polymers from the digital representation, then determining or providing subgroup characterization parameters of the subgroups, i.e. parameter values ​​of the characterization parameters quantifying the subgroups, and then determining the polymer characterization parameters based on the subgroup characterization parameters of each polymer.

[0013] Preferably, the molecular characterization parameters include physicochemical parameters, which refer to at least one of a composition descriptor, a count descriptor, a list of structural fragments, a fingerprint, a graph invariant, a 3D descriptor and / or a higher dimensional descriptor, which indicates a parameter that quantifies the physicochemical characteristics of a molecule. In addition, in the case of a polymer, it is preferred that the molecular characterization parameters include the molar mass distribution of the polymer as a physicochemical parameter. In the case of a polymer, it is further preferred that the molecular characterization parameters include a temperature profile as a process parameter. In addition, it is preferred that the molecular characterization parameters include the amount of the corresponding component as a recipe parameter. Typically, the molecular characterization parameters may also preferably include any combination of the possible parameters described above or below.

[0014] In a preferred embodiment, the molecular characterization parameters include 3D descriptors, in particular calculated physicochemical properties, and / or fingerprints. If the molecule is a polymer, the molecular characterization parameters can be obtained from the subgroup parameters, and therefore, the subgroup characterization parameters can also refer to the same characterization parameters as described above. Typically, the characterization parameters can also be obtained without using subgroups, for example by simulating the entire molecule. Possible characterization parameters will be defined in more detail below. In addition, in these cases, the defined characterization parameters can directly refer to the molecular characterization parameters, or alternatively for polymer embodiments, refer to the subgroup characterization parameters.

[0015] Composition descriptors may refer to any of the following: potential energy, average molecular weight, polydispersity, charge, spin, boiling point, melting point, enthalpy of fusion, dissociation constant, Hansen parameters, proton, polar and dispersion contributions, Abraham parameters, retention index, TPSA, torsion angle, degree of branching, nucleotide and / or amino acid composition, nucleotide and / or amino acid sequence, nucleotide and / or amino acid sequence conservation, isoelectric point, glycosylation pattern, receptor binding constant, inhibitor constant. Count descriptors may refer to any of the following: the sum of atomic electronegativity, the sum of atomic polarizabilities, the amount of components, the ratio of the amounts of components, the number of atoms and non-H atoms, the number of H, B, C, N, O, P, S, Hal and heavy atoms, the number of H donor and H acceptor atoms, the number of bonds, non-H or multiple bonds, the number of double bonds, triple bonds and aromatic bonds, the number of functional groups, the ratio of functional groups, the sum of bond orders, the aromatic ratio, the number of rings or loops, the number of unpaired electrons, the number of rotatable bonds, the rotatable bond fraction and the number of conformers. Molecular descriptors involving a list of structural fragment descriptors may refer to at least one of a list of molecular fractions, a list of functional groups, a list of bonds and a list of atoms. Fingerprint descriptors preferably include at least one of the following: MACCS keys (preferably in bit format or total format), Morgan and other circular fingerprints (preferably in bit format or total format), topological torsion, atom pairs, infrared and correlation spectra, fingerprint counts, PubChem fingerprints, substructure fingerprints and Klekota-Roth fingerprints. The graph invariant / topological index descriptor preferably comprises at least one of a topological structure index and a topological chemical index. In a preferred embodiment, the characterization parameters comprise 3D descriptors, including at least one of the following: total atomic volume sum, average volume per atom, total atomic area sum, average area per atom, area of ​​all atoms, average area per atom, solvent accessible surface, dispersion energy, dielectric energy, H-donors, H-acceptors, polar and non-polar surface areas, atomically resolved H-donors, H-acceptors, polar and non-polar surface areas, shape, sphericity, dipole moment and higher moment, polarizability, dielectric energy, protons, polar and non-polar surface areas, orbital energy and orbital gap, ionization energy, electron affinity, hardness, electronegativity, electrophilicity, excitation energy and intensity, infrared and ultraviolet absorption bands, reactivity measurements, redox potentials, bond standard points, partial charges, charge surface area, atomic orbital contributions, bond order, atomic radius.In particular, it is preferred that the molecular characterization parameters refer to 3D descriptors, which include at least one of the following: the sum of the volumes of all atoms, the average volume per atom, the sum of the areas of all atoms, the average area per atom, solvent accessible surface, dispersion energy, dielectric energy, H donors, H acceptors, polar and / or non-polar surface area, atomically resolved H donors, H acceptors, polar and / or non-polar surface area, shape, sphericity, cone angle, polarizability, dielectric energy, protons, polar and / or non-polar surface area, excitation energy and intensity, infrared and / or UV absorption bands, reactivity measurements, particle charge and / or charge surface area. Preferably, the higher dimensional descriptors utilized may include at least one of conformational partition function, solubility, vapor pressure, activity coefficient, diffusion coefficient, partition coefficient, interfacial activity, rotation constant, moment of inertia, radius of gyration, compositional drift of the molecule, density, viscosity, conformer weighted volume and area, conformer weighted H-donors, H-acceptors, protons, polar and / or non-polar surface area, charge distribution, conformational dipole moment, and molecular refraction. Preferably, higher dimensional descriptors are utilized that include at least one of solubility, vapor pressure and activity coefficient, interfacial activity, conformer weighted H-donors, H-acceptors, protons, polar and non-polar surface area, and charge distribution.

[0016] In the next step, the technical application characteristics of the molecule are determined using a trained determination model. Preferably, the determination model can include, for example, providing or receiving a determination model from a corresponding storage unit and utilizing the determination model provided or received. The trained determination model has been parameterized based on an explored data set and an unexplored data set so that the trained determination model is suitable for determining the technical application characteristics of the molecule based on one or more molecular characterization parameters of the molecule. Usually, parameterizing the trained determination model refers to determining the parameters of the determination model during the training process, so the determination model is trained based on an explored data set and an unexplored data set. The term "so that" should be interpreted as parameterization (i.e., training) here to make the determination model suitable for and therefore able to determine the technical application characteristics of the molecule based on one or more of the molecular characterization parameters of the molecule. In particular, the determination model is a data-driven model, and the term "data-driven" is used to emphasize that the model is mainly based on corresponding data input, rather than, for example, based on intuition, personal experience or knowledge. Preferably, the determination model is based on a known machine learning algorithm, such as a neural network, a regression model, a classification algorithm, etc. It has been found that for most applications in this context, in particular regression models based on linear regression, random forest, Lasso, boosted tree, ridge regression and MARS algorithms are suitable, while for classification models in particular random forest, logistic regression and SVM algorithms are suitable. Typically, the deterministic model is parameterized during a training process in which the deterministic model is trained using an explored data set and an unexplored data set.

[0017] Typically, the explored data set includes a) a plurality of technical application characteristics of the explored molecules and b) a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with the synthesis specifications and / or molecular structures corresponding to each of the explored molecules in the plurality of explored molecules. Therefore, the explored data set refers to a data set that includes only molecules whose at least one technical application characteristic has been determined at least once. Typically, an explored molecule refers to a molecule whose at least one technical application characteristic has been determined. The term "associated with..." in the statement "molecular characterization parameters are associated with synthesis specifications and / or molecular structures" will be interpreted herein as "defined" or "obtainable" by the corresponding synthesis specifications and / or molecular structures as described above with respect to the characterization parameters of potential target molecules. The technical application characteristics of the explored molecules can be determined in any known manner, such as based on experimental data and measurements, based on simulations or theoretical considerations. Preferably, the technical application characteristics refer to the measured technical application characteristics. For example, in order to measure the technical application characteristics, the corresponding molecules can be synthesized and the corresponding technical application characteristics can be measured in a suitable measurement process. However, the technical application characteristics can also be determined based on corresponding physical or chemical calculations or based on corresponding simulations of molecules. The molecular characterization parameter value associated with the molecule can also be measured, and can be determined in any other known manner (e.g., as described above, can be calculated). In an embodiment, the explored data set can further include producibility information for one or more of the multiple molecules in the explored data set, wherein the producibility information indicates which technical application characteristics of the molecule can be measured. For example, during the synthesis of the molecule, it can be determined that the molecule cannot usually be produced with a known synthesis process, or that the synthesized molecule is not suitable for certain measurement processes that produce corresponding technical application characteristics. This information can be provided as part of the producibility information. Typically, producibility information can refer to category information indicating whether a molecule is related to one or more categories. For example, a category can refer to whether a molecule is producible, or whether a certain application characteristic of a molecule can be measured. Providing at least some explored molecules with this information (which acts as a supplement or replacement for technical application characteristics as part of the explored data set) allows further training of the determination model based on these explored molecules to at least notice when a molecule may not be suitable for application (e.g., due to synthesis problems). Generally, for this embodiment, it is preferred that more explored molecules are provided with technical application characteristics than explored molecules are provided with manufacturability information rather than technical application characteristics.

[0018] The unexplored data set includes multiple molecular characterization parameter values ​​of multiple molecular characterization parameters, and the multiple molecular characterization parameters are associated with the synthesis specifications and / or molecular structures corresponding to each unexplored molecule in a variety of unexplored molecules. Preferably, the unexplored data set is composed of multiple molecular characterization parameter values ​​of multiple molecular characterization parameters, and the multiple molecular characterization parameters are associated with the synthesis specifications and / or molecular structures corresponding to each unexplored molecule in a variety of unexplored molecules. Therefore, the unexplored data set does not include the technical application characteristics associated with the unexplored molecules. This allows a large number of unexplored molecules, particularly molecular structures and / or synthesis specifications, and corresponding molecular characterization parameter values ​​for generating unexplored data sets to be easily and quickly generated. For example, in the case of a polymer based on an initial synthesis specification, the parameters of the synthesis specification can be changed to generate a new synthesis. In the case of a small molecule, a similar molecule database query based on a fingerprint can be performed. An alternative would be to select different subsets of molecules from a larger possible set of molecules across a certain chemical range. In addition, any variation from the initial synthesis specification and / or molecular structure can also be used to generate a large number of unexplored synthesis specifications and / or molecular structures. Then, based on these synthesis specifications and / or molecular structures, the corresponding molecular characterization parameter values ​​of the unexplored molecules associated with the corresponding synthesis specifications and / or molecular structures can be determined using known methods (e.g., as described above). In this way, a huge unexplored data set including, for example, thousands of unexplored molecules can be generated. In particular, the explored data set and the unexplored data set include at least some, preferably all molecular characterization parameter values ​​of the same molecular characterization parameter for each molecule. For example, an unexplored data set can be generated so that the molecular characterization parameters provided with their corresponding values ​​in the explored data set are calculated. However, in the explored data set, the molecular characterization parameters referring to the molecular characterization parameters provided in the unexplored data set can also be calculated. In particular, the unexplored data set generated in this way can be used to define the application context of the determined model to be trained. In particular, the corresponding constraints of the expected application of the model can be considered to generate the unexplored data set. Then, based on the unexplored data set defining the application context, the explored data set can be utilized to generate a training data set for determining the model. For example, from the explored data set, explored molecules and / or molecular characterization parameters can then be selected to form a training data set. The inventors have found that in this way, a training data set can be provided for training a determination model specifically tailored to the application context of the technical application properties without a cumbersome trial and error process and without consuming high computing resources. Such a trained determination model then allows a particularly accurate determination of the technical application properties of molecules that fall within the application context of the determination model.

