Upcycling of chemical byproducts

Abstract

The invention allows for upcycling of chemical products and relates to a method for generating synthesis instructions for producing one or more target product(s) associated with one or more target product properties, the method comprising: providing the one or more target product properties, providing a digital representation of a chemical structure of the one or more educt(s), determining a digital representation of a chemical structure of one or more product(s) from at least a part of the digital representation of the chemical structure of the one or more educt(s) by determining that the one or more product(s) are formed in one or more chemical reaction(s) of at least a part of the one or more educt(s) associated with a one or more formation score(s), determining if the one or more product(s) are the one or more target product(s) based on the one or more formation score(s) and determining if one or more product properties associated with the one or more product(s) correspond to the one or more target product properties, providing the digital representation of the chemical structure of the one or more product(s) in response to determining that the one or more product properties correspond the one or more target product(s) and optionally providing the digital representation of the one or more educt(s) associated with the one or more target product(s).

Description

[0001] UPCYCLING OF CHEMICAL BYPRODUCTS

[0002] TECHNICAL FIELD

[0003] The invention allows for upcycling of chemical products and relates to a computer-implemented method for generating a synthesis instructions of a target product associated with one or more target product properties, a computer-implemented method for producing a target product with one or more target product properties, a device or apparatus for generating a synthesis instructions of a target product associated with one or more target product properties, a use of a digital representation of a chemical structure of a target product for producing the target product, use of a digital representation of the chemical structure of the one or more educt(s) associated with the target product for producing the target product, and a non-transitory computer-readable storage medium.

[0004] TECHNICAL BACKGROUND

[0005] Chemicals are the basis for materials used and produced in industry. Therefore, sustainable chemicals are a necessity towards a sustainable industry. Currently, developing new chemicals relies on chemical experts and a lot of experimenting. This requires a lot of time and resources. Hence, it is desired to shorten the development time for chemicals and increase the resource-efficiency with respect to experimenting.

[0006] SUMMARY

[0007] In an aspect, this disclosure relates to a computer-implemented method for method for generating synthesis instructions for producing one or more target product(s) associated with one or more target product properties, the method comprising: providing the one or more target product properties, providing a digital representation of a chemical structure of the one or more educt(s), determining a digital representation of a chemical structure of one or more product(s) from at least a part of the digital representation of the chemical structure of the one or more educt(s) by determining that the one or more product(s) are formed in one or more chemical reaction(s) of at least a part of the one or more educt(s) associated with a one or more formation score(s), determining if the one or more product(s) are the one or more target product(s) based on the one or more formation score(s) and by determining if one or more product properties associated with the one or more product(s) correspond to the one or more target product properties, providing the digital representation of the chemical structure of the one or more product(s) in response to determining that the one or more product(s) correspond the one or more target product(s) and optionally receiving the digital representation of the one or more educt(s) associated with the one or more target product(s).

[0008] In another aspect, it relates to a method, in particular a computer-implemented method, for generating synthesis instructions for producing one or more target product(s) associated with one or more target product properties, the method comprising: providing the one or more target product properties, providing a digital representation of a chemical structure of one or more educt(s), determining a digital representation of a chemical structure of one or more product(s) from at least a part of the digital representation of the chemical structure of the one or more educt(s) by determining that the one or more product(s) are formed in one or more chemical reaction(s) of at least a part of the one or more educt(s), associated with one or more formation score(s), determining if the one or more product(s) are the one or more target product(s) according to the one or more formation score(s) and by determining if one or more product properties associated with the one or more product(s) correspond to the one or more target product properties, providing the digital representation of the chemical structure of the one or more product(s) in response to determining that the one or more product properties correspond the one or more target product(s), in particular for monitoring and / or controlling producing and / or processing the one or more product(s), and optionally providing the digital representation of the one or more educt(s) associated with the one or more target product(s).

[0009] In another aspect, it relates to a method, in particular a computer-implemented method, for monitoring and / or controlling producing and / or processing one or more target product(s) associated with one or more target product properties, the method comprising: providing the one or more target product properties, providing a digital representation of a chemical structure of one or more educt(s), determining a digital representation of a chemical structure of one or more product(s) from at least a part of the digital representation of the chemical structure of the one or more educt(s) by determining that the one or more product(s) are formed in one or more chemical reaction(s) of at least a part of the one or more educt(s), associated with one or more formation score(s), determining if the one or more product(s) are the one or more target product(s) according to the one or more formation score(s) and by determining if one or more product properties associated with the one or more product(s) correspond to the one or more target product properties, providing the digital representation of the chemical structure of the one or more product(s) in response to determining that the one or more product properties correspond the one or more target product(s), in particular for monitoring and / or controlling producing and / or processing the one or more product(s), and optionally providing the digital representation of the one or more educt(s) associated with the one or more target product(s), optionally monitoring and / or controlling producing and / or processing the one or more product(s) according to the digital representation of the chemical structure of the one or more product(s).

[0010] In another aspect, it relates to a use of a digital representation of the chemical structure of the one or more educt(s) associated with the target product as obtained by any one the methods as described herein for producing the target product.

[0011] In another aspect, it relates to a use of a digital representation of a chemical structure of a target product as generated according to any one the methods as described herein for producing the target product.

[0012] In another aspect, it relates to an apparatus for generating a synthesis instructions of a target product associated with one or more target product properties, the computing apparatus comprising: a processor configured for performing any one of the methods according to any one of the methods described herein.

[0013] In another aspect, it relates to a device or a system for generating a synthesis instructions of a target product associated with one or more target product properties, the device or system includes a processor configured for performing any one of the methods as described herein.

[0014] In another aspect, it relates to a non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a computer, cause the computer to perform any one of the methods as presented herein.

[0015] In another aspect, it relates to a method for producing one or more target product(s) associated with one or more target product properties, the method comprising: providing the one or more target product properties, providing a digital representation of a chemical structure of the one or more educt(s), determining a digital representation of a chemical structure of one or more product(s) from at least a part of the digital representation of the chemical structure of the one or more educt(s) by determining that the one or more product(s) are formed in one or more chemical reaction(s) of at least a part of the one or more educt(s) associated with a one or more formation score(s), determining if the one or more product(s) are the one or more target product(s) based on the one or more formation score(s) and determining if one or more product properties associated with the one or more product(s) correspond to the one or more target product properties, producing the one or more product(s) formed in one or more chemical reaction(s) of at least a part of the one or more educt(s) in response to determining that the one or more product properties correspond the one or more target product(s).

[0016] EMBODIMENTS

[0017] Any disclosure, embodiments and examples described herein relate to the methods, the systems, apparatuses, chemical products and computer elements lined out above and below.

[0018] Advantageously, the benefits provided by any of the embodiments and examples equally apply to all other embodiments and examples.

[0019] In the following, terminology as used herein and / or the technical field of the present disclosure will be outlined by ways of definitions and / or examples. Where examples are given, it is to be understood that the present disclosure is not limited to said examples.

[0020] Developing new materials constitutes the challenge of observing non-regular trends within the relation between chemical structure and product properties. Hence, it requires highly experienced experts to design a plurality of experiments that are usually resource-intensive due to the usage of different chemicals while producing potentially harmful waste requiring resource-intensive treatment. Further, chemical products are a result of a plurality of reaction steps each being associated with input products that are allowed to react to an intermediate product until the chemical product is formed. Because of this, educts are typically involved in a plurality of reactions and reaction pathways towards chemical products are designed to minimize waste resulting from unused byproducts. This makes chemical production very complex resulting in long time intervals needed for adapting to current developments, for example if a need for new chemical products arises.

[0021] The herein presented methods, devices, systems, uses, apparatuses and computer-readable storage media enable an accurate, fast and reliable development of new chemicals with target properties based on predefined educts. These educts may be defined by the user as the digital representations of the chemical structure of the one or more educt(s) may be provided by the user. Restricting the selection of possible educts to a subset of all possible educts by receiving a digital representation of a chemical structure of one or more educts enables to adapt the search for new chemical products among available educts. Such educts may be byproducts of other reactions, may have a lower environmental impact or may be part of another value chain with additional amounts of the educt. Hence, providing a predefined number of educts may provide the advantage of saving resources by upcycling educts into valuable chemical products. Furthermore, determining a digital representation of a chemical structure of one or more product(s) from the digital representation of a chemical structure of at least a part of the one or more educts allows to find robust synthesis routes towards a chemical product enabling scale-up of in silico determined synthesis routes, especially by using template-based approaches. This is even further enhanced by determining one or more formation score(s) associated with forming the one or more product(s) in one or more chemical reaction(s) from the one or more educt(s). These scores constitute an accurate measure for evaluating the robustness of a synthesis associated with the one or more product(s). Determining one or more product properties allows for evaluating the performance parameters associated with the one or more product(s). These application properties can be of different kinds and thus, allow an application fine-tuned chemical development. Furthermore, the amount of resources needed for experimentation is reduced and less hazardous waste needs to be processed to arrive at meaningful synthesis routes. By determining if the one or more product(s) are the one or more target product(s), a direct comparison between potential products is enabled. Thereby, products and respective reaction pathways to synthesize these products can be compared in terms of robustness, reproducibility and target performance parameters. Ultimately, this allows for an accurate and fast adaptation according to the need of new chemical products. Hence, this invention contributes to saving resources while accelerating the development of new materials with a variety of performance parameters.

[0022] These and other objects, which become apparent upon reading the following description, are solved by the subject matters of the independent claims. The dependent claims refer to embodiments of the invention.

[0023] In an embodiment, atom matching score may refer to a matching score indicating the corresponding atoms of the one or more educt(s) and the one or more product(s). The matching score may comprise a value, eg a percentage. Determining the atom matching score may refer to performing atom mapping. Atom matching score may comprise an atom mapping score.

[0024] In an embodiment, byproduct may be a product obtained from a chemical reaction. The byproduct may be obtained from a chemical reaction to a desired product. A conversion rate associated with obtaining the byproduct may be smaller than a conversion rate associated with obtaining the desired product. In an embodiment, the conversion rate associated with obtaining the byproduct may be smaller than 50%.

[0025] In an embodiment, chemical product may refer to a material associated with one or more chemical structures. In particular, chemical structure may refer to a molecular structure. Educts and / or products may be a chemical product.