[0019] Then, in a further step, the technical application characteristics of determined potential target molecule are compared with the target technical application characteristics. Based on this comparison, then make a decision between the following two items: a) the potential target molecule is determined as the target molecule, or b) the potential target molecule is provided and the technical application characteristics are repeatedly determined using a new potential target molecule. Preferably, if the determined technical application characteristics of the potential target molecule are determined to meet the target technical application characteristics, then step a) is performed, and if the determined technical application characteristics of the potential target molecule are determined to not meet the target technical application characteristics, then step b) is performed. In particular, the comparison of determined technical application characteristics and the target technical application characteristics allows to determine whether the determined technical application characteristics meet a predetermined standard, and this predetermined standard is for example that the determined technical application characteristics meet the target technical application characteristics within a predetermined limit. If this standard is met, then the potential target molecule is determined as the target molecule, and the method proceeds to the next step. However, if this comparison shows that the determined technical application characteristics do not meet the target technical application characteristics within a predetermined limit, then the next iteration step utilizing a new potential target molecule must be processed. In particular, for each iteration step of iteration, preferably based on previous potential target molecules, for example, by modifying one or more features of previous synthesis specifications and / or molecular structure of previous molecules, new potential target molecules are determined. However, new potential target molecules can also be generated by arbitrarily selecting new potential target synthesis specifications and / or molecular structure from a large number of previously generated potential target synthesis specifications and / or molecular structure. In addition, more complex methods can also be utilized to select new potential target molecules from a plurality of previously generated potential target molecules. Based on new potential target molecules, in each iteration step, the determined model trained is utilized again to determine the technical application characteristics, and the technical application characteristics thus determined are compared with the target technical application characteristics again, so that the comparison can lead to further iteration steps again, or if the corresponding standards are met, the corresponding new potential target molecules can be selected as the target molecules. In addition, the additional termination criteria for iteration can also be selected, for example, the number of iteration steps before the iteration termination can be determined, and the user is notified that the target molecules for the corresponding technical application characteristics can not be found. However, alternatively, after the iteration step of a predetermined amount, the method can further include, for example, by increasing the predetermined limit around the technical application characteristics and so as to modify the target technical application characteristics by repeating the iteration while utilizing the increased limit during the comparison. This can allow to find target molecules that meet the technical application characteristics as much as possible, even if it may not be possible to meet the original goal.After determining the target molecule as described above, the target molecule and optionally other information, such as molecular structure or synthesis specifications for producing the molecule, can be provided to the user, for example, via an output unit.In the case of small molecules, a data-driven retrosynthesis planner can be used, which provides potential starting materials for the synthesis of the target molecule. Then, a database can be used, which contains many disclosed synthetic routes. This provides possible instructions for synthesizing the target molecule from the starting materials of the suggestion. Alternatively, a data-driven tool can be used to provide synthetic instructions and an estimated probability of successful synthesis.

[0020] In an embodiment, providing determined target molecule comprises generating control signal based on target molecule and providing control signal, wherein, these control signals are configured to control production system to produce target molecule according to target synthesis specification.Preferably, control signal is provided to corresponding production system, and this production system is configured to produce target molecule based on control signal.Usually, control signal can be provided with any format that allows directly or indirectly controlling production system to produce target molecule according to target synthesis specification.For example, control signal can provide with the format that allows production system management application program to interpret control signal and then control production system (for example, synthesis robot) to produce target molecule.However, control signal can also provide with the format that directly allows corresponding production system to be controlled to produce target molecule (for example, directly allows control component, for example, starts or stops heating unit, opens or closes valve, starts or stops mixer etc.).Usually, production system can refer to any fully or partially automated production system provided in laboratory or industrial environment, and this production system is usually configured to produce molecule based on synthesis specification.

[0021] In an embodiment, the parameterization of the determination model based on machine learning is based on determining a similarity measure between members of the explored data set and members of the unexplored data set. Generally, in this context, a similarity measure can be defined as a measure of the distance between one or more members of the explored data set and one or more members of the unexplored data set in a predetermined space. Preferably, the predetermined space is defined by one or more molecular characterization parameters. Therefore, in this case, the similarity measure indicates how close one or more members of the explored data set are to one or more members of the unexplored data set with respect to one or more molecular characterization parameters. In this context, a member of the explored data set refers to a molecule provided by the corresponding explored data set and all quantities associated with the molecule, and a member of the unexplored data set refers to a molecule provided by the unexplored data set and all quantities associated with the molecule. In particular, it is preferred that the similarity measure between members of the explored data set and members of the unexplored data set is determined with respect to molecular characterization parameters of members of the explored data set and members of the unexplored data set. In particular, the similarity measure indicates the distance between members of the explored data set and members of the unexplored data set with respect to a subset of these molecular characterization parameters. Utilizing similarity measures between members of the explored dataset and members of the unexplored dataset allows determining which members of the explored dataset are most similar to members of the unexplored dataset for different molecular characterization parameters. This allows selecting not only members of the explored dataset, but also combinations of molecular characterization parameters that are most suitable as part of the training dataset for training a determined model for a technical application context defined by the unexplored dataset.

[0022] Preferably, parameterizing the machine learning-based determination model includes selecting a subset of the molecular characterization parameters of the explored data set as training molecular characterization parameters based on a similarity measure, wherein the parameterization of the machine learning-based determination model utilizes a training data set, the training data set comprising: a) technical application characteristics of at least two explored molecules of the plurality of explored molecules and b) molecular characterization parameter values ​​of the training molecular characterization parameters, the training molecular characterization parameters being associated with the synthesis specifications and / or molecular structures corresponding to each of the at least two explored molecules. Typically, the selected subset of the molecular characterization parameters of the explored data set may refer to a combination of one or more molecular characterization parameters for which the explored data set provides values. In particular, preferably, the subset of molecular characterization parameters is selected by comparing similarity measures for different combinations of molecular characterization parameters for members of the explored data set and members of the unexplored data set, and determining from these combinations combinations in which the members of the explored data set are most similar to the members of the unexplored data set, wherein the corresponding combination of molecular characterization parameters may then be selected as a subset. Thus, based on the similarity measure, it can be determined which combination of molecular characterization parameters allows the explored data set to best cover the technical application context of the determined model defined by the unexplored data set.

[0023] Preferably, the subset of these molecular characterization parameters is selected based on the optimization of the similarity measure between the members of the explored data set and the unexplored data set with respect to the subset of these molecular characterization parameters. In particular, it is preferred that the optimization refers to determining the subset of molecular characterization parameters that makes the members of the unexplored data set and the members of the explored data set most similar. In other words, the similarity measure is optimized so that in the space defined by the subset of molecular characterization parameters, the members of the explored data set cover the part of the space defined by the unexplored data set as well as possible. Therefore, training the deterministic model with such a determined training data set allows the generation of a deterministic model that is particularly well suited to the intended technical application context without the need for trial and error runs or performing further experiments to generate additional explored molecules.

[0024] Preferably, for optimization, the similarity measure is mathematically defined as

[0025]

[0026] in, and are the molecular characterization parameters from the unexplored (scr) and explored (tr) datasets, where f refers to a specific characterization parameter, and i, j refer to the corresponding molecules from the unexplored and explored datasets, respectively, and the coefficient c fcan be 0 or 1 - depending on whether the corresponding characterization parameter is selected, and the sum of f is equal to the number of characterization parameters K in the selected subset of characterization parameters, D is the universal distance function between any given two data points in the space defined by the selected subset of characterization parameters, and S is the number of data points in the unexplored data set. The number of characterization parameters K in the selected subset of characterization parameters can refer to a predetermined number. For example, it can be known from experience that when determining the technical application characteristics with such a trained determination model, less than three characterization parameters may lead to less accurate results, and more than 20 characterization parameters must be provided to the determination model to determine the technical application characteristics. It may be disadvantageous. For example, a large number of characterization parameters that must be provided to the determination model to determine the technical application characteristics may lead to model overfitting, which should be avoided, because this may cause the resulting model to have a reduced predictiveness on a data set different from the data set used to train the model. Therefore, for example, the number of characterization parameters in the subset of characterization parameters can be predetermined based on such considerations as outlined above. However, it is preferred that the number of characterization parameters in the selected subset of characterization parameters itself is subject to an optimization process and can therefore be regarded as a hyperparameter. In particular, the optimization of the similarity measure may also include optimizing over the number of characterization parameters in the selected subset of characterization parameters, such that this number does not have to be predetermined or fixed during the optimization process.

[0027] In a preferred embodiment, the subset is selected by further optimizing the determination accuracy of the determination model with respect to the subset of molecular characterization parameters. The selection of the subset of characterization parameters further based on the optimization of the determination accuracy of the determination model with respect to the subset of molecular characterization parameters allows to already take into account the potential accuracy of the determination model during the determination of the training data set. This allows to optimize the selection of the characterization parameters in the training data set also with respect to the determination accuracy already during the determination of the training data set. Preferably, for the optimization, the determination accuracy is mathematically defined as

[0028]

[0029] in, refers to the technical application characteristics determined by a determined model, which is trained with a corresponding training data set based on the corresponding selected subset, Y j tr refers to the technical application characteristics of the corresponding molecule, T is the number of data points in the explored dataset, and D * is the universal distance measured between the technical application properties of the molecules of the explored dataset and the technical application properties of the molecules of the explored dataset determined using the trained determination model, and the parameter λ refers to the weighting of the determination accuracy with respect to other aspects of the optimization.

[0030] In a preferred embodiment, the subset is selected by further optimizing the fitness of the determined model in the unexplored data set. In particular, the fitness of the determined model in the unexplored data set can be measured by utilizing a criterion that determines the degree of coverage of the space defined by the unexplored data set by the selected subset of the explored data set with respect to the molecular characterization parameters. For example, if the explored data set only covers a very small part of the space defined by the unexplored data set, that is, if the coverage is incomplete, the determined model may not be applicable to the complete intended technical application context defined by the unexplored data set. Therefore, by also considering the fitness of the determined model in the unexplored data set, for example by utilizing a coverage degree measure, the training data set can also be directly optimized with respect to the fitness, and the expected fitness of the determined model can be ensured as much as possible. Preferably, for optimization, the fitness is mathematically defined as

[0031]

[0032] in, refers to the technical application characteristics determined by the trained deterministic model, Y j tr refers to the technical application characteristics of the corresponding molecule, G is a function evaluating the fitness of the determination model in the unexplored data set (e.g., Euclidean distance), S is the number of data points in the unexplored data set, and the parameter μ refers to the weighting of the fitness metric relative to the other terms of the optimization. Typically, the weighting parameters μ and λ that weight the additional metrics relative to the similarity metric during optimization can refer to predetermined values ​​determined, for example, based on user experience, theoretical considerations, or resampling methods. For example, an increase in μ increases the application space of the determination model around the explored data set, wherein a decrease also reduces the application space of the determination model. λ largely determines whether the determination model follows the explored data set. However, these parameters can also be regarded as hyperparameters that can also be subject to optimization (i.e., can be changed during the optimization process).

[0033] In an embodiment, selecting a subset of molecular characterization parameters includes: a) determining a parameter training space based on an unexplored data set, wherein the parameter training space is defined by the molecular characterization parameter values ​​of the molecular characterization parameters of a plurality of unexplored molecules, b) determining a subspace of the parameter training space so that the molecular characterization parameter values ​​of the molecular characterization parameters of the explored data set cover the subspace according to a predetermined criterion based on a similarity metric, and c) selecting a subset of the molecular characterization parameters based on the determined subspace. Typically, the predetermined criterion can also be based on other metrics defined above. Therefore, optimizing the similarity metric and optionally one or more of the metrics defined above with respect to the molecular characterization parameters also refers to optimizing the subspace of the parameter training space so that the explored data set covers the subspace as well as possible.