[0026] In an embodiment, the chemical product may be obtained from at least one chemical reaction. The chemical product may include natural chemical products. Natural chemical products may include any chemical product that is produced by nature without human interaction or intervention, i.e. any unprocessed chemical substance that is found in nature, such as chemicals from plants, micro-organisms, animals, the earth and the sea or any chemical substance that is found in nature and extracted using a process that does not change its chemical composition. Natural chemical products may include biologicals like enzymes as well naturally occurring inorganic or organic chemical products. Natural chemical products may be isolated and purified prior to their use or they can be used in uninsulated and / or unpurified form. Chemical products may be synthetic chemical products. Synthetic chemical products may include chemical products produced with human interaction or intervention. Synthetic chemical products may be produced with the same chemical reactions occurring in nature or with different chemical reactions. Chemical products may be any inorganic or organic chemical product obtained by reacting inorganic and / or organic chemical reactants. The inorganic and organic chemical reactants may be natural chemical products or may be synthetic chemical products. Chemical reactions may include any chemical reaction commonly known in the state of the art in which the reactants are converted to one or more different chemical products. Chemical reactions may involve the use of catalysts, enzymes, bacteria, etc. to achieve the chemical reaction between the reactants. The chemical product may include a raw material. The chemical product may include a chemical material produced by reacting at least two raw materials. The chemical product may include a component. The chemical product may include a component assembly. The chemical product may include an end product.

[0027] Preferably, the chemical product may be an organic chemical product. The chemical product may be associated with a molecular mass. Molecular mass may specify the mass associated with the atoms comprised in one molecule of the one or more chemical products. The molecular mass may be expressed in the unit u or Da. 1 u and / or 1 Da may refer to 1 / 12 of the mass of a 12-C atom. The molecular mass of the one or more chemical products may be less than 1000 u and / or less than 1000 Da.

[0028] In an embodiment, a chemical reaction may change the chemical structure of one or more educt(s) to the chemical structure of one or more products, preferably one or more products. A chemical reaction may further change the relation between the two or more atoms of the one or more educt(s) to the arrangement of atoms associated with the one or more products, preferably one or more products. A chemical reaction may change one or more bonds between one or more atoms associated with the one or more educt(s) to result in the one or more products, preferably one or more products. Hence, a chemical reaction may include breaking one or more bonds in the chemical structure of the one or more educt(s) towards the one or more products, preferably one or more products and / or establishing one or more bonds in the chemical structure of the one or more educt(s) towards the one or more products, preferably one or more products.

[0029] In an embodiment, data-driven model may refer to a model suitable for describing one or more non-linear relations between input data and output data. Input data may refer to data to be provided to the data-driven model and / or to data being received by the data-driven model. Output data may be data to be received from the data- driven model and / or to be provided by the data-driven model. Hence, the data-driven model may determine the output data based on transforming the input data via one or more non-linear relations.

[0030] Digital representation of a chemical structure may be a machine-readable and / or a machine-interpretable representation of the chemical structure. The digital representation of a chemical structure may comprise a data structure indicative of the chemical structure. The chemical structure may specify one or more atoms associated with a chemical product such as one or more educt(s) and / or one or more products Specifying the one or more atoms may refer to specifying one or more elements associated with the one or more atoms. Further, the digital representation of a chemical structure may be indicative of a relation between the one or more atoms, preferably an interaction between the one or more atoms, most preferably one or more bonds between the one or more atoms. Additionally or alternatively, digital representation of the chemical structure may be indicative of an arrangement of the one or more atoms, in particular in relation to a predefined point. Examples for digital representation of a chemical structure can include SMILES, SMARTS, mol files, pdb files, files indicative of the two or more atoms associated with the chemical structure and their arrangement.

[0031] In an embodiment, educt may refer to a chemical product suitable for synthesizing one or more products.

[0032] In an embodiment, formation score may be indicative of a certainty that the one or more product(s) may be formed in one or more chemical reaction(s) from at least a part of the one or more educt(s). Formation score may be indicative of an amount of the one or more product(s) formed by a chemical reaction of at least a part of the one or more educt(s).

[0033] In an embodiment, forward modelling template may refer to a template describing the connection of at least one educt with itself, e.g. to form a ring, or at least two educts, e.g. to form a chain-like structure or a structure comprising a ring. Forward modelling template may specify the product resulting from the connection of one or more educts and / or the leaving atoms during the formation of the product.

[0034] In an embodiment, predefined range may be indicative of a range of numerical values associated with the target product. Preferably, the predefined range may be indicative of a range of numerical values associated with the formation of the target product. Predefined range may include a predefined score range and / or a predefined structure range.

[0035] In an embodiment, product may refer to a chemical product suitable for being synthesized based on one or more educt(s). In an embodiment, product property may refer to a property associated with a chemical product. Product property may refer to an environmental attribute, to a physical property, to a chemical property or a combination thereof. In an embodiment, environmental attribute may comprise at least one of emission data of the chemical product, recyclate content of the chemical product, bio-based content of the chemical product, renewable content of the chemical product, chemical product declaration data, chemical product safety data or a combination thereof.

[0036] In an embodiment, emission data may comprise any data related to environmental footprint. The environmental footprint may refer to an entity and its associated environmental footprint. The environmental footprint may be entity specific. For instance, the environmental footprint may relate to a chemical product, a company, a process such as a manufacturing process, a raw material or basic substance, a chemical product or material, a component, a component assembly, an end product, combinations thereof or additional entity-specific relations. Emission data may include data relating to carbon footprint of a chemical product. Emission data may include data relating to greenhouse gas emissions e.g. released in production of the chemical product. Emission data may include data related to greenhouse gas emissions. Greenhouse gas emissions may include emissions such as carbon dioxide (CO2) emission, methane (CH4) emission, nitrous oxide (N2O) emission, hydrofluorocarbons (HFCs) emission, perfluorocarbons (PFCs) emission, sulphurhexafluoride (SFe) emission, nitrogen trifluoride (NF3) emission, combinations thereof and additional emissions. Emission data may include data related to greenhouse gas emissions of an entities or companies own operations (production, power plants and waste incineration). Scope 2 comprise emissions from energy production which is sourced externally. Scope 3 comprise all other emissions along the value chain. Specifically, this includes the greenhouse gas emissions of raw materials obtained from suppliers. Product Carbon Footprint (PCF) sum up greenhouse gas emissions and removals from the consecutive and interlinked process steps related to a particular product. Cradle-to-gate PCF sum up greenhouse gas emissions based on selected process steps: from the extraction of resources up to the factory gate where the product leaves the company. Such PCFs are called partial PCFs. In order to achieve such summation, each company providing any products must be able to provide the scope 1 and scope 2 contributions to the PCF for each of its products as accurately as possible, and obtain reliable and consistent data for the PCFs of purchased energy (scope 2) and their raw materials (scope 3).

[0037] Chemical property may be a property that can be established by changing the structure of the at least one chemical product. Examples for chemical properties may be acidity, oxidation state or reactivity. Physical property may be one of the following: mechanical properties, electrical properties, optical properties, thermal properties or the like. For example, physical property may comprise one or more of the following density, scratch resistance, electrical conductivity, color, absorption, heat capacity or the like.

[0038] In an embodiment, synthesis instructions may specify the synthesis associated with the formation of the one or more target product(s), in particular the synthesis for generating the one or more target product(s). Preferably, synthesis instructions may indicate the one or more educts, the one or more products and / or the reaction conditions. Synthesis instructions may indicate at least one of qualitative educt composition, quantitative educt composition, qualitative product composition, quantitative product composition, synthesis conditions, purification steps, qualitative product composition after purification, quantitative product composition after purification, or a combination thereof. The synthesis instructions may be indicative of a chemical structure of the one or more target product(s) and optionally a chemical structure of the one or more educts the one or more target product(s) can be synthesized from. Synthesis instructions may be computer-readable and / or computer-executable instructions suitable for monitoring and / or controlling at least a part of a chemical plant. Synthesis instructions may be instruction suitable for producing a chemical product, such as one or more product(s). Synthesis instructions may be computer-readable and / or computer-executable instructions suitable for producing a chemical product, such as one or more product(s). Synthesis instructions may be associated with a digital representation of one or more target product(s).

[0039] In an embodiment, determining the one or more product properties from the digital representation of the chemical structure of the one or more product(s) may comprise providing the digital representation of the chemical structure of the one or more educt(s) to a property data-driven model and / or receiving the one or more product properties from the property data-driven model, preferably in response to having provided the property data-driven model with the digital representation of the chemical structure of the chemical structure of the second product, wherein the property data-driven model may be trained and / or parametrized based on a plurality of digital representations of the chemical structure of products and corresponding product properties. The property data-driven model may be configured for providing the one or more product properties in response to being provided with a digital representation of the chemical structure of the one or more product(s). By doing so, target products can be determined fast and reliably while saving resources for carrying out experiments.

[0040] In an embodiment, determining the one or more product properties from the digital representation of the chemical structure of the chemical structure of the one or more product(s) may comprise providing the digital representation of the chemical structure of the one or more educt(s) to a property database for receiving the one or more product properties, wherein the property database may be configured for receiving the digital representations of the chemical structure products and providing product properties in response to receiving the digital representation of the chemical structure of the products. The property database may comprise a plurality of digital representations of the chemical structure of products and corresponding product properties. By doing so, target products can be determined fast and reliably while saving resources for carrying out experiments as historical data from already conducted experiments are used.

[0041] In an embodiment, determining a digital representation of a chemical structure of the one or more product(s) from the digital representation of the chemical structure of the one or more educt(s) may comprise providing the digital representation of the chemical structure of the one or more educt(s) to a forward modelling data-driven model for generating the digital representation of the chemical structure of the one or more product(s). The forward modelling data-driven model may be parametrized and / or trained to receive digital representations of the chemical structure of one or more educt(s) and generate the digital representation of the chemical structure of products in response to receiving the digital representation of the chemical structure of the one or more educt(s). The forward modelling data-driven model may be trained and / or parametrized based on a forward modelling training data set. The forward modelling training data set may comprise a plurality of digital representations of a chemical structure of educts and a plurality of corresponding digital representations of a chemical structure of products. The forward modelling data-driven model may be parametrized and / or trained to receive the one or more of the digital representations of a chemical structure of one or more educts at an input layer of the forward modelling data- driven model, and / or to provide, in particular in response to having received the digital representation of a chemical structure of the one or more educt(s), the digital representation of a chemical structure of the one or more product(s). By doing so, products can be obtained while resources for carrying out experiments can be saved.

[0042] In an embodiment, the digital representation of the chemical structure of the one or more educt(s) may be provided and / or received via a user interface. Further, the digital representation of the chemical structure of the one or more educt(s) may be provided by a user.

[0043] Additionally or alternatively, determining a digital representation of a chemical structure of one or more product(s) from the digital representation of the chemical structure of the one or more educt(s) may comprise providing the digital representation of the chemical structure of the one or more educt(s) to a compound database for receiving the digital representation of the chemical structure of the one or more product(s), wherein the compound database is configured for receiving digital representations of the chemical structure of one or more educt(s) and providing the digital representation of the chemical structure of the one or more product(s) in response to receiving the digital representation of the chemical structure of the one or more educt(s). The compound database may comprise at least one of a plurality of digital representations of a chemical structure of educts, a plurality of digital representations of a chemical structure of products, a plurality of product properties with respect to at least one of the digital representations of the chemical structure of products or a combination thereof. Furthermore, the compound database may relate at least one of the plurality of digital representations of the chemical structure of educts and at least one of the plurality of digital representation of a chemical structure of products. Hence, the compound database may be suitable for providing the digital representation of a chemical structure of the one or more product(s) in response to having received the digital representation of the chemical structure of the one or more educt(s). In an embodiment, determining a digital representation of a chemical structure of the one or more product(s) from the digital representation of the chemical structure of the one or more educt(s) may comprise applying a forward modelling template to digital representation of the chemical structure of the one or more educt(s). Applying the forward modelling template to at least a part of the digital representation of a chemical structure of the one or more educt(s) may result in the digital representation of a chemical structure of the one or more product(s). The one or more forward modelling templates may be received, e.g. from a database. One or more forward modelling templates may be received and / or provided based on a selection. The selection may be received e.g. via a user interface and / or from a user. The forward modelling template may indicate a change in the chemical structure of at least a part of the one or more educt(s) to result in the chemical structure of the one or more product(s). Hence, applying the forward modelling template may change the chemical structure of at least the part of the one or more educt(s) to the chemical structure of the one or more product(s).

[0044] In an embodiment, the target product may be produced from one or more byproducts of a chemical reaction and / or wherein at least one of the one or more educt(s) is a byproduct of the chemical reaction. A byproduct may be a product obtained in one or more chemical reaction(s) to obtain the one or more target product(s). The byproduct may be obtained in the one or more chemical reaction(s) in addition to the one or more target product(s). A quantity associated with the byproduct may be lower than a quantity associated with the target product. An inventory may comprise at least a part of the one or more educt(s). In an embodiment, the one or more educt(s) may be selected from a plurality of educts by determining the one or more educts available at a location of a chemical production facility for producing the one or more target product(s). By doing so, target products with target properties can be obtained from already available educts. Hence, chemical products can be produced in line with the demand.

[0045] In an embodiment, determining a digital representation of a chemical structure of the one or more product(s) from at least the part of the digital representation of the chemical structure of the one or more educt(s) may comprise providing a request for providing the digital representation of the chemical structure of the one or more product(s) to a compound database and providing the digital representation of the chemical structure of the one or more product(s) by the compound database. The request may be indicative of the one or more educt(s). The request may comprise a structured query for providing the digital representation of the chemical structure of the one or more product(s). The compound database may be a structured compound database. The compound database may be configured for providing digital representations of the chemical structure of one or more product(s) in response to receiving requests for providing the digital representations of the chemical structure of one or more product(s). Additionally or alternatively, the request may comprise an embedded digital representation of the chemical structure of the one or more educt(s). The compound database may be configured for determining a distance between the embedded request and embedded digital representation of the chemical structure of the one or more product(s) and selecting the digital representation of the chemical structure of the one or more product(s) based on determining that the distance associated with the digital representation of the chemical structure of the one or more product(s) is within a predefined structure range. This is advantageous since already generated knowledge can be used and computational resources are saved. Hence, this embodiment contributes to an efficient use of resources such as electricity and computational cost. Furthermore, the computational resources may be used for other operations, e.g. determining a new products based on another combination of educts and / or a new products based on another reaction template.

[0046] In an embodiment, the predefined score range and / or the predefined structure range may be provided via a user interface. This allows for tailoring of the target product towards requirements specified by the use case.

[0047] In an embodiment, target product may be synthesized and / or produced from one or more byproducts of a synthesis associated with a second synthesis instructions and / or wherein at least one of the one or more educt(s) may be a byproduct of the one or more chemical reaction(s). This is beneficial since this allows to use byproducts of other synthesis. Consequently, resources are used more efficiently and material is not wasted. More than that, it allows for exploiting the potential of byproducts. Followingly, the method for generating a synthesis instructions of a target product associated with one or more target product properties may refer to a method for generating a synthesis instructions of a target product associated with one or more target product properties based on one or more byproducts.

[0048] In an embodiment, the digital representation of the chemical structure of the one or more educt(s) and / or one or more product(s) may comprise one or more machine-readable indication(s) and / or a machine-readable indication of the chemical structure associated with the one or more educt(s) and / or the one or more products. The machine- readable indication of the chemical structure associated with the one or more educt(s) and / or the one or more product(s) may comprise one or more chemical fingerprint(s). Thus, the digital representation of a chemical structure of the one or more educt(s) and / or the one or more products may comprise a chemical fingerprint.

[0049] In an embodiment, the digital representation of the chemical structure of the one or more educt(s) and / or one or more product(s) may be a numerical representation of the chemical structure of the one or more educt(s) and / or one or more product(s), in particular a vector representation of the chemical structure of the one or more educt(s) and / or one or more product(s). Preferably, the numerical representation of the chemical structure of the one or more educt(s) and / or one or more product(s) may be obtained by mapping an indication on the chemical structure of the one or more educt(s) and / or one or more product(s) to the digital representation. The indication may be suitable for identifying a chemical compound, ie the one or more educt(s) and / or one or more product(s). The data structure and / or data format associated with the numerical representation may be equal to a data structure and / or data format associated with a plurality of different educts and / or products, in particular a plurality of educts and / or products with different functional groups and / or different number of atoms. In an embodiment, the method may further comprise generating one or more chemical fingerprints associated with the one or more educt(s) from the digital representation of the chemical structure of the one or more educt(s). The chemical fingerprint may be a multidimensional representation of the digital representation of the chemical structure of the one or more educts. The chemical fingerprint may be indicative of a presence of one or more subunits of the one or more product(s) and / or educt(s). Further, the chemical fingerprint may be indicative of a relation between two or more subunits of the one or more product(s) and / or educt(s). The subunits of the one or more product(s) and / or educt(s) may include for example a fragment, a functional group, an atom or the like. Providing the digital representation of the chemical structure of the one or more educt(s) may comprise providing the chemical fingerprint of the one or more educt(s). Determining a digital representation of a chemical structure of the one or more product(s) from at least a part of the digital representation of the chemical structure of the one or more educt(s) may comprise determining a digital representation of a chemical structure of the one or more product(s) from at least a part of the chemical fingerprint of the one or more educt(s). For example, the chemical fingerprint may be a molecular descriptor. The fingerprint may indicate the atoms and / or the chemical surrounding of at least a part of the atoms within a predefined distance. In an embodiment, the chemical fingerprint of an educt may be obtained by encoding the digital representation of a chemical structure of the one or more educt(s) and / or one or more product(s). Encoding in this context may refer to transforming the digital representation of a chemical structure of the one or more educt(s) and / or one or more product(s) into one or more chemical fingerprints associated with the one or more educt(s) and / or one or more product(s). The digital representation of a chemical structure of the one or more educt(s) and / or one or more product(s) may indicate the chemical structure of the educt and the chemical fingerprint may indicate the atoms and / or the chemical surrounding of at least a part of the atoms within a predefined distance. Followingly, the fingerprint may be obtained based on the digital representation of a chemical structure of the one or more educt(s) and / or one or more product(s) by determining the atoms of the chemical structure of the one or more educt(s) and determining the chemical surrounding of the atoms in a predefined distance. The chemical surrounding may indicate the chemical nature of the atoms surrounding the respective atom. The chemical surrounding may indicate the reaction behavior may indicate the atoms and / or the surrounding of at least a part of the atoms within a predefined distance of the respective atom. By doing so, one takes the chemical nature of the atomic interactions into account defining the reaction behavior of the atoms in a corresponding chemical product.

[0050] In an embodiment, the method may further include folding the one or more chemical fingerprints by reducing the dimensionality of the chemical fingerprints. Reducing the dimensionality of the chemical fingerprint may refer to reducing a size of the chemical fingerprint and / or folding the chemical fingerprint. An example for reducing the dimensionality of the chemical fingerprint may include hashing the chemical fingerprint. Folding the chemical fingerprint may result in a data structure with a higher information density. Thus, the computational resources for determining the products based on the educts may be used more efficiently. In an embodiment, determining one or more formation score(s) may comprise determining one or more formation score(s) by determining one or more atom matching score(s) of the atoms of the one or more educt(s) with the atoms of the one or more product(s), determining a reaction likelihood associated with the formation of the one or more product(s) based on the one or more educt(s) by providing the one or more of the digital representation of the chemical structure of the one or more educt(s) sets to a scoring data-driven model. Determining one or more formation score(s) may be further based on retrosynthetic educts. Retrosynthetic educts may be obtained by determining the educts of a retrosynthetic reaction of the one or more product(s). The retrosynthetic educts may be obtained by applying a retrosynthetic template to the digital representation of a chemical structure of the one or more product(s).

[0051] In an embodiment, the scoring data-driven model may be parametrized and / or trained to determine the polarization and / or the reactivity associated with at least an atom of a molecule. The scoring data-driven model may be parametrized and / or trained based on a scoring training data set. The scoring training data set may comprise digital representations of the chemical structure of the one or more educt(s) and / or digital representations of the chemical structure of the one or more product(s) and one or more formation score(s) associated with the formation of the one or more products based on the one or more educts. The scoring data- driven model may be parametrized and / or trained to receive digital representation of the chemical structure of the one or more educt(s) and the digital representation of the chemical structure of the one or more product(s), e.g. at an input layer, and / or to provide one or more formation score(s) associated with the formation of the one or more products.

[0052] In an embodiment, determining if one or more product properties associated with the one or more product(s) may correspond to the one or more target product properties may comprise determining one or more product properties associated with the one or more product(s) from the digital representation of the chemical structure of the one or more product(s) and optionally the digital representation of the one or more educt(s) and comparing the one or more product properties and the one or more target product properties. The digital representation of the chemical structure of two or more product(s) associated with two or more formation score(s) may be determined, and wherein two or more product properties are determined associated with the two or more product(s) from the digital representation of the chemical structure of the two or more product(s) and optionally the digital representation of the one or more educt(s). The two or more product properties may be compared with the one or more target product properties. The digital representation of the chemical structure of at least one of the two or more product(s) may be provided in response to determining that the at least one product corresponds to the one or more target product(s). Determining the one or more product properties may comprise providing the digital representation of the chemical structure of the one or more product(s) to a property data-driven model and / or receiving the one or more product properties from the property data-driven model, preferably in response to having provided the digital representation of the chemical structure of the one or more product(s). The property data-driven model may be configured to provide product properties in response to being provided with digital representation of the chemical structure of products. The property data-driven model may be trained and / or parametrized based on a property training data set. The property training data set may comprise a plurality of digital representation of the chemical structure of the one or more product(s) and a plurality of corresponding product properties. By doing so, resources and time otherwise used for determining the product properties e.g. within a laboratory can be saved.