[0034] In an embodiment, providing an unexplored data set includes: a) receiving information indicating constraints for generating molecules, and b) generating an unexplored data set based on the received information. Constraints may refer to any constraints for generating molecules, for example, constraints provided by technical constraints of a production system for generating molecules, which refer to constraints on the availability of starting materials or catalysts or other necessary materials for generating molecules, and physical or chemical constraints that do not allow the generation of specific molecules. Such constraints may generally apply to all molecules, for example, if they refer to physical or chemical constraints and are therefore generally considered when generating an unexplored data set, or may refer to specific constraints, for example, constraints on the technical facilities of potential users that must be considered when generating an unexplored data set. For example, if a technical constraint indicates that a potential user's production facility is only allowed to produce molecules within a certain temperature range and only based on certain starting materials, then considering these constraints, the synthesis specifications of the unexplored data set can be easily generated, so that only molecules that can be generated in the corresponding production facilities of the potential user in the unexplored data set are considered. In a preferred embodiment, the constraints result in constraining the potential target molecule to a molecule for which the corresponding synthesis specifications are known or can be determined. Typically, the above constraints also allow the determination of the training data set for the determination model so that the determination model is trained specifically for the intended application, i.e. only for molecules that can be produced by the production facility. In addition, preferably, during the determination of the target molecule, the corresponding constraints are also taken into account to generate new potential target molecules, so that during the iteration only potential target molecules that can actually be produced by the corresponding target facility of the potential user are considered.

[0035] In an embodiment, parameterizing the determination model based on machine learning includes selecting a subset of molecular characterization parameters from the molecular characterization parameters of the explored data set as training molecular characterization parameters based on the explored data set and the unexplored data set, wherein the parameterization of the determination model based on machine learning utilizes a training data set, and the training data set includes a) technical application characteristics of at least two of the multiple explored molecules from the explored data set, and b) molecular characterization parameter values ​​of the training molecular characterization parameters, which are associated with the synthesis specifications and / or molecular structures corresponding to each of the at least two explored molecules. Preferably, selecting the subset includes: a) determining a parameter training space based on the unexplored data set, wherein the parameter training space is defined by the molecular characterization parameter values ​​of the molecular characterization parameters of the multiple unexplored molecules, b) determining a subspace of the parameter training space so that the molecular characterization parameter values ​​of the molecular characterization parameters of the explored data set cover the subspace according to a predetermined criterion, and c) selecting the training molecular characterization parameters based on the determined subspace. Further, preferably, the predetermined criterion refers to the similarity between members of the explored data set and members of the unexplored data set. In particular, the similarity may be determined using a similarity measure as already described above.In addition, the predetermined criterion may also refer additionally to other measures already defined above, and determining the corresponding subspace may refer to optimization of the similarity measure, and optionally additionally to optimization of any one of the other measurements already described above.

[0036] In another aspect of the present invention, a computer-implemented method for generating a determination model based on machine learning is proposed, so that the trained determination model is suitable for determining the technical application characteristics of a molecule based on one or more molecular characterization parameters, wherein the molecular characterization parameters are associated with the synthesis specifications and / or molecular structures corresponding to the molecule, wherein the method comprises: i) providing an explored data set, the explored data set comprising a) technical application characteristics of a plurality of explored molecules and b) a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with the synthesis specifications and / or molecular structures corresponding to each of the plurality of explored molecules, ii) providing an unexplored data set, the unexplored data set comprising a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with the synthesis specifications and / or molecular structures corresponding to each of the plurality of unexplored molecules, iii) generating the trained determination model by parameterizing the determination model based on the explored data set and the unexplored data set, so that the trained determination model is suitable for determining the technical application characteristics of the molecule based on one or more molecular characterization parameters of the molecule, and iv) providing the generated trained determination model. Preferably, providing the unexplored data set comprises generating the unexplored data set based on the intended application of the determined model.In particular, as already described above, a plurality of synthesis specifications and / or molecular structures may be generated based on the intended application, in particular based on constraints.

[0037] On the other hand, a computer-implemented method for determining one or more molecular characterization parameters for training a machine learning-based determination model is provided, so that the trained determination model is suitable for determining the technical application characteristics of a molecule based on the one or more molecular characterization parameters, wherein the molecular characterization parameters are associated with the synthesis specifications and / or molecular structure corresponding to the molecule, wherein the method comprises: i) providing an explored data set, the explored data set comprising a) the technical application characteristics of a plurality of explored molecules and b) a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with the synthesis specifications and / or molecular structure corresponding to each of the plurality of explored molecules, ii) providing an unexplored data set, the unexplored data set comprising a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with the synthesis specifications and / or molecular structure corresponding to each of the plurality of unexplored molecules, iii) selecting one or more molecular characterization parameters from the characterization parameters of the explored data set based on the explored data set and the unexplored data set, and iv) providing the selected molecular characterization parameters. In an embodiment, preferably, providing the unexplored data set comprises generating the unexplored data set based on the expected application of the determination model. In particular, as already described above, a plurality of synthesis specifications and / or molecular structures may be generated based on the intended application, in particular based on constraints.

[0038] Generally, the same embodiments and definitions described with respect to the method for determining a target molecule may also apply to corresponding features of the method for determining a determination model and for determining molecular characterization parameters for training a determination model.

[0039] On the other hand, an interface method for providing a deterministic model is proposed, wherein the interface includes: i) receiving an explored data set via an input unit, the explored data set including a) technical application characteristics of multiple explored molecules and b) multiple molecular characterization parameter values ​​of multiple molecular characterization parameters, the multiple molecular characterization parameters being associated with the synthesis specifications and / or molecular structures corresponding to each of the multiple explored molecules, ii) receiving information indicating an unexplored data set via the input unit, the unexplored data set including multiple molecular characterization parameter values ​​of multiple molecular characterization parameters, the multiple molecular characterization parameters being associated with the synthesis specifications and / or molecular structures corresponding to each of the multiple unexplored molecules, iii) interfacing with a processor that performs the method as described above via the interface unit to provide the explored data set and information and receive the trained deterministic model, and iv) providing the trained deterministic model via the output unit. In an embodiment, the interface method may include steps for providing the explored data set and information performed by the processor according to the method as described above.

[0040] In another aspect, an interface method for providing one or more molecular characterization parameters for training a machine learning-based determination model is proposed, wherein the interface includes: i) receiving an explored data set via an input unit, the explored data set including a) a plurality of technical application characteristics of a plurality of explored molecules and b) a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with a synthesis specification and / or a molecular structure corresponding to each of the plurality of explored molecules, ii) receiving information indicating an unexplored data set via the input unit, the unexplored data set including a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with a synthesis specification and / or a molecular structure corresponding to each of the plurality of unexplored molecules, iii) interfacing with a processor performing the method as described above via the interface unit to provide the explored data set and information and receive the determined molecular characterization parameters for training a machine learning-based determination model, and iv) providing the determined molecular characterization parameters via an output unit. In an embodiment, the interface method may include steps for providing the explored data set and information and receiving the determined molecular characterization parameters for training a machine learning-based determination model performed by a processor according to the method as described above.

[0041] In another aspect, an interface method for providing a target synthesis specification indicating a target molecule including a target technical application characteristic is provided, wherein the interface method comprises: i) receiving the target technical application characteristic via an input unit, ii) interfacing with a processor performing the method as described above via an interface unit to provide the target technical application characteristic and receive the determined target synthesis specification, and iii) providing the determined target synthesis specification via an output unit. In an embodiment, the interface method may include steps for providing the target technical application characteristic and receiving the determined target synthesis specification performed by a processor according to the method as described above.

[0042] On the other hand, a device for generating a determination model based on machine learning is proposed, so that the trained determination model is suitable for determining the technical application characteristics of a molecule based on one or more molecular characterization parameters, wherein the molecular characterization parameters are associated with the synthesis specifications and / or molecular structure corresponding to the molecule, wherein the device includes: i) an input interface, which is configured to provide an explored data set, the explored data set including a) technical application characteristics of multiple explored molecules and b) multiple molecular characterization parameter values ​​of multiple molecular characterization parameters for each of the multiple explored molecules, and provide an unexplored data set, the unexplored data set including multiple molecular characterization parameter values ​​of multiple molecular characterization parameters for each of the multiple unexplored molecules, ii) a processor, which is configured to generate the trained determination model by parameterizing the machine learning-based determination model based on the explored data set and the unexplored data set, so that the trained determination model is suitable for determining the technical application characteristics of the molecule based on one or more molecular characterization parameters of the molecule, and iii) an output interface, which is configured to provide the generated trained determination model.

[0043] On the other hand, a device for determining one or more molecular characterization parameters for training a determination model based on machine learning is proposed, so that the trained determination model is suitable for determining the technical application characteristics of the molecule based on the one or more molecular characterization parameters, wherein the molecular characterization parameters indicate the characterization features of the molecule and / or are obtainable from one or more characterization features of the molecule, wherein the device includes: i) an input interface, which is configured to provide an explored data set, the explored data set including a) technical application characteristics of multiple explored molecules and b) multiple molecular characterization parameter values ​​of the multiple molecular characterization parameters for each of the multiple explored molecules, and to provide an unexplored data set, the unexplored data set including multiple molecular characterization parameter values ​​of the multiple molecular characterization parameters for each of the multiple unexplored molecules, ii) a processor, which is configured to select one or more molecular characterization parameters from the characterization parameters of the explored data set based on the explored data set and the unexplored data set, and iii) an output interface, which is configured to provide the selected molecular characterization parameters.

[0044] On the other hand, a device for determining a target synthesis specification indicating a target molecule including a target technical application characteristic is provided, wherein the device comprises: i) an input interface configured to provide the target technical application characteristic, and to provide a digital representation of a potential target molecule, the digital representation indicating or being associated with a characterization parameter of the potential target molecule, ii) a processor configured to determine the technical application characteristic of the molecule using a trained determination model, wherein the trained determination model has been parameterized based on an explored data set and an unexplored data set, such that the trained determination model is suitable for determining the technical application characteristic of the molecule based on one or more molecular characterization parameters of the molecule, wherein the explored data set comprises a) a plurality of technical application characteristics of a plurality of explored molecules and b) a plurality of molecular representation parameters. The invention relates to a method of manufacturing a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameter values, wherein the molecular characterization parameters are associated with the synthesis specifications and / or molecular structures corresponding to each of the plurality of explored molecules, and the unexplored data set includes a plurality of molecular characterization parameter values ​​of the plurality of molecular characterization parameters, wherein the molecular characterization parameters are associated with the synthesis specifications and / or molecular structures corresponding to the plurality of unexplored molecules, and the determined technical application characteristics of the potential target molecule are compared with the target technical application characteristics, and based on the comparison, a) the potential target molecule is determined as the target molecule, or b) a new potential target molecule is provided and the determination of the technical application characteristics is repeated using the new potential target molecule, and iii) an output interface configured to provide the determined target molecule and the corresponding determined target technical application characteristics.

[0045] In another aspect, an interface device for providing a deterministic model is proposed, wherein the interface device comprises: i) an input unit configured to receive an explored data set, the explored data set comprising a) technical application characteristics of a plurality of explored molecules and b) a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with the synthesis specifications and / or molecular structures corresponding to each of the plurality of explored molecules, and receiving information indicating an unexplored data set, the unexplored data set comprising a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with the synthesis specifications and / or molecular structures corresponding to each of the plurality of unexplored molecules, ii) an interface unit configured to interface with a processor that performs the method as described above to provide the explored data set and information and receive the trained deterministic model, and iii) an output unit configured to provide the trained deterministic model. In an embodiment, the interface device may comprise a processor that performs the method as described above to provide the explored data set and information.

[0046] On the other hand, an interface device for providing one or more molecular characterization parameters for training a machine learning-based determination model is proposed, wherein the interface device includes: i) an input unit configured to receive an explored data set, the explored data set including a) technical application characteristics of multiple explored molecules and b) multiple molecular characterization parameter values ​​of multiple molecular characterization parameters, the multiple molecular characterization parameters being associated with synthesis specifications and / or molecular structures corresponding to each of the multiple explored molecules, and receiving information indicating an unexplored data set, the unexplored data set including multiple molecular characterization parameter values ​​of multiple molecular characterization parameters, the multiple molecular characterization parameters being associated with synthesis specifications and / or molecular structures corresponding to each of the multiple unexplored molecules, iii) an interface unit configured to interface with a processor that performs the method as described above to provide the explored data set and information and receive the determined molecular characterization parameters for training the machine learning-based determination model, and iii) an output unit configured to provide the determined molecular characterization parameters. In an embodiment, the interface device may include a processor that performs the method described above to provide explored data sets and information and receive determined molecular characterization parameters for training a machine learning based determination model.

[0047] In another aspect, an interface device for providing a target molecule including a target technical application characteristic is provided, wherein the interface device comprises: i) an input unit configured to receive the target technical application characteristic, ii) an interface unit configured to interface with a processor performing the method as described above to provide the target technical application characteristic and receive the determined target synthesis specification, and iii) an output unit configured to provide the determined target molecule. In an embodiment, the interface device may include a processor performing the method as described above to provide the target technical application characteristic and receive the determined target synthesis specification.

[0048] In another aspect, the use of the determined training molecular characterization parameters for training a determination model based on machine learning is proposed, wherein the training molecular characterization parameters have been generated using the apparatus and method as described above.

[0049] In another aspect, a computer program product for determining a target molecule comprising properties for a target technical application is proposed, wherein the computer program product comprises program code means for causing a computing system to perform the method as described above.

[0050] In another aspect, a control signal generated using the method as described above is proposed.

[0051] In another aspect, use of a control signal generated using the method as described above for controlling a production system, in particular a laboratory device, to produce a target molecule.

[0052] In another aspect, a trained deterministic model generated using the method described above is presented.

[0053] In another aspect, a data set comprising characterization parameters is presented, wherein the characterization parameters are determined using the method as described above.

[0054] In another aspect of the present invention, a device for optimizing a training data set for training a determination model based on machine learning is proposed, so that the trained determination model is suitable for determining the technical application characteristics of a molecule based on the one or more molecular physicochemical parameters, wherein the molecular physicochemical parameters indicate the physicochemical characteristics of the molecule and / or can be obtained from one or more physicochemical characteristics of the molecule, wherein the device comprises: i) an explored data providing unit for providing an explored data set, wherein the explored data set comprises a) characteristics of a plurality of molecules and b) a plurality of molecular physicochemical parameter values ​​of a plurality of molecular physicochemical parameters in the plurality of molecules, ii) a parameter training space providing unit for providing a parameter training space, wherein the parameter training space is about the determination model The intended application is defined in the molecular physicochemical parameter space, iii) a subspace determination unit for determining a subspace of the parameter training space defined by one or more molecular physicochemical parameters of the molecular physicochemical parameters defining the parameter training space, wherein the subspace is determined so that the training data set covers the subspace according to a predetermined criterion, and iv) a training molecule determination unit for determining the training molecular physicochemical parameters for training the determination model based on the determined molecular physicochemical parameters defining the determined subspace, and iv) a training data selection unit for selecting data from the explored data set so that the training data set includes the characteristics of a plurality of molecules and the molecular physicochemical parameter values ​​of the determined training molecular physicochemical parameters of the plurality of molecules. In particular, it is preferred that the parameter training space providing unit is suitable for determining the parameter training space based on the unexplored data set as defined above. Preferably, the predetermined criterion refers to the similarity between the explored data set and the unexplored data set, and the subspace is determined by optimizing the similarity measure between the explored data set and the unexplored molecular data set with respect to the molecular physicochemical parameters. Generally, the same embodiments and definitions as already described above are also applicable to this aspect.

[0055] In a preferred embodiment, the molecule is a polymer. In this embodiment, a computer-implemented method for determining a target synthesis specification indicative of a target polymer including a target technical application property is provided, wherein the method comprises: i) providing the target technical application property, ii) providing a digital representation of a potential target synthesis specification, the digital representation indicative of or associated with a characterization parameter of the potential target polymer, iii) determining the technical application property of the potential target polymer using a trained determination model, wherein the trained determination model has been parameterized based on an explored data set and an unexplored data set, such that the trained determination model is suitable for determining the technical application property of the polymer based on one or more polymer characterization parameters of the polymer, wherein the explored data set comprises a) a plurality of technical application properties of a plurality of explored polymers and b) a plurality of polymer characterization parameters of the plurality of polymer characterization parameters. The method comprises the steps of: determining a target polymer and a target synthesis specification for the unexplored polymer; determining a target polymer and a target synthesis specification for the unexplored polymer; and providing a target polymer and a target synthesis specification for the unexplored polymer. The method comprises the steps of: determining a target polymer and a target synthesis specification for the unexplored polymer; and providing a target polymer and a target synthesis specification for the unexplored polymer; and providing a target polymer and a target synthesis specification for the unexplored polymer. The method comprises the steps of: determining a target polymer and a target synthesis specification for the unexplored polymer; and providing a target polymer and a target synthesis specification for the unexplored polymer. The method comprises the steps of: generating a control signal based on the target synthesis specification and providing the control signal, wherein the control signal is configured to control the production system to produce the target polymer according to the target synthesis specification.In addition, a computer-implemented method for generating a determination model based on machine learning is proposed, so that the trained determination model is suitable for determining the technical application characteristics of a polymer based on one or more polymer characterization parameters, wherein the polymer characterization parameters are associated with the synthesis specifications corresponding to the polymer, wherein the method comprises: i) providing an explored data set, the explored data set comprising a) technical application characteristics of multiple explored polymers and b) multiple polymer characterization parameter values ​​of multiple polymer characterization parameters, the multiple polymer characterization parameters being associated with the synthesis specifications corresponding to each of the multiple explored polymers, ii) providing an unexplored data set, the unexplored data set comprising multiple polymer characterization parameter values ​​of multiple polymer characterization parameters, the multiple polymer characterization parameters being associated with the synthesis specifications corresponding to multiple unexplored polymers, iii) generating the trained determination model by parameterizing the determination model based on the explored data set and the unexplored data set, so that the trained determination model is suitable for determining the technical application characteristics of the polymer based on one or more polymer characterization parameters of the polymer, and iv) providing the generated trained determination model. Further, a computer-implemented method for determining one or more polymer characterization parameters for training a machine learning-based determination model is proposed, so that the trained determination model is suitable for determining the technical application characteristics of a polymer based on the one or more polymer characterization parameters, wherein the polymer characterization parameters are associated with the synthesis specifications corresponding to the polymer, wherein the method comprises: i) providing an explored data set, the explored data set comprising a) technical application characteristics of multiple explored polymers and b) multiple polymer characterization parameter values ​​of multiple polymer characterization parameters, the multiple polymer characterization parameters being associated with the synthesis specifications corresponding to each of the multiple explored polymers, ii) providing an unexplored data set, the unexplored data set comprising multiple polymer characterization parameter values ​​of multiple polymer characterization parameters, the multiple polymer characterization parameters being associated with the synthesis specifications corresponding to multiple unexplored polymers, iii) selecting one or more polymer characterization parameters from the characterization parameters of the explored data set based on the explored data set and the unexplored data set, and iv) providing the selected polymer characterization parameters. In addition, an interface method for providing a target synthesis specification indicating a target polymer including a target technical application characteristic is proposed, wherein the interface method comprises: i) receiving the target technical application characteristic via an input unit, ii) interfacing with a processor that executes the method according to any one of claims 1 to 8 via an interface unit to provide the target technical application characteristic and receive the determined target synthesis specification, and ii) providing the determined target synthesis specification via an output unit.Furthermore, a device for determining a target synthesis specification indicative of a target polymer including a target technical application characteristic is proposed, wherein the device comprises: i) an input interface configured to a) provide the target technical application characteristic, and b) provide a digital representation of a potential target synthesis specification, the digital representation indicating or associated with a characterization parameter of the potential target polymer, ii) a processor configured to a) determine the technical application characteristic of the polymer using a trained determination model, wherein the trained determination model has been parameterized based on an explored data set and an unexplored data set, such that the trained determination model is suitable for determining the technical application characteristic of the polymer based on one or more polymer characterization parameters of the polymer, wherein the explored data set comprises I) a plurality of technical application characteristics of the explored polymers and II) a plurality of polymer characterization parameter values ​​of the plurality of polymer characterization parameters, These polymer characterization parameters are associated with the synthesis specifications corresponding to each of the multiple explored polymers, and the unexplored data set includes multiple polymer characterization parameter values ​​for multiple polymer characterization parameters, which are associated with the synthesis specifications corresponding to multiple unexplored polymers, and b) comparing the determined technical application characteristics of the potential target polymer with the target technical application characteristics, and based on the comparison, I) determining the potential target polymer as the target polymer and determining the potential target synthesis specification as the target synthesis specification, or II) providing a new potential target synthesis specification for the potential target polymer, and repeatedly determining the technical application characteristics by using the new potential target synthesis specification of the potential target polymer, and iii) an output interface, which is configured to provide the determined target polymer and the corresponding determined target technical application characteristics. Further, a computer program product for determining a target synthesis specification of a target polymer indicating a target technical application characteristic is proposed, wherein the computer program product includes program code means for causing a computing system to execute the method as described above. In addition, the use of a control signal generated by the method as described above for controlling a production system, in particular a laboratory device, to produce a target polymer according to the target synthesis specification is proposed. It shall be understood that the method as described above, the device as described above and the computer program product as described above have similar and / or identical preferred embodiments, in particular as defined in the dependent claims.

[0056] It shall be understood that a preferred embodiment of the present invention may also be any combination of the dependent claims or the above-described embodiments with the corresponding independent claim.

[0057] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1Schematically and exemplarily shown is a method for determining target synthesis specifications indicative of target molecules including target technical application properties,

[0059] Figure 2 Schematically and exemplarily a method for providing a model for determining a technical application characteristic usable in a method for determining a target synthesis specification is shown,

[0060] Figure 3 The principles of the present invention are schematically and exemplarily shown.

[0061] Figures 4 to 6 Schematic diagrams and flow charts schematically and exemplarily show exemplary embodiments of the present invention, and

[0062] Figure 7 and Figure 8 A block diagram schematically and exemplarily illustrates an exemplary system architecture utilizing a system of the present invention. DETAILED DESCRIPTION

[0063] Figure 1 The computer-implemented method for determining the target synthesis specification of the target molecule indicating the target technology application characteristic is shown schematically and exemplarily. The method includes providing the target technology application characteristic, which can refer to any technology application characteristic that the corresponding target molecule should comply with. For example, the target technology application characteristic can be provided in the form of a target value together with a predetermined limit, and within the predetermined limit, the target molecule should comply with the target value. However, the target technology application characteristic can also be provided in the form of a target value range, wherein, in this case, the target molecule should include a technology application characteristic value that falls within the target range. In addition, the target technology application characteristic can also be provided in the form of an upper limit or a lower limit, so that the target molecule should include a target technology application characteristic value that is correspondingly lower than or higher than the corresponding upper limit or lower limit.