[0053] In an embodiment, wherein determining the one or more product properties may comprise evaluating the digital representation of the chemical structure of the one or more product(s). The digital representation may be indicative of one or more product properties, preferably may relate to one or more product properties. Hence, evaluating the digital representation may comprise relating the digital representation to the one or more product properties. In an embodiment, at least a part of the one or more product properties may be obtained by evaluating the digital representation of the chemical structure of the one or more product(s) and providing the digital representation of the chemical structure of the one or more product(s) to the property data-driven model. Properties obtained by evaluating the digital representation may be include properties related to a presence of one or more subunits of the one or more educt(s) and / or product(s) and / or a relation between two or more subunits of the one or more educt(s) and / or product(s). Hence, examples may include a molecular weight, a number of a type of a subunit of the one or more educt(s) and / or product(s).

[0054] In an embodiment, determining a digital representation of a chemical structure of one or more product(s) from the digital representation of the chemical structure of the one or more educt(s) may comprise providing the digital representation of the chemical structure of the one or more educt(s) to a compound database for receiving the digital representation of the chemical structure of the one or more product(s), wherein the compound database is configured for receiving digital representations of the chemical structure of one or more educt(s) and providing the digital representation of the chemical structure of the products in response to receiving the digital representation of the chemical structure of the one or more educt(s). Providing the digital representation of the chemical structure of the one or more educt(s) may comprise providing a query indicative of the digital representation of the chemical structure of the one or more educt(s). Additionally or alternatively, providing the digital representation of the chemical structure of the one or more educt(s) may comprise embedding digital representation of the chemical structure of the one or more educt(s) to result in an embedded digital representation of the chemical structure of the one or more educt(s) and determining the distance between the embedded digital representation of the chemical structure of the one or more educt(s) and the embedded digital representations of the chemical structure of the products. The embedded digital representations of the chemical structure may be obtained by embedding the digital representations of the chemical structure of the products. Embedding the digital representation of the chemical structure may comprise providing the digital representations of the chemical structure to an encoder configured for reducing the dimensionality of the digital representations of the chemical structure. Reducing the dimensionality of the digital representations of the chemical structure may comprise transforming the digital representations of the chemical structure into tensors associated with the digital representations of the chemical structure. Determining the distance between the embedded digital representation of the chemical structure of the one or more educt(s) and the embedded digital representations of the chemical structure of the products may comprise determining an euclidean distance and / or a cosine similarity between the embedded digital representation of the chemical structure of the one or more educt(s) and the embedded digital representations of the chemical structure of the products. The digital representations of the chemical structure of the one or more product(s) may be provided in response to determining that the embedded digital representation of the chemical structure of the one or more product(s) may be within a predefined distance of the embedded digital representation of the chemical structure of the one or more educt(s) and / or in response to determining that the embedded digital representation of the chemical structure of the one or more product(s) may be closest in distance to the embedded digital representation of the chemical structure of the one or more educt(s) in comparison to the other digital representations of the chemical structure of the products.

[0055] In an embodiment, the methods may further comprise generating one or more chemical fingerprints of the one or more educt(s) from the digital representation of the chemical structure of the one or more educt(s), wherein the chemical fingerprint may be a dimensionality-reduced representation of the digital representation of the chemical structure of the one or more educt(s) and wherein providing digital representation of the chemical structure of the one or more educt(s) comprises providing the chemical fingerprint of the one or more educt(s) and wherein determining a digital representation of a chemical structure of one or more product(s) from at least a part of the digital representation of the chemical structure of the one or more educt(s) comprises determining a digital representation of a chemical structure of the one or more product(s) from at least a part of the chemical fingerprint of the one or more educt(s). The chemical fingerprint may be a bit string. The chemical fingerprint may be indicative of two or more atoms of one or more chemical elements the one or more educt(s) comprises, preferably contains, and one or more atoms of one or more chemical elements within a predefined distance of the two or more atoms. By doing so, less computational resources are deployed for obtaining the target product with tailored target properties.

[0056] In an embodiment, determining the one or more product properties from the digital representation of the chemical structure of the chemical structure of the one or more product(s) may comprise providing the digital representation of the chemical structure of the one or more educt(s) to a property database for receiving the one or more product properties. The property database may be configured for receiving the digital representations of the chemical structure of the one or more product(s) and providing the one or more product properties in response to receiving the digital representation of the chemical structure of the one or more product(s). Providing the digital representation of the chemical structure of the one or more product(s) may comprise providing a query indicative of the digital representation of the chemical structure of the one or more product(s). Additionally or alternatively, providing the digital representation of the chemical structure of the one or more product(s) may comprise embedding digital representation of the chemical structure of the one or more product(s) to result in an embedded digital representation of the chemical structure of the one or more product(s) and determining the distance between the embedded digital representation of the chemical structure of the one or more product(s) and the embedded digital representation of the chemical structure of the plurality of products associated with product properties. The embedded digital representations of the chemical structure of the one or more product(s) may be obtained by embedding the digital representation of the chemical structure of the one or more product(s). Embedding the digital representation of the chemical structure of the one or more product(s) may comprise providing the digital representation of the chemical structure of the one or more product(s) to an encoder configured for reducing the dimensionality of the digital representation of the chemical structure of the one or more product(s). Reducing the dimensionality of the digital representation of the chemical structure of the one or more product(s) may comprise mapping the digital representation of the chemical structure of the one or more product(s) into tensors associated with the digital representation of the chemical structure of the one or more product(s). Determining the distance between the embedded digital representation of the chemical structure of the one or more product(s) and the embedded digital representations of the chemical structure of the plurality of products may comprise determining an euclidean distance and / or a cosine similarity between the embedded digital representation of the chemical structure of the one or more product(s) and the embedded digital representations of the chemical structure of the plurality of products. The one or more product properties may be provided in response to determining that the embedded digital representation of the chemical structure of the one or more product(s) may be within a predefined distance of one of the embedded digital representation of the chemical structure of the plurality of products and / or in response to determining that the embedded digital representation of the chemical structure of the one or more product(s) may be closest in distance to the one embedded digital representation of the chemical structure of the plurality of products in comparison to the other digital representations of the chemical structure of the products.

[0057] In an embodiment, determining one or more product properties associated with the one or more product(s) from the digital representation of the chemical structure of the one or more product(s) may comprise providing one or more product properties associated with a chemical compound other than the one or more product(s). Preferably, the chemical compound may be associated with a similarity score indicative of a matching of one or more subunits of the chemical compound and the one or more product(s). The one or more product properties associated with the chemical compound may be obtained by providing the embedded digital representation of the chemical structure of the one or more product(s). In particular, at least a part of the product properties of the one or more product(s) may be product properties related to at least 90 % of the chemical structure. Examples may be biological activity, lipophilicity or the like. In an embodiment, the providing the digital representation of the chemical structure of the one or more product(s) and / or educt(s) may comprise providing a digital identifier associated with the one or more product(s) and / or educt(s). The property database and / or the compound database may be configured for providing the one or more product properties and / or the one or more digital representations of the chemical structure of the one or more educt(s) based on receiving the digital identifier. The digital identifier may be associated with and / or may be linked to and / or may be retrieved based on the digital representation of the chemical structure of the one or more product(s) and / or educt(s).

[0058] In an embodiment, determining if the one or more product(s) are the one or more target product(s) based on the one or more formation score(s) may comprise determining the one or more formation score(s) associated with forming the one or more product(s) in the one or more chemical reaction(s), and providing a predefined score range associated with determining that the one or more product(s) may be formed in the one or more chemical reaction(s) from the one or more educt(s) and determining if at least one of the one or more formation score(s) may be within the predefined score range.

[0059] In an embodiment, a plurality of predefined score ranges may be provided. The plurality of predefined score ranges are associated with a plurality of different chemical reactions and selecting the predefined score range associated with the chemical reaction in which the one or more product(s) are formed. It may be determined if at least one of the one or more formation score(s) may be within the selected predefined score range.

[0060] In an embodiment, the digital representation of a chemical structure of the one or more educt(s) and / or the one or more product(s) may include at least one of one or more SMILES strings, one or more graph data structures associated with the chemical structure of the one or more educt(s), wherein the graph data structure comprises two or more nodes indicative of two or more atoms comprised in the one or more educt(s) and one or more edges indicative of one or more bonds between the two or more atoms, one or more SMARTS strings or a combination thereof.

[0061] In an embodiment, determining if the one or more product(s) are the one or more target product(s) may depend on the one or more formation score(s) and on determining if one or more product properties associated with the one or more product(s) correspond to the one or more target product properties.

[0062] BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0063] In the following, the present disclosure is further described with reference to the enclosed figures. The same reference numbers in the drawings and this disclosure are intended to refer to the same or like elements, components, and / or parts. FIG. 1 illustrates an embodiment of producing products by one or more chemical production facilities 102.

[0064] FIG. 2 illustrates an embodiment of a chemical production network.

[0065] FIG. 3 illustrates an embodiment of a method for generating synthesis instructions for producing a target product associated with one or more target product properties.

[0066] FIG. 4 illustrates combinations of a plurality of educts 402a - 402b to form a plurality of products 404a - 404b.

[0067] FIG. 5 illustrates the principle of templates, retrosynthetic modelling and forward synthetic modelling.

[0068] FIG. 6 illustrates representations of a chemical structure associated with a molecule.

[0069] FIG. 7 illustrates an embodiment of determining a formation score.

[0070] FIG. 8 illustrates an embodiment of an apparatus for generating synthesis instructions of a target product.

[0071] FIG. 9 illustrates an embodiment of input and output data associated with the forward modelling data-driven model, the scoring data-driven model and / or the property data-driven model.

[0072] DETAILED DESCRIPTION

[0073] The following embodiments are mere examples for implementing the method, the system or application device disclosed herein and shall not be considered limiting.

[0074] FIG. 1 illustrates an embodiment of producing products by one or more chemical production facilities 102.

[0075] Products may be produced by the one or more chemical production facilities 102. The products produced by the one or more chemical production facilities 102 may be associated with 1 16. The product may be processed by one or more product processing facilities 1 14. For processing the product by the one or more product processing facilities 114 the 1 16 may be desired to correspond to one or more target product properties 1 12. The one or more target product properties 112 may be necessary for producing end products with a desired quality. End products with a lower quality may become waste immediately as such end products may not be suited for the intended application or at least faster than the end products with the desired quality due to faster degradation. Hence, ensuring sufficient quality of the products produced by the one or more chemical production facilities 102 is important to reduce resource waste. Furthermore, by producing products from byproducts of other chemical reactions in a chemical production network with hundreds of different chemical reaction pathways, byproducts can be processed towards en products. Thereby, the waste of chemical production is further lowered while the production of the products by the one or more chemical production facilities 102 may be tailored to the desired properties of the products. The challenge in chemical production is to select the educts for producing the desired products as the properties of educts may be independent of the properties of the product obtained by a chemical reaction of the educts to yield the product. The reason is that chemical reactions change the chemical structure of the educts to yield the products. The properties of a product highly depend on the chemical structure of the product. Even a change in an orientation of one subgroup of a molecule with hundreds of atoms may typically result in a change of the properties of the product. Hence, no direct relation between the educts at hand and the target products can be established due to the nature of chemical reactions. Nevertheless, chemical production facilities 102 are controlled according to the educts for a chemical reaction. Therefore, this disclosure provides methods, uses and apparatuses for generating synthesis instructions for producing one or more target product(s) associated with one or more target product properties based on available educts. Followingly, the disclosure enables a targeted steering of chemical production to reduce waste and energy.