[0064] Further, the method includes providing a digital representation of a potential target molecule, which indicates or is associated with a characterization parameter of the potential target molecule. A digital representation may refer to any representation that allows a potential target molecule to be obtained from a digital representation, and in particular, a digital representation may indicate a chemical structure (i.e., a molecular structure) or a synthesis specification of a molecule. A characterization parameter of a potential target molecule may refer to any type of parameter that allows characterization of a feature of a potential target molecule in particular. For example, a characterization parameter may include a process parameter that characterizes at least a portion of a process that can produce a potential target molecule. Further, these characterization parameters may also include a recipe parameter that characterizes a recipe (e.g., a material utilized during a production process, such as a starting material or a catalyst) that characterizes the recipe according to which the potential target molecule can be produced. Preferably, these characterization parameters include the physicochemical parameters of the potential target molecule itself, such as a molecular descriptor, such as a quantum mechanical descriptor. Preferably, the characterization parameter indicates the molecular connectivity or other characteristics of a molecule. A digital representation of a potential target molecule indicates or is associated with a characterization parameter of a potential target molecule. For example, a characterization parameter of a potential target molecule may be derived from a molecular structure and / or a synthesis specification provided as a digital representation. In some embodiments, the characterization parameters of the present invention can be used to represent the characteristics of the target molecule. However, when the potential target molecule is known, at least some characterization parameters may also be a part of the digital representation of the potential target molecule, or may be available on the corresponding database. If necessary, it is possible to utilize the commonly known method for obtaining the characterization parameters from the target molecule, such as 3D simulation, quantum mechanical simulation, electronic structure calculation, etc., for example, based on molecular structure and / or synthesis specification. However, the characterization parameters may also be directly read from the potential target molecule, such as, when the characterization parameters relate to the molecular connectivity or atomic quantity that are usually provided as a part for molecular identification. In addition, the suitable characterization parameters of a plurality of potential target molecules may have been stored, and then the characterization parameters associated with specific potential molecules may be provided by accessing the corresponding storage device and selecting them.

[0065] In the next step, the technical application characteristics of the potential target molecule are determined using the determination model, wherein the trained determination model has been parameterized based on the explored data set and the unexplored data set, so that the trained determination model is suitable for determining the technical application characteristics of the molecule based on one or more molecular characterization parameters of the molecule. Figure 2The further details of training, especially parameterization, of the determination model are described. Generally, a determination model suitable for the intended application is utilized, that is, the determination model has been trained to determine the technical application characteristics related to the target technical application characteristics based on the obtained or provided characterization parameters, and is generally applicable to potential target molecules. Therefore, generally for different molecule types, different technical application characteristics to be determined and / or different characterization parameters, different determination models can be stored in, for example, a storage unit, and can then be used to determine the technical application characteristics. The determination model is a parameterized data-driven model, so that the model can determine the technical application characteristics associated with the molecule based on a digital representation (for example, based on a molecular descriptor indicating the physicochemical characteristics of a subgroup of molecules as a polymer). In a preferred embodiment, the data-driven model refers to a machine learning model, for example, an algorithm based on a regression model or an algorithm based on a classifier model. The algorithm based on the regression model can be based on any one of a neural network algorithm, a LASSO algorithm, a ridge regression algorithm, a MARS algorithm, and a random forest algorithm. The model algorithm based on the classifier can be based on any one of a random forest algorithm and a SVM algorithm. The inventor has found that for most applications, particularly a random forest algorithm and an algorithm based on MARS are suitable. Based on the characterization parameters associated with or indicated by the potential target polymer and the determined model, the corresponding technological application properties of the potential target molecule are determined.

[0066] In a further step, determined technical application characteristics are compared with target technical application characteristics. In the case where the determined technical application characteristics of potential target molecule do not meet the target technical application characteristics, new potential target molecule is provided, particularly including new molecular structure and / or synthesis specification. For example, new potential target molecule can be provided from a storage device having stored a plurality of potential target molecules. Then new potential target molecule can be selected arbitrarily or based on a predetermined rule from a plurality of potential target molecules having stored. However, new potential target molecule can also be generated, for example, by modifying one or more of the parameters of the potential target synthesis specification and / or molecular structure previously utilized based on a predetermined rule, in the form of new potential target synthesis specification and / or molecular structure. Then, the technical application characteristics are determined by utilizing a determination model repeatedly with new potential target molecule. If the determined technical application characteristics meet the target technical application characteristics within predetermined limits, the potential target molecule is determined as a target molecule, and for example, this determined target molecule is provided in the form of an executable control file, which can for example be used in a laboratory to control the synthesis process for synthesizing a target molecule.

[0067] Figure 2A computer-implemented method for providing a determination model is schematically and exemplarily shown, and the determination model can be used in the above-mentioned computer-implemented method. In particular, the provided determination model is parameterized based on an explored data set and an unexplored data set. The method includes providing an explored data set, the explored data set including a) a plurality of technical application characteristics of a plurality of explored molecules and b) a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with the synthesis specifications and / or molecular structures corresponding to each of the explored molecules in the plurality of explored molecules. Further, an unexplored data set is provided, the unexplored data set including a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with the synthesis specifications and / or molecular structures corresponding to each of the unexplored molecules in the plurality of unexplored molecules. In particular, at least some of the molecular characterization parameters in the unexplored data set and the molecular characterization parameters whose characterization parameter values ​​are provided in the explored data set refer to the same molecular characterization parameters. Then, the explored data set and the unexplored data set thus provided are used to generate a trained determination model by parameterizing the determination model based on the machine learning based on the explored data set and the unexplored data set, so that the trained determination model is suitable for determining the technical application characteristics of the molecule based on one or more molecular characterization parameters of the molecule. In general, the inventors have found that parameterizing the determination model based on the explored data set and the unexplored data set not only allows the training process of the determination model to be more efficient, i.e., consumes less computing resources and less time, but also further makes the trained determination model better suitable for the intended application context, and thus produces more accurate results when using the trained determination model to determine the technical application characteristics of the molecule.

[0068] In particular, it is preferred that the explored data set and the unexplored data set are used to determine a training data set for training a determined model, the determined model being specifically suitable for an application context in which the determined model should be used. Figure 3 The context of determining a training data set based on an unexplored data set and an explored data set is exemplarily and schematically shown. Figure 3Two different sections, i.e., subspaces, of the space spanned by all characterization parameters provided by the explored and unexplored datasets are shown. In this illustrative example, a section 310 of the space spanned by characterization parameter 1 and characterization parameter 2 is shown. Further, a section 320 spanned by characterization parameter 3 and characterization parameter 4 is shown. Based on the values ​​provided for the corresponding characterization parameters in the unexplored and explored datasets, members of the unexplored and explored datasets can be recorded in the corresponding sections 310, 320. In particular, each point or cross shown in the corresponding sections 310, 320 can be regarded as being associated with a molecule in the unexplored or explored dataset. In the example shown here, the cross refers to the explored dataset, and the point refers to the unexplored dataset. Typically, an explored dataset refers to a molecule for which at least one technical application property is known (e.g., has been measured), and thus defines a dataset that can generally be used to train a determination model due to the known technical application property. As Figure 3 As exemplarily shown in , in section 310 based on characterization parameters 1 and 2, for this combination of characterization parameters, the explored molecules (i.e., members of the explored data set) can only be found in a specific region 311 (i.e., a subspace) of section 310. An unexplored data set generally refers to molecules, in particular molecular structures and / or synthesis specifications, that are artificially generated in huge quantities. For example, more than one thousand unexplored molecules, in particular molecular structures and / or synthesis specifications can be generated to form an unexplored data set. The unexplored data set is then used to define the intended technical application context of the determined model. For example, taking into account specific production constraints or application constraints, an unexplored data set can be generated to include molecular structures and / or synthesis specifications and associated molecules. Therefore, it is preferred that the training data set used to train the determined model covers a space defined by members of the unexplored data set according to predetermined criteria. As Figure 3 As exemplarily shown in , for the first section 310, i.e. for the first subspace of the characterization parameter space defined by the characterization parameters, the members of the explored data set and the members of the unexplored data set do not mainly cover the same area, i.e. the explored data set covers area 311, while the unexplored data set covers area 312. Typically, for this case, a predetermined criterion will determine that the degree of coverage is not good enough. In contrast, in section 320, i.e. in the subspace of the characterization parameter space spanned by characterization parameters 3 and 4, it is shown that the area 322 defined by the unexplored data set is also well covered by the members of the explored data set. Therefore, changing the perspective (i.e. selecting the corresponding subspace of the characterization parameter space) allows finding such characterization parameters, in particular combinations of characterization parameters: for these characterization parameters, the explored data set is similar to the unexplored data set and is therefore very suitable for determining the intended technical application of the model. Therefore, for Figure 3In the example shown, the training data set may be selected to include at least two technical application characteristics of two explored molecules of the explored data set, and further include values ​​of characterization parameters 3 and 4 of the corresponding explored molecules.

[0069] According to the above Figure 3 The principle described, utilizes an unexplored data set and an explored data set to select a training data set for parameterizing a determined model. In particular, the selection refers to selecting characterization parameters in the training data set for training the determined model. Preferably, these characterization parameters (i.e., a corresponding subset of these characterization parameters) are selected based on a similarity measure, which indicates the similarity between members of the explored data set and members of the unexplored data set with respect to the corresponding subset of these characterization parameters. One possibility of mathematically finding the best subset of characterization parameters so that the members of the explored data set are as similar as possible to the members of the unexplored data set involves defining an optimization function. For example, the possibility of an optimization function can be defined as

[0070]

[0071] in, and are the molecular characterization parameters from the unexplored (scr) and explored (tr) datasets, where f refers to a specific characterization parameter, and i, j refer to the corresponding molecules from the unexplored and explored datasets, respectively, and the coefficient c f can be 0 or 1 - depending on whether the corresponding characterization parameter is selected, and the sum of f is equal to the number of characterization parameters K in the selected subset of characterization parameters, D is the universal distance function between any given data point in the space defined by the selected subset of characterization parameters, and S is the number of data points in the unexplored dataset. The phase function can then be optimized with respect to the selected subset of characterization parameters so that the members of the explored dataset are as similar as possible to the members of the unexplored dataset.

[0072] Furthermore, the optimization function can be modified to take into account further advantageous features. In particular, a subset can be selected by further optimizing the determination accuracy of the model with respect to the subset of molecular characterization parameters. For example, for this case, the optimization function can be defined as

[0073]

[0074] in, refers to the technical application characteristics determined by a determined model, which is trained with a corresponding training data set based on the corresponding selected subset, Y j trrefers to the technical application characteristics of the corresponding molecules, T is the number of data points in the explored dataset, D* is a universal distance measure between the technical application characteristics of the molecules of the explored dataset determined by the trained deterministic model and the technical application characteristics of the molecules of the explored dataset, and the parameter λ refers to the weighting of the similarity between the technical application characteristics determined by the deterministic model and the technical application characteristics of the explored dataset. In addition, the subset can be selected by further optimizing on the fitness of the deterministic model in the unexplored dataset. In this case, the optimization function can be defined as

[0075]

[0076] in, refers to the technical application characteristics determined by the trained deterministic model, Y j tr refers to the technical application characteristics of the corresponding molecule, G is a function for evaluating the fitness of the determined model in the unexplored data set, S is the number of data points in the unexplored data set, and the parameter μ refers to the weighting of the fitness measure of the determined model in the unexplored data set. However, other mathematical formulas may also be suitable as optimization functions or for obtaining a corresponding subset of characterization parameters for training the determined model.

[0077] The corresponding selected subset of characterization parameters then allows definition of a training data set for parameterizing the determination model, wherein the training data set comprises technical application characteristics of at least two explored molecules and characterization parameter values ​​associated with the at least two explored molecules, the characterization parameter values ​​referring to the selected subset of characterization parameters. The determination model can then be parameterized based on the training data set using known machine learning and training techniques. Then, the determination model can be parameterized, for example, by Figure 1 The target molecule identification method provides and utilizes such a trained identification model.

[0078] Preferred examples of the above-mentioned method and corresponding apparatus will be described in more detail below. In particular, for example, the method as described above can be used to determine molecules for target technical application properties, which molecules for target technical application properties belong to at least the group of the following items: a) mechanical properties, such as adhesion, tensile strength, stiffness, hardness, shrinkage, elongation, crack tear, tear strength, rebound, compressibility, wear, overflow, morphology, tactile properties, breaking stress, elongation at break, particle size, filling degree, b) optical properties, such as coloration, turbidity, opacity, transparency, reflection, appearance, absorption, scattering, color intensity, cloud point, extinction, optical density, spectrum, refractive index, c) physicochemical properties, such as density, viscosity, K value, molar mass, dispersibility, molar mass distribution, particle size distribution, solubility, partition coefficient, interface properties, surface tension, dispersibility, storage stability, odor, segregation, conductivity, capacitance, surface area, flow time, vapor pressure, VOC, solid content, hygroscopicity, magnetism, miscibility, thixotropy, phase change properties, glass transition temperature, corrosion inhibition, solvent separation, aggregation, self-heating energy force, impact sensitivity, drying loss, response angle, static charge, minimum film forming temperature and, charge density, dart impact strength, melt volume rate, flowability, tear propagation resistance, seal strength, permeability, d) chemical properties, such as functional group count, atom type count, functional group density, atom type density, chemical resistance, reaction time, demolding time, growth, hard / soft segment content, crystallinity, reaction temperature, reaction pressure, decomposition, thermal decomposition, photodegradation, acidity, pKa, pH, carbon footprint, production cost, waste formation, moisture / water content, flammability, burning rate, autoignition, flash point, formation of flammable gases, reaction to fire, deflagration rate, residual monomer count, by-product formation, degree of polymerization, salt content, temperature resistance, oxidizing properties, reducing properties, reactivity, ash content, non-volatile matter content, stability, chelating ability, calorific value, saponification value, and e) biological properties such as biodegradability, bioresistance, toxicity, biotransformation, ecotoxicology, sensitization, bacterial count, enzyme activity, distribution in the environment, bioaccumulation.

[0079] Therefore, this method can be applied in several technical fields like agricultural molecules, coatings, dispersions, structural molecules, soluble molecules, polymer foams (e.g. for thermal and acoustic insulation, shoes and automotive applications).

[0080] An exemplary embodiment of a method for determining a technical application characteristic using a determination model trained as described above may consist of the steps described below. In particular, this method may be used in embodiments where the molecule is a polymer, such as in Figure 1 The technical application characteristics are determined in the method described. Figure 4A schematic and exemplary flow chart of an exemplary embodiment of the method 400 is shown in FIG. In a first step 410, a digital representation indicating a polymer is provided. The digital representation may directly include polymer characterization parameters, in particular polymer descriptors, wherein, in this case, the polymer descriptors may be omitted. Figure 4 The steps shown are up to step 450. However, in many cases, polymer characterization parameters must first be determined based on the provided digital representation, in which case these polymer characterization parameters refer, for example, to the synthesis specifications used to synthesize the polymer. In general, the digital representation may include or allow for any one or more of the following information: the amount of monomer components; the amount of non-monomer components like initiators, fillers, additives; reaction conditions, like temperature, container, pressure, stirring rate; condition profile, such as temperature profile, pH value, solvent; feed profile; polymerization type, such as free radical, cationic, anionic, polycondensation, polyaddition, polyether formation; post-treatment, such as the amount of components, conditions and temperature and feed profile; type of post-treatment, such as free radical, cationic, anionic, polycondensation, polyaddition, polyether formation; chemical information about the components, such as mixtures, connectivity of non-polymer pure compounds, composition of polymer pure compounds based on subgroups, connectivity of monomers associated with subgroups in polymer pure components; for block copolymers, also information in which block each monomer and reactive prepolymer is contained; for structured / layered materials and composites, also information in which phase / layer each component is contained. If such information is not directly provided by the digital representation, in optional step 420, reactive components and subgroups may also be derived from the digital representation (eg, from a synthesis specification).

[0081] If the information provided indicates that a mixture is present, then in the next step the mixture is resolved into its pure components and each polymer component is considered as input polymer. In addition, the polymer composition can also be converted to mol % if desired.

[0082] In the next step 430, the polymerizable components can be converted into subgroups, such as repeating units, and the subgroups are determined as different types. For example, the polymerizable subgroups can be determined based on the connectivity information of the non-polymer pure compound by using SMARTS, for example, via the KNIME workflow. In addition, the connectivity information of all possible subgroups can also be obtained from the connectivity information of the non-polymer pure compound by using reaction SMARTS (e.g., also via the KNIME workflow).

[0083] After the subgroups and their types have been determined, in step 440, the type of characterization parameters that should be utilized can be provided. However, the characterization parameters can also be determined without first selecting the subgroup type. In order to reduce the computational resources of the method, it is preferred that in step 441, it is determined whether the subgroup characterization parameters associated with the corresponding type of subgroup are already stored in the database, for example, it can be determined whether the entries of the subgroups with the same connectivity information already exist in the database. If this is the case, the corresponding associated subgroup characterization parameters can be downloaded directly, for example in step 444. If the determined subgroup type is not stored in the database, the subgroup characterization parameters associated with the corresponding subgroup type can be determined, for example in step 442. For example, the 3D structure of the corresponding type of subgroup can be obtained based on the connectivity information, and the automatic calculation of the subgroup characterization parameters can be started using, for example, computer clustering, or the results can be determined as subgroup characterization parameters using existing machine learning. Generally, it is preferred that if calculations on new subgroups are necessary, in step 443, after the calculations are completed, the results are stored in the database. Alternatively, additional subgroup characterization parameters can be provided by topological analysis of the subgroup, quantum chemical calculations, molecular dynamics calculations, coarse-grained methods, finite element calculations and kinetic simulations. In particular, polymer reaction engineering methods can be used to obtain subgroup characterization parameters that allow consideration of polymer microstructure.

[0084] In step 431, determine the amount of subgroups, i.e., the amount of each type of subgroup, for example, based on the synthesis specifications of the provided polymer. For example, the amount can be determined by calculating the amount of polymerizable groups of each polymerizable component (optionally including prepolymers). In this case, the information of the polymerizable groups can be obtained from the non-polymer component, and the amount thus determined can be added to the number count of the optionally non-polymerizable polymerizable groups of the polymer component subgroup based on the composition of the polymer component to determine the resulting amount. Further, it is preferred to remove the amount of polymerizable groups derived from the reagents used for post-treatment after polymerization from the resulting amount. In step 432, the subgroup amount thus determined can be provided and, for example, saved in a database. Before further processing the determined subgroup amount, the subgroups represented entirely by other subgroups can be deleted. In addition, subgroups with the same connectivity can also be merged.

[0085] Alternatively, the obtained subgroup amounts can be used for further interpretation of the polymer composition. For example, the total number of polymerized functional groups, such as double bonds, amine groups, alcohol groups, thiol groups, carboxylic acid groups, isocyanate groups, epoxy groups, and formed functional groups, such as amide groups, ester groups, thioester groups, urea groups, carbamate groups, thiocarbamate groups, ether groups can be determined. In addition, the mole-weighted total number of polymerized functional groups, the mass-weighted total number of polymerized functional groups, the total number of residual functional groups, such as double bonds, amine groups, alcohol groups, thiol groups, carboxylic acid groups, isocyanate groups, epoxy groups, the mole-weighted total number of residual functional groups, the mass-weighted total number of residual functional groups, the sum of all residual functional groups, the ratio between functional groups after polymerization, the number of crosslinks in the polymer, the mole fraction of crosslinks in the polymer (optionally, also weighted by mass), the average number of atoms per subgroup (optionally, also by weight), the average number of non-H atoms per subgroup (optionally, also by weight), the average number of bonds per subgroup (optionally, also by weight), the average number of bonds between non-H atoms per subgroup (optionally, also by weight), the average number of rotors per subgroup (optionally, also by weight), the average number of rotors between non-H atoms per subgroup (optionally, also by weight), the average number of rings per subgroup (optionally, also by weight), the average number of atoms per subgroup (optionally, also by weight), the average number of The average polar surface area of ​​the group (optionally, also by weight), the average refractive index of each subgroup (optionally, also by weight), the total number of blocks, the molar size of the first block, the molar size of the last block, the HLB value of the polymer (optionally, as an area-weighted HLB value), the HLB value of the block with the lowest HLB value (optionally, as an area-weighted HLB value), the HLB value of the block with the largest HLB value (optionally, as an area-weighted HLB value), the HLB value of the first block (optionally, as an area-weighted HLB value), the HLB value of the last block (optionally, as an area-weighted HLB value), the mass of the first block, the mass of the last block, the area of ​​the block with the lowest HLB value, the area of ​​the block with the largest HLB value, the difference in HLB values ​​of the blocks (optionally, as an area-weighted HLB value), the hydrophilic area of ​​the polymer, the lipophilic area of ​​the polymer, the number of arms for ring-opening polymerization or the arm length for ring-opening polymerization.

[0086] In step 450, the determined amount and type of subgroups and the associated subgroup characterizing parameters can be used to calculate polymer characterizing parameters. For example, the polymer characterizing parameters can be determined by performing one or more of mole weighting (e.g., arithmetic, harmonic, or logarithmic average), mass weighting (e.g., arithmetic, harmonic, or logarithmic average), volume weighting (e.g., arithmetic, harmonic, or logarithmic average), and surface area weighting (e.g., arithmetic, harmonic, or logarithmic average) on the associated descriptors of the subgroups. In addition, the polymer characterizing parameters can be determined by determining one or more of a mole weighted standard deviation, a mass weighted standard deviation, a volume weighted standard deviation, a surface area weighted standard deviation, a mole weighted maximum, a mass weighted maximum, a volume weighted maximum, a surface area weighted maximum, a mole weighted minimum, a mass weighted minimum, a volume weighted minimum, a surface area weighted minimum, a mole weighted sum, a mass weighted sum, a volume weighted sum, a surface area weighted sum, and a maximum difference from the associated subgroup characterizing parameters.

[0087] Then, in step 460, the obtained or provided polymer characterization parameters can be provided to the trained determination model to determine the application characteristics. Generally, as described above, for determining different polymer characteristics, determination models trained in different ways can also be used. The determination model is trained using, for example, a training data set determined based on an unexplored data set and an explored data set, as described above with respect to Figure 2 and 3 Described. The determination model can generally refer to sparse, such as spline function, LASSO regression, PLS, and non-sparse, such as ridge regression, tree method, kernel-based method, statistical learning model for associating polymer characterization parameters with application characteristics of interest. In addition, the determination model can further provide a reliability estimate of the determined technical application characteristics, depending on the corresponding determination model used. In step 470, the determined technical application characteristics can be provided to the user, for example, via a user interface, or can be further used for comparison with the target technical application characteristics, such as with respect to Figure 1 In another application, embodiments of the above method can be used for virtual screening using, for example, one of the following: composition optimization, Pareto optimization, low-dimensional visualization of the screened formulations, feature selection for achieving the maximum applicable domain of the model, and a domain checker for the applicability of new formulations at the descriptor level.