[0076] To do so, one or more target product properties 112 may be provided by one or more product processing facilities 114 to an operating system 1 18 of the one or more chemical production facilities 102. The operating system 118 may comprise an inventory of available educts 104 and an apparatus for generating synthesis instructions for producing one or more target product(s) associated with one or more target product properties. The apparatus may be configured for being provided with digital representation of the one or more educt(s) 108 and to provide the digital representation of the chemical structure of the one or more target product(s). The apparatus may be further described in the context of FIG. 8.

[0077] The educt provided to the one or more chemical production facilities 102 can be represented with a digital representation of the one or more educt(s) 108. For example, the digital representation of the one or more educt(s) 108 may be stored in an inventory of available educts 104. For determining the synthesis instructions for producing a product 106, the digital representation of the one or more educt(s) 108 may be provided. From the digital representation of the one or more educt(s) 108, a digital representation of the chemical structure of the one or more product(s) obtained in one or more chemical reaction(s) from the one or more educt(s) may be determined. Details on determining the digital representation of the chemical structure of the one or more product(s) may be described in the context of FIG. 3 and FIG. 5. Further, it may be determined if the one or more product(s) may be associated with the one or more target product properties by determining the product properties of the one or more product(s). Determining the product properties may be described in further detail in the context of FIG. 3. Further, the one or more formation score(s) associated with forming the one or more product(s) in one or more chemical reaction(s) of the one or more educt(s) may be determined and used for determining the one or more target product(s). The one or more formation score(s) may be indicative of one or more quantities associated with forming the product from the one or more educt(s). Hence, a high formation score may indicate a high conversion rate for producing the product associated with the high formation score. Typically, the formation score may indicate a numerical value between 0 and 1 including 0 and 1 . By using the formation score for determining the one or more target product(s), an efficient production of the one or more target product(s) may be enabled as products with high conversion rates can be selected over products with low conversion rates. Higher conversion rates associated with a product may come with the benefit of lower conversion rates of byproducts. Followingly, determining the target product based on the formation score increases the resource-efficiency of the chemical production.

[0078] The determined synthesis instructions for producing a product 106 may be provided to the one or more chemical production facilities 102 for producing the product associated with the 1 16. The 116 may correspond to the one or more target product properties 1 12 when the synthesis instructions for producing a product 106 may be determined as described herein.

[0079] FIG. 2 illustrates an embodiment of a chemical production network.

[0080] The chemical production network may comprise a plurality of chemical production facilities 202. The chemical production facilities 202 may produce the educts 1-5. Educts 1 , 2 and 4 may be main products of one or more chemical reaction(s) conducted in the chemical production facilities 202. Educts 2 and 3 may be byproducts of one or more chemical reaction(s) conducted in the chemical production facilities 202. The chemical production facilities 202 may be connected via paths for producing products with one or more target product properties. In the example, educt 3 and educt 5 may be byproducts of one or more chemical reaction(s). By determining synthesis instructions as described in this disclosure, in particular the following Figures, target products may be identified and produced from byproducts such as educt 3 and educt 5. These byproducts may otherwise be burned or require energy-intensive treatment to avoid harm to the environment. Followingly, this example shows how an efficient controlling of a complex chemical production network can be enabled by this disclosure.

[0081] FIG. 3 illustrates an embodiment of a method for generating synthesis instructions for producing a target product associated with one or more target product properties.

[0082] One or more target product properties may be provided 302. The one or more target product properties may be provided by providing a request for providing a product with the one or more target product properties comprising the one or more target product properties. The providing of the one or more target product properties may trigger providing a digital representation of a chemical structure of the one or more educt(s) 304. The digital representation of the chemical structure of the one or more educt(s) may be provided by a compound database. The compound database may comprise the digital representation of the chemical structure of the one or more educt(s) available. Hence, digital representation of the chemical structure of the one or more educt(s) may be provided according to an inventory of a chemical production facility and / or a chemical production network. In an embodiment, a location of the chemical production facility and / or network may be provided. The one or more educt(s) may be selected by determining the one or more educt(s) available at the location of the chemical production facility and / or network. The one or more educt(s) may be selected from a plurality of educts. The plurality of educt(s) may be associated with a location. By comparing the location associated with the plurality of educts with the location of the chemical production facility, the one or more educt(s) available at the location of the chemical production facility may be determined. The one or more educt(s) available at the location of the chemical production facility may be associated with a location within a predefined distance from the location of the chemical production facility and / or with a transportation route associated with the location of the chemical production facility.

[0083] The digital representation of a chemical structure may comprise a machine-readable format such as CDX, CDXML, SMILES, SMARTS or a graph representation. This may be further explained in the context of FIG. 6. In an embodiment, the digital representations of the chemical structure of the one or more educt(s) may be transformed into digital representations of the chemical structure of the one or more educt(s) suitable for being provided to a forward modelling data-driven model. The forward modelling data-driven model may require specific data input. In an example, the provided digital representations of the chemical structure of the one or more educt(s) may be associated with a CDX format and / or the forward modelling data-driven model may require SMILES input.

[0084] A digital representation of the chemical structure of the one or more product(s) may be determined from at least a part of the digital representation of the chemical structure of the one or more educt(s) 306. Determining the digital representation of the chemical structure of the one or more product(s) may include determining that the one or more product(s) are formed in one or more chemical reaction(s) of at least a part of the one or more educt(s) associated with a one or more formation score(s). Determining the digital representation of the chemical structure of the one or more product(s) may include providing the digital representation of the chemical structure of the one or more educt(s) to the forward modelling data-driven model. The forward modelling data-driven model may be parametrized and / or trained to receive the digital representation of the chemical structure of the one or more educt(s) and to provide the digital representation of the chemical structure of the one or more product(s). For this purpose, the forward modelling data-driven model may comprise one or more input layer(s). The input layer(s) may be configured for receiving the digital representation of the chemical structure of the one or more educt(s). Receiving the digital representation of the chemical structure of the one or more educt(s) may comprise mapping the digital representation of the chemical structure of the one or more educt(s) to a educt tensor related to the digital representation of the chemical structure of the one or more educt(s). Further, the forward modelling data- driven model may comprise one or more hidden layer(s). The one or more hidden layer(s) may be connected to the one or more input layer(s). The hidden layers may be configured for mapping the educt tensor related to the digital representation of the chemical structure of the one or more educt(s) to a product tensor related to the digital representation of the chemical structure of the one or more product(s). Further, the forward modelling data- driven model may comprise one or more output layer(s). The output layer(s) may be configured for mapping the product tensor related to the digital representation of the chemical structure of the one or more product(s) to the digital representation of the chemical structure of the one or more product(s). Hence, the forward modelling data- driven model may comprise a neural network. The forward modelling data-driven model may be parametrized and / or trained based on a forward modelling training data set. The forward modelling training data set may comprise a plurality of digital representations of the chemical structure of the one or more educt(s) and a plurality of corresponding digital representations of the chemical structure of the one or more product(s). For parametrizing and / or training the forward modelling data-driven model, the forward modelling data-driven model may be provided with the plurality of digital representations of the chemical structure of the one or more educt(s). During the parametrizing and / or training of the forward modelling data-driven model, the parameters associated with the forward modelling data-driven model may be adapted. The forward modelling data-driven model may generate a plurality of digital representations of the chemical structure of one or more product(s). The parameters associated with the forward modelling data-driven model may be adapted to reduce the deviation between the generated digital representations of the chemical structure of the one or more product(s) and the digital representations of the chemical structure of the one or more product(s) associated with the forward modelling training data set. A loss function may be determined and / or used for determining the deviation. The deviation may be reduced by applying a gradient descent algorithm. The forward modelling data-driven model may be parametrized and / or trained if the deviation between the generated digital representations of the chemical structure of the one or more product(s) and the digital representations of the chemical structure of the one or more product(s) associated with the forward modelling training data set may be within a predefined range.

[0085] An example of input and output data associated with the forward modelling data-driven model can be seen in FIG. 9.

[0086] Forward modelling data-driven model may be further based on physical equations. Hence, the forward modelling data-driven model may be a hybrid model. For example, the reactivity of at least one of the educts may be derived from the digital representation of the chemical structure of the one or more educt(s). The reactivity may be further used as input and / or the digital representation of the chemical structure of the one or more educt(s) may be enriched by the reactivity. The reactivity may be determined based on physical equations and / or assumptions known in the art. For example, the reactivity may be determined based on electronegativity and / or electro positivity associated with the one or more educt(s). Additionally or alternatively, determining the digital representation of the chemical structure of the one or more product(s) may include providing the digital representation of the chemical structure of the one or more product(s) by a compound database in response to providing a query associated with the digital representation of the chemical structure of the one or more educt(s) to the compound database. This might be especially advantageous where the product may be already known or where in a preceding workflow the digital representation of the chemical structure of the one or more product(s) may have been determined.

[0087] Additionally or alternatively, determining the digital representation of the chemical structure of the one or more product(s) may be based on applying reaction templates. The reaction template may be indicative of a change in the chemical structure of the one or more educt(s) towards the one or more product(s). The reaction template may be described in more detail within the context of FIG. 5.