[0088] Preferably, the molecular characterization parameters (in the above examples, polymer characterization parameters) utilized in the above method are derived from quantum chemical calculations with solvation treatment. The scale of quantum chemical calculations is very disadvantageously scaled with the system size, which makes the calculation of polymers or shorter monomer sequences impractical. This obstacle is solved by the above method, which includes cutting the polymer into subgroups at a preferably non-polarized bond. The resulting subgroups have a size similar to that of the monomer, and the characterization parameters can be calculated using quantum chemical methods.

[0089] In the following, a more detailed preferred example of a preferred embodiment of the method using the determination model trained as described above is described. Figure 5 A schematic and exemplary flow chart of an exemplary and preferred embodiment of the method is provided. In this exemplary embodiment, the method starts from, for example, requesting a target value for a target application via a user interface. In addition, in the next step, optimization is initialized by providing a potential target molecule, for example, by providing synthesis specifications and / or molecular structure (i.e., initial synthesis specifications). Optionally, for example, if a user provides constraints on the chemical structure of the potential target molecule, such constraints can be considered in this process. Constraints can refer to, for example, constraints on molecular production, constraints on the starting materials that should be used to synthesize molecules, etc. In addition, additional application conditions that particularly indicate further possible information about the target molecule that should be met can also be requested. Based on the above steps, an optimization for determining a target molecule, i.e., a target synthesis specification and / or molecular structure, can be initialized. In the first step of optimization, it is possible to initialize the optimization from the provided initial potential target molecule to a molecular characterization parameter value (e.g., such as about Figure 4 For polymer described in more detail). However, obtaining molecular characterization parameters may also refer to accessing a storage device that has stored corresponding molecular characterization parameters of corresponding potential target molecules. In addition, if the digital representation of the potential target molecule provided already includes molecular characterization parameters, this step may also be omitted. Based on the requested additional application conditions, a corresponding determination model may be provided. Based on the provided determination model and the digital representation of the potential target molecule, the value of the target technical application characteristics of the potential target molecule may be provided. In the next step, the determined performance value is determined, that is, whether the determined technical application characteristics meet the target value within predetermined limits. If this is not the case, that is, if the condition is not met, the formulation of the potential target molecule is modified, and a new target molecule is optionally determined taking into account the constraints previously provided. Then, a new iteration may be started for the new potential target molecule. If the determined performance value at a certain moment meets the target value within limits, that is, if the corresponding conditions are met, the potential target molecule is determined as the target molecule, and provided, for example, to a user or a control unit to produce the corresponding determined target molecule.

[0090] Figure 6Another preferred embodiment of the above method for determining a target molecule having a predetermined first target technical application characteristic is schematically and exemplarily shown, wherein, in this embodiment, in addition to the first target technical application characteristic, the target molecule is expected to also satisfy another second target value, i.e., the target technical application characteristic. The additional target technical application characteristic may also refer to any technical application characteristic. In particular, the method follows the above description of Figure 5 The same principle described. However, due to the additional target value, additional conditions must be met during optimization. Therefore, hereinafter, only the main differences relative to the above method are pointed out. In particular, in this preferred embodiment, the optimizer module is optimized not only on the first target value, i.e., on the first target application characteristic, but also on the second target value. Preferably, for the second target value, a determination model suitable for determining the value of the technical application characteristic based on the molecular physicochemical parameters is also utilized. Therefore, in addition to the above method, for the second target application, a second determination model for the second target application is also provided, which allows the application characteristic value to be determined based on the molecular physicochemical parameters. The second determination model can be, for example, based on the same algorithm as the first determination model, and is only trained according to the same principle as described above for determining another technical application characteristic. Then, the comparison refers not only to determining whether the determined first application characteristic meets the target application characteristic within the limit, but also to determining whether the determined second application characteristic value meets the target second application characteristic value within the limit. Predetermined rules can be utilized, which determine under which circumstances to continue iteration (i.e., provide new potential target molecules), and under which conditions to determine the potential target molecule as the target molecule. For example, the user can predetermine the weights used for weighting, that is, which conditions must be met to what extent. For example, it may be more important for the user to meet the first technical application characteristics, while other target application characteristics are less important. In this case, the restrictions that the second target application characteristics can meet can be set wider, or the compliance with this condition can be weighted less. Then, if it is determined that the conditions are met and the predetermined rules are met at a certain point in the iteration, the corresponding potential target molecule can be determined as the target molecule and provided to the user as an output, or the potential target molecule can be used to generate a control file to produce the corresponding target molecule.

[0091] Figure 7A block diagram of an exemplary system architecture of an automated laboratory system 1000 is illustrated for synthesizing molecules using a laboratory equipment control device 1102, a network 1150, and a synthesis specification (i.e., recipe) module 1100 / 1110 and a client device 1108. The automated laboratory system includes a laboratory equipment control device layer 1152 (as part of the laboratory equipment control device 1102), a synthesis specification module layer 1154 associated with the synthesis specification module, and a remote control or client layer 1156 associated with the client device 1108. The laboratory equipment control device layer can be divided into multiple hierarchical layers: a hardware layer, a middleware layer, and an interface layer. The hardware layer is associated with hardware resources such as sensors and actuators, particularly for controlling the synthesis of molecules. Middleware is associated with any known middleware for laboratory or factory synthesis operations. An example is LABS / QM, which provides different abstractions to hardware, networks, and operating systems, such as low-level device control and message passing. The communication layer is related to the communication protocol, which can be REST, which can be implemented through different transport protocols (i.e. UDP, TCP, telemetry) that allow the exchange of messages between the laboratory equipment control device and the laboratory equipment device. This software architecture allows the control and monitoring of laboratory equipment without the need to interact with the hardware.

[0092] The synthesis specification module layer 1154 may include: a large-capacity storage layer, a computing layer, and an interface layer. The storage layer is configured to provide large-capacity storage to the data-driven determination model so as to provide the synthesis specifications of the molecule based on the technical application characteristics as described above. In particular, as described above, the functions performed by the device can be provided as program code devices stored on a large-capacity storage device. In addition, the synthesis specifications of a variety of molecules can be stored in a large-capacity storage device. Such data can be stored in a structured database (such as an SQL database), or in a distributed file system (such as HDFS, NoSQL database (such as HBase, MongoDB), etc.). The computing layer may include an application layer, which allows the functions provided by the standard cloud service to be customized based on the target characteristics to perform the calculation process. Such functions may include: determining the digital representation of the target molecule based on the target technical application characteristics and the determination model, generating synthesis specifications from the digital representation of the target molecule, and providing the synthesis specifications as control data (i.e., control signals) to the laboratory equipment control device.

[0093] The interface layer can implement network services, network interfaces as UDP or TCP or Websocket interfaces. In order to communicate with the laboratory equipment control device, a REST API is implemented.

[0094] The client layer 1156 provides an interface for the end user. For the end user, the client layer 1156 can run client-side network applications that provide interfaces to the synthetic specification module layer 1154 or the laboratory equipment control device layer 1152. A UI for selecting a target technical application characteristic and a target value for this characteristic can be provided to the user, and the target value can also include a range of technical application characteristics. In other instances, a UI for selecting more than one technical application characteristic and corresponding values ​​can be provided to the user. The application can be configured for the user to remotely monitor and control the laboratory equipment control device and operation. In other instances, the client device layer and the synthetic specification module layer can be integrated into one device. The alternatives described here are for illustrative purposes only and should not be considered restrictive.

[0095] Figure 8 A block diagram of an exemplary system architecture is illustrated, with a system and apparatus for generating a determination model for determining technical application characteristics, a network 2150, and a model generation module 2100 / 2110 that can be viewed as or include a training model apparatus, a synthesis specification module 1100 / 1110, and a client device 2108. The system for generating a determination model includes a model generation module layer 2154 (as part of the model generation module) and a client layer 2156 associated with the client device 2108.

[0096] The model generation module layer 2154 may include: a large-capacity storage layer, a computing layer, and an interface layer. The storage layer is configured to provide large-capacity storage to the data-driven determination model as described above. Further, the large-capacity storage is configured to store the synthesis specifications and technical application characteristics of the molecule. Such data can be stored in a structured database (such as an SQL database), or in a distributed file system (such as HDFS, NoSQL database (such as HBase, MongoDB), etc.). The computing layer may include an application layer that allows customization of the functions provided by the standard cloud service to perform the calculation process, thereby generating a determination model for determining the characteristics of the molecule. Such functionality may include: receiving digital representations of at least two previously explored molecules each associated with the molecule, and measurement data of at least one technical application characteristic of each of the at least two previously explored molecules; receiving a digital representation of at least one unmeasured molecule at a model generation module; training a model based on the digital representations of at least two previously explored molecules, at least one technical application characteristic of each of the at least two previously explored molecules, and preferably a measure of similarity between the digital representations associated with the at least two previously explored molecules and the corresponding digital representations associated with at least one unmeasured molecule according to the above-mentioned training principles; and providing a determined model of technical application characteristics via an output interface. The model generation module layer may be configured to deploy the generated model and synthesis specification database to the synthesis specification module layer. This may include storing the generated model and synthesis specification database in a mass storage device associated with the synthesis specification module.

[0097] The model generation module layer can be further configured to determine the digital representation of the molecule associated with the synthesis specification according to the synthesis specification. The digital representation can include a set of molecular characterization parameters and molecular characterization parameter values, and these parameter values ​​are associated with the synthesis specification of each explored molecule. A way to obtain these molecular characterization parameters can be to apply the SMILES algorithm or any other principle described above. In the case where the model is generated based on the digital representation obtained from the synthesis specification, the relationship between the synthesis specification and the characterization parameters can be stored in a large-capacity storage device associated with the model generation module. In this case, the deployment model includes providing this relationship.

[0098] The interface layer can implement network services, network interfaces as UDP or TCP or Websocket interfaces. In order to communicate with the client device, a REST API is implemented in this example. The client layer 2156 provides access to large-capacity storage devices, which contain synthetic specifications of molecules and, for at least two molecules, contain at least one technical application characteristic. The client layer further provides an interface for end users. For end users, the client layer 2156 can run client-side network applications, which provide interfaces to the model generation module layer 2154 or large-capacity storage devices associated with the client layer. A UI for selecting technical application characteristics can be provided to the user. A UI for selecting molecular data and technical application characteristic data associated with the molecular data can be further provided to the user. The user interface can also provide options for uploading selected data to the model generation module layer, and optionally options for starting model generation.

[0099] Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.

[0100] For the processes and methods disclosed herein, the operations performed in the processes and methods may be implemented in different orders. In addition, the operations outlined are provided only as examples, and some of the operations may be optional, may be combined into fewer steps and operations, may be supplemented with more operations, or may be expanded into more operations without departing from the essence of the disclosed embodiments.

[0101] In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality.

[0102] A single unit or device may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0103] The procedures such as providing digital representation and target technology application characteristics, determining technology application characteristics, comparing technology application characteristics, etc. performed by one or more units or devices may be performed by any other number of units or devices. These processes may be implemented as program code means of a computer program and / or dedicated hardware.