[0088] One or more formation score(s) associated with forming the one or more product(s) in one or more chemical reaction(s) of at least the part of the one or more educt(s) may be determined 308. The formation score may be indicative of a likelihood that the one or more product(s) may be obtained by the one or more chemical reaction(s) from the one or more educt(s). To determine the one or more formation score(s), the digital representation of the chemical structure of the one or more product(s) may be provided to a scoring data-driven model. The scoring data-driven model may be configured to determine the formation scores from digital representation of the chemical structure of the one or more product(s) and optionally the digital representation of the chemical structure of the one or more educt(s). For this purpose, the scoring data-driven model may comprise one or more input layer(s). The input layer(s) may be configured for receiving the digital representation of the chemical structure of the one or more educt(s) and / or one or more product(s). Receiving the digital representation of the chemical structure of the one or more educt(s) and / or one or more product(s) may comprise mapping the digital representation of the chemical structure of the one or more educt(s) and / or one or more product(s) to a chemical product tensor related to the digital representation of the chemical structure of the one or more educt(s) and / or one or more product(s). Further, the scoring data-driven model may comprise one or more hidden layer(s). The one or more hidden layer(s) may be connected to the one or more input layer(s). The hidden layers may be configured for mapping the chemical product tensor related to the digital representation of the chemical structure of the one or more educt(s) and / or one or more product(s) to a score tensor related to the formation score associated with forming the one or more product(s) in one or more chemical reaction(s) of at least a part of the one or more educt(s). Further, the scoring data-driven model may comprise one or more output layer(s). The output layer(s) may be configured for mapping the score tensor to the formation score. Hence, the scoring data-driven model may comprise a neural network. An example of input and output data associated with the scoring data-driven model can be seen in FIG. 9. The scoring data-driven model may be trained and / or parametrized based on a scoring training data set. The scoring training data set may comprise a plurality of digital representations of the chemical structure of products and / or educts and corresponding formation scores. For parametrizing and / or training the scoring data-driven model, the scoring data-driven model may be provided with the plurality of digital representations of the chemical structure of the educts and / or products. The scoring data-driven model may generate a plurality of formation scores associated with forming the products in one or more chemical reaction(s) from the educts. During the parametrizing and / or training of the scoring data-driven model, the parameters associated with the scoring data- driven model may be adapted. The parameters associated with the scoring data-driven model may be adapted to reduce the deviation between the generated formation scores and the formation score associated with the scoring training data set. A loss function may be determined and / or used for determining the deviation. The deviation may be reduced by applying a gradient descent algorithm. The scoring data-driven model may be parametrized and / or trained if the deviation between the generated formation scores and the formation scores associated with the scoring training data set may be within a predefined range.

[0089] Additionally or alternatively, determining the one or more formation score(s) may comprise applying atom mapping. Atom mapping may include determining the share of atoms remaining unchanged during the one or more chemical reactions. Atom mapping may be described in more detail in the context of FIG. 7.

[0090] One or more product properties associated with the one or more product(s) from the digital representation of the chemical structure of the one or more product(s) and optionally the digital representation of the one or more educt(s) may be determined 310.

[0091] In an embodiment, determining the one or more product properties from the digital representation of the chemical structure of the chemical structure of the one or more product(s) may comprise providing the digital representation of the chemical structure of the one or more educt(s) to a property database for providing the one or more product properties. The property database may be configured for receiving the digital representations of the chemical structure of the one or more product(s) and providing one or more product properties in response to receiving the digital representation of the chemical structure of the one or more product(s). The property database may comprise a plurality of digital representations of the chemical structure of products and corresponding product properties. By doing so, target products can be determined fast and reliably while saving resources for carrying out experiments as historical data from already conducted experiments are used.

[0092] In an embodiment, determining the one or more product properties from the digital representation of the chemical structure of the chemical structure of the one or more product(s) may comprise providing the digital representation of the chemical structure of the one or more educt(s) to a property data-driven model. The property data-driven model may be configured for providing the one or more product properties in response to receiving digital representation of the chemical structure of the one or more product(s) and optionally of the one or more educt(s). For this purpose, the property data-driven model may comprise one or more input layer(s). The input layer(s) may be configured for receiving the digital representation of the chemical structure of the one or more product(s). Receiving the digital representation of the chemical structure of the one or more product(s) may comprise mapping the digital representation of the chemical structure of the one or more product(s) to a chemical product tensor related to the digital representation of the chemical structure of the one or more product(s). Further, the property data-driven model may comprise one or more hidden layer(s). The one or more hidden layer(s) may be connected to the one or more input layer(s). The hidden layers may be configured for mapping the chemical product tensor related to the digital representation of the chemical structure of the one or more product(s) to a property tensor related to the one or more product properties associated with the one or more product(s). Further, the property data-driven model may comprise one or more output layer(s). The output layer(s) may be configured for mapping the property tensor to the one or more product properties. Hence, the property data-driven model may comprise a neural network. An example of input and output data associated with the property data-driven model can be seen in FIG. 9. The property data-driven model may be trained and / or parametrized based on a plurality of digital representations of the chemical structure of products and corresponding product properties. By doing so, target products can be determined fast and reliably while saving resources for carrying out experiments.

[0093] The property data-driven model may be trained and / or parametrized based on a property training data set. The property training data set may comprise a plurality of digital representations of the chemical structure of products and corresponding product properties. For parametrizing and / or training the property data-driven model, the property data-driven model may be provided with the plurality of digital representations of the chemical structure of the products. The property data-driven model may generate a plurality of product properties associated with the products. During the parametrizing and / or training of the property data-driven model, the parameters associated with the property data-driven model may be adapted. The parameters associated with the property data-driven model may be adapted to reduce the deviation between the generated product properties and the product properties associated with the property training data set. A loss function may be determined and / or used for determining the deviation. The deviation may be reduced by applying a gradient descent algorithm. The property data-driven model may be parametrized and / or trained if the deviation between the generated product properties and the product properties associated with the property training data set may be within a predefined range.

[0094] Further, determining the one or more product properties may include providing the one or more product properties via a user interface. In an example, the one or more product properties may be determined by conducting an experiment. Providing the one or more product properties via the user interface allows for taking into account experimental data. A predefined score range associated with determining that the one or more product(s) may be formed in the one or more chemical reaction(s) from the one or more educt(s) may be provided, e.g. via a user interface 312. This allows for tailoring the generation of the synthesis instructions to requirements regarding conversion rates of products and to exclude inefficient chemical reactions from producing the one or more target product(s).

[0095] It may be determined if the one or more product(s) are the one or more target product(s) based on the one or more formation score(s) and the one or more product properties of the one or more product(s) 312. This may include determining if the one or more formation score(s) may be within the predefined score range by comparing the one or more formation score(s) and the predefined score range. Further, determining if the one or more product(s) may be the one or more target product(s) may include comparing the one or more product properties with the one or more target product properties. Preferably, the at least one product with one or more product properties more similar to the one or more target product properties than the other properties of the one or more product(s) may be selected to be the one or more target product(s).

[0096] The selected digital representation of the chemical structure of the one or more product(s) may be provided 316. In an embodiment, digital representations of the chemical structure of the one or more educt(s) may be further provided in particular to a chemical production facility for producing the one or more target product(s) as described within the context of FIG. 1 .

[0097] FIG. 4 illustrates the formation of a plurality of products 404a - 404c based on plurality of educts 402a - 402b. The educts 402a - 402b may be represented based on their chemical structure. In FIG. 4 the chemical structure of the educts 402a - 402b and the products 404a - 404b may be represented via a skeletal formula. This representation may be suitable for humans to recognize the chemical structure. For performing machine learning, the educts 402a - 402b and the products 404a - 404b may be represented via a machine-readable representation of the educts 402a - 402b and the products 404a - 404b . Such a machine-readably representation may be for example Simplified Molecular Input Line Entry Specification (SMILES), SMILES arbitrary target specification (SMARTS), mol files, pdb files or the like. An example for a SMILES 602 of a educt molecule 402a can be found in FIG. 6. The educts 402a - 402b and the products 404a - 404b are preferably organic molecules optionally being associated with a molecular weight smaller than 1000 u, wherein u may refer to an unified atomic mass number specifying 1 / 12 of a mass of a carbon atom comprising 6 protons and 6 neutrons in the nucleus. Educt 402a may be a carboxylic acid and may be associated with a digital representation of the chemical structure of the educt 402a. Educt 402a may undergo different reactions depending on the reaction partner. For example, educt 402a may form an acid anhydride together with educt 402c as it can be seen in reaction 1 of FIG. 4. Educt 402c may be associated with a digital representation of the chemical structure of educt 402c. In reaction 1 of FIG. 4 educt 402a and 402c may form a product, here product 404a, preferably under acidic conditions. The product may be associated with a digital representation of the chemical structure of product 404a. As educt 402c comprises another functional group being a hydroxyl group independent of the carboxylic acid group, another reaction mechanism may be plausible. For example, the hydroxyl group of 402c and the carboxylic acid group of 402a may form an ester group resulting in a product different than 404a.

[0098] In an example, reactions 2&3 show the reaction of educts 402a and 402b. As described in the former passage, again more than one product may be plausible based on combining educts 402c and 402b leading to at least two products 404d and 404c as shown in reactions 2&3 in FIG. 4. Product 404d may be a result of an esterification of the carboxylic acid group of 402c and the hydroxyl group of 402b. Product 404c may be a result of a etherification of the hydroxyl group independent of the carboxylic acid group of educt 402c and the hydroxyl group of educt 402b. Similar to reaction 1 , reactions 2&3 may be carried out under acidic conditions.

[0099] Additionally or alternatively, an educt may react with itself. In an example, educt 402a may react with itself to form product 404b in reaction 4. Similar to reactions 1 -3, reaction 4 may be carried out under acidic conditions.

[0100] FIG. 5 illustrates the principle of templates, retrosynthetic modelling and forward synthetic modelling.

[0101] Determining a digital representation of the chemical structure of the one or more product(s) may be based on reaction templates. A reaction template may specify one or more functional groups, preferably one or more functional groups involved into a chemical reaction described by the reaction template, associated with one or more educts and / or products, preferably one or more educts involved into a chemical reaction described by the reaction template and / or one or more products involved into a chemical reaction described by the reaction template. Followingly, the reaction template may represent a schematic of a group of molecules potentially undergoing the reaction associated with the reaction template.

[0102] In the example shown in FIG. 5, reaction templates 502, 504, 506 may be represented in the skeletal formula. The reaction templates 502, 504, 506 in FIG. 5 may depict the functional groups of an acyl chloride 502, a primary amine 504 and an amide 506. The amide 506 may be a product and may be formed based on the acyl chloride 502 and the amine 504. Determining a product formed based on or from one or more educts may include forward synthetic modelling. Forward synthetic modelling may specify the direction of the reaction associated with the formation of the product. In contrast to forward synthetic modelling, retrosynthetic modelling describes the opposite direction. Retrosynthesis in general refers to a technique for determining one or more educts based on one or more products. Retrosynthetic modelling may describe in silico techniques for determining one or more educts based on one or more educt(s). Reaction templates may be used for forward synthetic modelling and retrosynthetic modelling. Commonly, in both forward synthetic modelling and retrosynthetic modelling n to n relations may be established between the one or more educts and one or more products. Followingly, the reaction template associated with one or more products and one or more educts of a chemical reaction may be applied for determining the one or more products and / or one or more educts of a chemical reaction associated with the reaction template. An example of a reaction template may include reaction SMILES, reaction SMARTS, SMIRKS or the like.