[0104] The computer program product may be stored / distributed on a suitable medium, such as an optical storage medium or a solid-state medium provided together with or as a part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems.

[0105] Any unit described herein may be a processing unit as part of a classical computing system. The processing unit may include a general-purpose processor, and may also include a field programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or any other dedicated circuit. Any memory may be a physical system memory, which may be volatile, non-volatile, or some combination of the two. The term "memory" may include any computer-readable storage medium, such as a non-volatile mass storage device. If the computing system is distributed, the processing and / or storage capabilities may also be distributed. The computing system may include multiple structures as "executable components". The term "executable component" is a structure that is fully understood in the computing field as a structure that can be software, hardware, or a combination thereof. For example, when implemented in software, a person of ordinary skill in the art will understand that the structure of the executable component may include software objects, routines, methods, etc. that can be executed on the computing system. This may include executable components in the computing system stack or on a computer-readable storage medium. The structure of the executable component may exist on a computer-readable medium so that when interpreted by one or more processors of the computing system (e.g., by a processor thread), the computing system performs a function. Such a structure may be directly computer-readable by a processor, for example, as in the case where the executable component is binary, or it may be constructed to be interpretable and / or compiled, for example, whether in a single level or multiple levels, thereby generating such binary that can be directly interpreted by a processor. In other cases, the structure may be hard-coded or hard-wired logic gates that are implemented exclusively or almost exclusively in hardware, such as in a field programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or any other dedicated circuit. Therefore, the term "executable component" is a term for a structure that is well known to those of ordinary skill in the field of computing, whether implemented in software, hardware, or a combination. Any embodiment herein is described with reference to actions performed by one or more processing units of a computing system. If such actions are implemented in software, one or more processors direct the operation of the computing system in response to having executed the computer-executable instructions that constitute the executable component. The computing system may also include a communication channel that allows the computing system to communicate with other computing systems, such as through a network. "Network" is defined as one or more data links that enable electronic data to be transmitted between computing systems and / or modules and / or other electronic devices. When information is transferred or provided to a computing system over a network or another communications connection (e.g., hardwired, wireless, or a combination of hardwired or wireless), the computing system properly views the connection as a transmission medium. Transmission media may include networks and / or data links that may be used to carry desired program code means in the form of computer-executable instructions or data structures and may be accessed by general purpose computing systems or special purpose computing systems or a combination thereof.Although not all computing systems require a user interface, in some embodiments, a computing system includes a user interface system for interacting with a user. The user interface serves as an input or output mechanism for the user, such as via a display.

[0106] Those skilled in the art will appreciate that at least part of the present invention can be practiced in a network computing environment with various types of computing system configurations, including personal computers, desktop computers, laptop computers, message processors, handheld devices, multiprocessor systems, microprocessor-based or programmable consumer electronics, network PCs, minicomputers, mainframe computers, mobile phones, PDAs, pagers, routers, switches, data centers, wearable devices (such as glasses), etc. The present invention can also be practiced in a distributed system environment, where, for example, a local computing system and a remote computing system linked by a network by a hardwired data link, a wireless data link, or a combination of hardwired and wireless data links perform tasks together. In a distributed system environment, program modules can be located in both local and remote memory storage devices.

[0107] Those skilled in the art will also appreciate that at least part of the present invention can be practiced in a cloud computing environment. The cloud computing environment can be distributed, but this is not necessary. When the cloud computing environment is distributed, it can be distributed in multiple countries within an organization and / or have components across multiple organizations. In this specification and the appended claims, "cloud computing" is defined as a model for implementing on-demand network access to a shared pool of configurable computing resources (such as networks, servers, storage devices, applications and services). The definition of "cloud computing" is not limited to any of the other numerous advantages that can be obtained from this model when deployed. The computing system of the accompanying drawings includes various components or functional blocks that can implement various embodiments disclosed herein as explained. Various components or functional blocks can be implemented on a local computing system, or can be implemented on a distributed computing system that includes elements residing in the cloud or implementing various aspects of cloud computing. Various components or functional blocks can be implemented as software, hardware, or a combination of software and hardware. The computing system shown in the figure can include more or fewer components than those shown in the figure, and some of these components can be combined as needed according to the situation.

[0108] Any reference signs in the claims should not be construed as limiting the scope.

[0109] The present invention relates to a method for determining synthesis specifications. A digital representation of a target characteristic and a potential synthesis specification is provided. Then, a model is used to determine the characteristics of a potential target molecule. The model has been parameterized based on an explored data set and an unexplored data set. The explored data set includes the characteristics of a plurality of explored molecules and the molecular characterization parameter values ​​of a plurality of explored molecules. The unexplored data set includes the characterization parameter values ​​of a plurality of unexplored molecules. The determined characteristics of the potential target molecule are compared with the target characteristics. Based on the comparison, i) the potential target molecule is determined as the target molecule, or ii) a new potential target synthesis specification is determined and the determination of the characteristics is repeated. Then, the determined target molecule is provided.

Claims

1. A computer-implemented method for determining the synthesis specifications and / or molecular structure of a target molecule, in particular a target polymer, including properties for a target technical application, wherein: The method includes: Provide target technology application characteristics, providing a digital representation of a potential target molecule, the digital representation indicating or correlating with the characterization parameters, The technical application characteristics of the potential target molecule are determined using a trained determination model, wherein the trained determination model has been parameterized based on an explored data set and an unexplored data set, so that the trained determination model is suitable for determining the technical application characteristics of the molecule based on one or more molecular characterization parameters of the molecule, wherein the explored data set includes a) technical application characteristics of a plurality of explored molecules and b) a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, wherein the molecular characterization parameters are associated with synthesis specifications and / or molecular structures corresponding to each of the plurality of explored molecules, and the unexplored data set includes a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, wherein the molecular characterization parameters are associated with synthesis specifications and / or molecular structures corresponding to a plurality of unexplored molecules, comparing the determined technical application property of the potential target molecule with the target technical application property, and based on the comparison, i) determining the potential target molecule as the target molecule, or ii) providing a new potential target molecule and repeating the determination of the technical application property using the new potential target molecule, The identified target molecule is provided.

2. The method according to claim 1, wherein: Providing the determined target molecule includes generating control signals based on the target molecule and providing the control signals, wherein the control signals are configured to control a production system to produce the target molecule.

3. The method according to any one of claims 1 and 2, wherein: Parameterizing the machine learning based determination model is based on determining a similarity measure between members of the explored data set and members of the unexplored data set.

4. The method according to claim 3, wherein: Parameterizing the machine learning-based determination model includes selecting a subset of the molecular characterization parameters of the explored data set as training molecular characterization parameters based on the similarity measure, wherein the parameterization of the machine learning-based determination model utilizes the training data set, which includes a) technical application characteristics of at least two explored molecules among the multiple explored molecules from the explored data set, and b) molecular characterization parameter values ​​of these training molecular characterization parameters, which are associated with the synthesis specifications and / or molecular structures corresponding to each of the at least two explored molecules.

5. The method according to claim 4, wherein: The subset of the molecular characterization parameters is selected based on optimization of the similarity measure between members of the explored data set and the unexplored data set with respect to the subset of the molecular characterization parameters.

6. The method according to any one of claims 3 to 5, wherein: The similarity measure indicates the distance between members of the explored and unexplored datasets with respect to a subset of the molecular characterization parameters.

7. The method according to any one of claims 5 to 6, wherein: The subset is selected by further optimizing on the determination accuracy of the determination model with respect to the subset of molecular characterization parameters.

8. The method according to any one of claims 5 to 7, wherein: The subset is selected by further optimizing on the fitness of the determined model in the unexplored dataset.

9. A computer-implemented method for generating a determination model based on machine learning, such that the trained determination model is suitable for determining technical application properties of a molecule based on one or more molecular characterization parameters, wherein: The molecular characterization parameters are associated with the synthesis specifications and / or molecular structure of the molecule, wherein the method comprises: - providing an explored data set, the explored data set comprising a) technical application characteristics of a plurality of explored molecules and b) a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, the plurality of molecular characterization parameters being associated with synthesis specifications and / or molecular structures corresponding to each of the plurality of explored molecules, - providing an unexplored data set, the unexplored data set comprising a plurality of molecular characterization parameter values ​​for a plurality of molecular characterization parameters associated with synthesis specifications and / or molecular structures corresponding to each unexplored molecule of a plurality of unexplored molecules, - generating the trained determination model by parameterizing the determination model based on the machine learning based on the explored data set and the unexplored data set, such that the trained determination model is suitable for determining the technical application characteristics of the molecule based on one or more molecular characterization parameters of the molecule, - Providing the generated trained deterministic model.

10. A computer-implemented method for determining one or more molecular characterization parameters for training a determination model based on machine learning, such that the trained determination model is suitable for determining technical application properties of a molecule based on the one or more molecular characterization parameters, wherein: The molecular characterization parameters are associated with synthesis specifications and / or molecular structures corresponding to the molecule, wherein the method comprises: providing an explored data set, the explored data set comprising a) a plurality of technical application characteristics of a plurality of explored molecules and b) a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters associated with synthesis specifications corresponding to each of the plurality of explored molecules, - providing an unexplored data set, the unexplored data set comprising a plurality of molecular characterization parameter values ​​for a plurality of molecular characterization parameters associated with synthesis specifications and / or molecular structures corresponding to each unexplored molecule of the plurality of unexplored molecules, - selecting one or more molecular characterization parameters from the characterization parameters of the explored dataset based on the explored dataset and the unexplored dataset, and - Provide selected molecular characterization parameters.

11. An interface method for providing target synthesis specifications indicating target molecules including target technical application characteristics, wherein: The interface methods include: - receiving target technology application characteristics via an input unit, - interfacing via an interface unit with a processor executing a method according to any one of claims 1 to 8 to provide the target technology application characteristics and receive the determined target synthesis specifications, and - providing the determined target synthesis specification via an output unit.

12. An apparatus for determining target synthesis specifications indicative of a target molecule including target technical application properties, wherein: The equipment includes: - an input interface, which is configured as i) provide target technology application characteristics, and ii) providing a digital representation of the potential target molecule, the digital representation indicating or correlating to the characterization parameters, - a processor, the processor being configured to i) determining the technical application characteristics of the molecule using a trained determination model, wherein the trained determination model has been parameterized based on an explored data set and an unexplored data set, such that the trained determination model is suitable for determining the technical application characteristics of the molecule based on one or more molecular characterization parameters of the molecule, wherein the explored data set includes a) technical application characteristics of a plurality of explored molecules and b) a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, wherein the molecular characterization parameters are associated with synthesis specifications and / or molecular structures corresponding to each of the plurality of explored molecules, and the unexplored data set includes a plurality of molecular characterization parameter values ​​of a plurality of molecular characterization parameters, wherein the molecular characterization parameters are associated with synthesis specifications and / or molecular structures corresponding to a plurality of unexplored molecules, and ii) comparing the determined technical application property of the potential target molecule with the target technical application property, and based on the comparison, i) determining the potential target molecule as the target molecule, or ii) providing a new potential target molecule and repeating the determination of the technical application property using the new potential target molecule, and - an output interface configured to provide the determined target molecules and the corresponding determined target technical application properties.

13. A computer program product for determining target synthesis specifications indicating target molecules including target technical application characteristics, wherein: The computer program product comprises program code means for causing a computer system to execute the method according to any one of claims 1 to 8.

14. A control signal generated using the method of any one of claims 2 to 8.

15. Use of a control signal generated by the method according to any one of claims 2 to 8 for controlling a production system, in particular a laboratory device, to produce the target molecule.

Images