[0103] Forward synthetic modelling may include applying a forward modelling template, preferably to at least a part of the digital representation of the chemical structure of the one or more educt(s). Applying the forward modelling template to at least a part of the digital representation of a chemical structure of the one or more educt(s) may result in the digital representation of a chemical structure of the one or more product(s). The forward modelling template may specify the connection of two or more functional groups in relation to the one or more educts. The connection of two or more functional groups may result in the formation of a new functional group, preferably a product functional group. In the current example of FIG. 5, the acyl chloride group of 502 and the amine group of 504 may form the amide group of 506. Furthermore, the template may specify lost atoms during the reaction. For example, the amide formation of FIG. 5 results in the loos of hydrochloric acid.

[0104] Retrosynthetic modelling may be suitable for verifying a reaction pathway. A reaction may be more likely if the products may be determined based on one or more educts and if the one or more educts may be determined from the product based on retrosynthetic modelling.

[0105] Retrosynthetic modelling may include applying a retrosynthetic template, preferably to the digital representation of the chemical structure of the one or more product(s). Applying the retrosynthetic template may result in the digital representation of the chemical structure of the one or more educt(s). The retrosynthetic template may specify the separation of one or more functional groups related to the one or more product(s). The separation of one or more functional groups may result in the formation of two or more new functional group, preferably two or more educt functional groups separated among one or more educts. In the current example of FIG. 5, the amide group 506 was formed based on the acyl chloride group of 502 and the amine group of 504. Furthermore, the template may specify addition atoms during the reaction. For example, the amide separation of FIG. 5 results in the addition of one hydrogen atom to one educt 504 and one chloric atom to the other educt 502.

[0106] FIG. 6 illustrates representations of a chemical structure associated with a carboxylic acid. A molecule, in particular organic molecules may be represented in the skeletal structure 604. Skeletal structure 604 may be a representation of the chemical structure associated with the molecule, wherein each line constitutes a chemical bond and edges constitute carbon atoms optionally together with the corresponding number of hydrogen atoms to result in the adequate valency associated with the carbon atoms. Atoms other than carbon atoms optionally together with the corresponding number of hydrogen atoms may be represented by the element symbol. Similar to the skeletal structure 604, a graph representation 606 may specify edges and vertices. Edges associated with the graph representation 606 may specify atoms and vertices may specify the chemical bonds between atoms. The graph representation 606 may be easily machine-readable while allowing to associate every edge representing an atom by its very own and characteristic atomic properties. For example, the carboxylic acid represented in FIG. 6 comprises a plurality of hydrogen atoms with the same valency. The hydrogen atom bound to the oxygen atom of the hydroxyl group may polarized and / or shielded differently from the hydrogen atoms bound to the most left carbon atom. The hydrogen atoms bound to the most left carbon atom are again different from the hydrogen atoms bound to the hydrogen atoms of the second carbon atoms counted from the right side. Followingly, the graph representation 606 is advantageous for representing accurately the complex specifics of an organic molecule.

[0107] Additionally or alternatively, a molecule may be represented by a SMILES 602. The SMILES 602 may follow specific rules as known in the art. A further varied representation may be SMARTS. Both representations may be developed for being machine-readable.

[0108] FIG. 7 illustrates an embodiment of determining a formation score.

[0109] The one or more formation score(s) may be determined based on atom mapping, in particular atom mapping with respect to the digital representation of the chemical structure of the one or more educt(s) and / or one or more product(s). Atom mapping may refer to determining a matching score of the atoms associated with the one or more educts and / or the atoms associated with the one or more products, preferably called the atom matching score. Hence, the one or more formation score(s) may comprise one or more atom matching score(s). The atom matching score may specify a matching of atoms in the one or more educts and the one or more products. The atom matching score may be determined by providing a digital representation of the chemical structure of the one or more educt(s) and one or more product(s) indicative of a position of the one or more atoms of the one or more educt(s) and / or one or more product(s) 706. From the digital representation of the chemical structure of the one or more educt(s) and one or more product(s), a total number of atoms of at least a part of the one or more educt(s), in particular atoms of at least the part of the one or more educt(s) participating in the one or more chemical reaction(s) forming the one or more product(s) 708. Further, a fraction number of atoms of the one or more educt(s) associated with a position unchanged by the one or more chemical reaction(s) forming the one or more product(s) from the one or more educt(s) may be determined 708. The atom matching score may be determined by dividing the fraction number by the total number 710. The atom matching score may be provided 712 e.g. by the formation score determining engine 840.

[0110] The atom matching score may be for example a number between 0 and 1 . In FIG. 7 a condensation reaction may be depicted. The hexanoic acid 702a may undergo a reaction with hexanoic acid 702b under loss of one water molecule to form a hexanoic acid anhydride 704. During the condensation reaction, the oxygen atom of the hydroxyl group of the hexanoic acid 702a may bound to the carbon atom of the carboxylic acid group of another hexanoic acid 702b to form the acid anhydride. In a condensation reaction, a water molecule may be separated from the reactants. The water molecule of the reaction shown in FIG. 7 may be formed based on the hydrogen atom of the hydroxyl group with respect to hexanoic acid 702a and the hydroxyl group with respect to hexanoic acid 702b. Followingly, these atoms may not be part of the formed product and these atoms may not be matched to the product. Since the rest of the atoms of the hexanoic acid 702a and hexanoic acid 702b may remain, the corresponding hexanoic acid anhydride 704 may comprise the majority of atoms of the hexanoic acid 702a and the hexanoic acid 702b. In this example, hexanoic acid 702a and hexanoic acid 702b may comprise 17 atoms, whereas hexanoic acid anhydride 704 may comprise 31 atoms. Hence, the total number may be 34. The fraction number may be 31. Thus, the matching score of the educts hexanoic acid 702a and hexanoic acid 702b with respect to the product hexanoic acid anhydride 704 may be around 91 .2 %. As the number of matching atoms and / or the matching score may be higher than a predefined threshold, the resulting one or more formation score(s) may signify the reaction pathway. Hence, atom matching score may be determined based on dividing the number of atoms of the one or more educt(s) remaining in a position of the corresponding number of atoms of the one or more product(s) by a total number of atoms of the one or more educt(s). Further, the atom matching score may be determined based on a corresponding weighting of the atoms. This may increase the significance since hydrogen atoms e.g. in an alkyl group may be less of interest than more reactive groups such as hydroxyl hydrogen atoms or oxygen atoms.

[0111] Additionally or alternatively, the matching score may be used and / or may be averaged with other scores. Other parameters for evaluating the likelihood of a reaction pathway may be used to strengthen the reliability of the one or more formation score(s). For example, in reaction 2&3 of FIG. 4 two possible reaction pathways may be shown, one being an esterification and the other being an etherification. In both reactions, a water molecule may be lost. Hence, atom mapping would classify reactions 2&3 both identical. Nevertheless, for evaluation the feasibility of a reaction pathway, the reactivity and the resulting tendency of the one or more educts to form the one or more products, in particular in relation to each other, may be relevant to be considered. For example, when comparing the etherification and the esterification, the esterification may be more likely due to the increased reactivity of the hydrogen atom corresponding to the carboxyl acid group than the hydrogen atom corresponding to the hydroxyl group independent of the carboxyl acid group because of the higher polarization of a hydrogen atom in a carboxyl acid group. Followingly, the likelihood of the esterification may be higher than compared to the etherification due to the difference in polarization resulting in a difference in reactivity.

[0112] Ultimately, the one or more formation score(s) may be based on the retrosynthetic approach, the scoring data- driven model, in particular the output of the scoring data-driven model and / or the atom mapping. In an embodiment, the one or more formation score(s) may be based on weighting of the retrosynthetic approach, the scoring data-driven model, in particular the output of the scoring data-driven model and / or the atom mapping. FIG. 8 illustrates an embodiment of an apparatus for generating synthesis instructions for producing a target product.

[0113] In an embodiment, the apparatus may comprise an intake interface 802, a chemical structure generating engine 842, a compound database 806 comprising digital representations of a chemical structure of digital representations of the chemical structure of the one or more educt(s) 824, a property determining engine 844, a formation score determining engine 840, a product determining engine 846 and / or an output interface 848. The one or more target product properties may be provided to the chemical structure generating engine 842. Further, digital representations of the chemical structure of the one or more educt(s) 824 may be provided, in particular by the compound database 806 to the chemical structure generating engine 842. The providing of the one or more target product properties may trigger the providing of the digital representations of the chemical structure of the one or more educt(s) 824.

[0114] A request for receiving the digital representations of the chemical structure of the one or more educt(s) 824 may be provided by the chemical structure generating engine 842 to the compound database 806. The request for receiving the digital representations of the chemical structure of the one or more educt(s) 824 may be a query suitable for being provided to the compound database 806. The query may comprise retrieving instructions for retrieving the digital representations of the chemical structure of the one or more educt(s) 824. The compound database 806 may be a structured compound database 806. Additionally or alternatively, the compound database 806 may be an embedding compound database 806. The request may comprise an embedded request. The embedded request may be a representation of the request, in particular a representation comprising a vector. The embedding compound database 806 may be configured for determining the digital representations of the chemical structure of the one or more educt(s) 824 corresponding to the received request by determining a distance between embedded digital representations of the chemical structure of the one or more educt(s) 824 and the embedded request, and selecting the digital representations of the chemical structure of the one or more educt(s) 824 associated with the distance within a predefined range.

[0115] The selected digital representations of the chemical structure of the one or more educt(s) 824 may be provided by the embedding compound database 806.

[0116] The providing of the digital representations of the chemical structure of the one or more educt(s) 824 may trigger the chemical structure generating engine 842 to determine one or more digital representations of a chemical structure of the one or more product(s). The chemical structure generating engine 842 may be configured for determining the one or more digital representations of a chemical structure of the one or more product(s) from at least a part of the digital representations of the chemical structure of the one or more educt(s) 824. Determining the one or more digital representations of a chemical structure of the one or more product(s) may be described in further detail in the context of FIG. 3 and FIG. 5.

[0117] The one or more digital representations of a chemical structure of the one or more product(s) may be provided to a formation score determining engine 840. The formation score determining engine 840 may be configured for determining a formation score associated with a formation of the one or more product(s) from the one or more educt(s). For example, a high formation score may indicate a high rate of formation of the products. The formation score may be obtained by calculating the degree of atomic configurations unchanged by the chemical reaction of the one or more educt(s) to the products. As the production of products may be associated with equilibria and incomplete conversions, the formation score may be indicative of the efficiency of a production process. The one or more products may be selected if the formation score associated with the formation of the one or more products may be within a predefined range. This allows for increasing the efficiency of the production of products and reduces the amount of undesired byproducts. Thereby, waste in the chemical production is reduced. Determining the formation score may be described in further detail in the context of FIG. 3 and FIG. 7.

[0118] Further, the one or more digital representations of a chemical structure of the one or more product(s) may be received by the property determining engine 844. The property determining engine 844 may be configured for determining a property of the one or more products from at least a part of the digital representations of the chemical structure of the one or more educt(s) 824. For example, the property determining engine 844 may comprise a classification model. The classification model may be configured for classifying the digital representation of the chemical structure of the one or more products according to the one or more properties associated with the one or more products. Determining the one or more properties associated with the one or more products may be as described in the context of FIG. 3.

[0119] The one or more determined properties may be provided to a product determining engine 846. The determined formation score may be provided to the product determining engine 846. The digital representations of a chemical structure of the one or more product(s) may be provided to the product determining engine 846. The product determining engine 846 may be configured for selecting the one or more product(s) associated with the one or more target product properties e.g. by comparing the one or more properties of the one or more product(s) with the one or more target product properties. The product determining engine 846 may further select the one or more product(s) by determining that the formation score associated with the one or more product(s) may be within a predefined range. Hence, the property determining engine 844 may select the one or more product(s) with the one or more target product properties from the one or more product(s) associated with the digital representations digital representation of the chemical structure of the one or more product(s) as obtained by the chemical structure generating engine 842. The chemical structure generating engine 842 may provide the digital representation of the chemical structure of the one or more target product(s) to the output interface 848. Further, the product determining engine 846 may provide the one or more formation score(s) and / or the one or more properties associated with the selected one or more product(s) to the output interface 848. By using the abovedescribed engines and / or conducting the above described acts, the products with the one or more target product properties can be obtained efficiently from e.g. educts available to the one or more chemical production facilities 102.

[0120] The digital representation of the chemical structure of the one or more target product(s) may be provided to the one or more chemical production facilities 102 for producing the one or more target product(s) as described in the context of FIG. 1 .

[0121] FIG. 9 illustrates an embodiment of input and output data associated with the forward modelling data-driven model, the scoring data-driven model and / or the property data-driven model.

[0122] In an example, the forward modelling data-driven model may receive SMILES input 602 associated with pentanoic acid. The SMILES 602 may be an example of a digital representation of the chemical structure of an educt. The forward modelling data-driven model may be configured for providing a digital representation of the chemical structure of the one or more product(s) in response to receiving the digital representation of the chemical structure of the one or more educt(s) as described in the context of FIG. 3. The SMILES output 958 may be an example of the digital representation of the chemical structure of the one or more product(s). In the example, the SMILES 958 may be associated with the product of an esterification of pentanoic acid. The corresponding chemical structure can be seen in FIG. 4, specifically reaction 4.

[0123] In an example, the scoring data-driven model may receive SMILES input 602 associated with pentanoic acid. The scoring data-driven model may be configured for providing a formation score associated with the pentanoic acid in response to receiving the digital representation of the chemical structure of the one or more educt(s) as described in the context of FIG. 3. In this example, the formation score may be indicative of a chemical stability of the pentanoic acid.

[0124] In an example, the property data-driven model may be provided with the SMILES 602 to determine one or more product properties associated with the product, in particular associated with pentanoic acid. For example, the property data-driven model may be configured for determining a pKavalue associated with a product. The pKamay be an indicator of acidic properties of a product. For this purpose, the property data-driven model may generate a pKavalue vector indicative of a plurality of confidence scores associated with a plurality of pKavalues ranging from 1 to n, wherein n may be a natural number. The pKavector may be associated with a pKavalue. Hence, the property data-driven model may determine the pKavalue. The present disclosure has been described in conjunction with preferred embodiments and examples as well. However, other variations can be understood and effected by those persons skilled in the art and practicing the claimed invention, from the studies of the drawings, this disclosure and the claims. Notably, in particular, the any steps presented can be performed in any order, i.e. the present invention is not limited to a specific order of these steps. Moreover, it is also not required that the different steps are performed at a certain place or at one node of a distributed system, i.e. each of the steps may be performed at different nodes using different equipment / data processing.

[0125] As used herein ..determining" also includes ..initiating or causing to determine", “generating" also includes ..initiating and / or causing to generate" and “providing” also includes “initiating or causing to determine, generate, select, send and / or receive”. “Initiating or causing to perform an action” includes any processing signal that triggers a computing node or device to perform the respective action.

[0126] In the claims as well as in the description the word “comprising” or “including” or similar wording does not exclude other elements or steps and shall not be construed limiting to the elements or steps lined out. The indefinite article “a” or “an” does not exclude a plurality. A single element or other unit may fulfill the functions of several entities or items recited in the claims. The mere fact that certain measures are recited in the mutual different dependent claims does not indicate that a combination of these measures cannot be used in an advantageous implementation or further elements may be included.

[0127] Providing in the scope of this disclosure may include any interface configured to provide data. This may include an application programming interface, a human-machine interface such as a display and / or a software module interface. Providing may include communication of data or submission of data to the interface, in particular display to a user or use of the data by the receiving entity.

[0128] Any disclosure and embodiments described herein relate to the methods, the systems, devices, the computer program element lined out above and vice versa. Advantageously, the benefits provided by any of the embodiments and examples equally apply to all other embodiments and examples and vice versa.

Claims

CLAIMS1 . A method, in particular a computer-implemented method, for generating synthesis instructions for producing one or more target product(s) associated with one or more target product properties, the method comprising: providing the one or more target product properties, providing a digital representation of a chemical structure of one or more educt(s), determining a digital representation of a chemical structure of one or more product(s) from at least a part of the digital representation of the chemical structure of the one or more educt(s) by determining that the one or more product(s) are formed in one or more chemical reaction(s) of at least a part of the one or more educt(s), associated with one or more formation score(s), determining if the one or more product(s) are the one or more target product(s) according to the one or more formation score(s) and by determining if one or more product properties associated with the one or more product(s) correspond to the one or more target product properties, providing the digital representation of the chemical structure of the one or more product(s) in response to determining that the one or more product properties correspond the one or more target product(s), in particular for monitoring and / or controlling producing and / or processing the one or more product(s), and optionally providing the digital representation of the one or more educt(s) associated with the one or more target product(s).

2. The method of claim 1 , wherein determining if one or more product properties associated with the one or more product(s) correspond to the one or more target product properties comprises determining one or more product properties associated with the one or more product(s) from the digital representation of the chemical structure of the one or more product(s) and optionally the digital representation of the one or more educt(s) and comparing the one or more product properties with the one or more target product properties.

3. The method of claim 2, wherein the digital representation of the chemical structure of two or more product(s) is determined associated with two or more formation score(s), and wherein two or more product properties associated with the two or more product(s) are determined from the digital representation of the chemical structure of the two or more product(s) and optionally the digital representation of the one or more educt(s), and wherein the two or more product properties are compared with the one or more target product properties, and wherein the digital representation of the chemical structure of at least one of the two or more product(s) are provided in response to determining that the at least one product corresponds to the one or more target product(s).

4. The method of claim 2 or 3, wherein determining the one or more product properties comprises providing the digital representation of the chemical structure of the one or more product(s) to a property data-driven model and / or receiving the one or more product properties from the property data-driven model, preferably in response to having provided the digital representation of the chemical structure of the one or more product(s), wherein the property data-driven model is configured to provide product properties in response to being provided with digital representation of the chemical structure of products.

5. The method of claim 2 or 4, wherein determining the one or more product properties comprises providing the digital representation of the chemical structure of the one or more product(s) to a property database for receiving the one or more product properties, wherein the property database is configured for receiving the digital representations of the chemical structure products and providing the one or more product properties in response to receiving the digital representations of the chemical structure of the one or more product(s).

6. The method of any one of claims 1 to 5, wherein determining if the one or more product(s) correspond to the one or more target product(s) according to the one or more formation score(s) comprises determining the one or more formation score(s) associated with forming the one or more product(s) in the one or more chemical reaction(s), and providing a predefined score range associated with determining that the one or more product(s) are formed in the one or more chemical reaction(s) from the one or more educt(s) and determining if at least one of the one or more formation score(s) is within the predefined score range.

7. The method of any one of claims 1 to 6, wherein determining a digital representation of a chemical structure of the one or more product(s) from at least the part of the digital representation of the chemical structure of the one or more educt(s) may comprise providing at least the part of the digital representation of the chemical structure of the one or more educt(s) to a forward modelling data-driven model for generating the digital representation of the chemical structure of the one or more product(s), wherein the forward modelling data-driven model is configured to receive digital representations of the chemical structure of educt(s) and generate the digital representations of the chemical structure of products from the digital representation of the chemical structure of the one or more educt(s).

8. The method of any one of claims 1 to 7, wherein determining a digital representation of a chemical structure of the one or more product(s) from at least the part of the digital representation of the chemical structure of the one or more educt(s) comprises providing a request for providing the digital representation of the chemical structure of the one or more product(s) to a compound database and providing the digital representation of the chemical structure of the one or more product(s) by the compound database, wherein the request is indicative of the one or more educt(s), and wherein the compound database is configured for providing digital representations of the chemical structure of one or more product(s) in response toreceiving requests for providing the digital representations of the chemical structure of one or more product(s).

9. The method of any one of the preceding claims, wherein determining a digital representation of a chemical structure of the one or more product(s) from the digital representation of the chemical structure of the one or more educt(s) comprises applying a forward modelling template to digital representation of the chemical structure of the one or more educt(s), wherein the forward modelling template indicates a change in the chemical structure of at least a part of the one or more educt(s) to result in the chemical structure of the one or more product(s).

10. The method of any one of claims 7 to 9, further comprising generating one or more chemical fingerprints associated with the one or more educt(s) from the digital representation of the chemical structure of the one or more educt(s), wherein the chemical fingerprint is a multidimensional representation of the digital representation of the chemical structure of the one or more educts and wherein providing the digital representation of the chemical structure of the one or more educt(s) comprises providing the chemical fingerprint of the one or more educt(s) and wherein determining a digital representation of a chemical structure of the one or more product(s) from at least a part of the digital representation of the chemical structure of the one or more educt(s) comprises determining a digital representation of a chemical structure of the one or more product(s) from at least a part of the chemical fingerprint of the one or more educt(s).1 1 . The method of any one of claims 1 to 10, wherein an inventory comprises at least a part of the one or more educt(s).

12. The method of any one of claims 1 to 11 , wherein determining if the one or more product(s) correspond to the one or more target product(s) according to the one or more formation score(s) comprises determining the one or more formation score(s) associated with forming the one or more product(s) in the one or more chemical reaction(s), and providing a plurality of predefined score ranges associated with a plurality of different chemical reactions, selecting the predefined score range associated with the chemical reaction in which the one or more product(s) are formed, and determining if at least one of the one or more formation score(s) is within the selected predefined score range.

13. An apparatus for generating a synthesis instructions of a target product associated with one or more target product properties, the computing apparatus comprising: a processor configured for performing any one of the methods according to any one of claims 1 to14. Use of a digital representation of a chemical structure of a target product as generated according to any one of claims 1 to 12 for producing the target product.

15. Use of a digital representation of the chemical structure of the one or more educt(s) associated with the target product as obtained by any one the any one of claims 1 to 12 for producing the target product.

Images