An ai-based platform for predicting impacts of compounds and combinations thereof on a tissue

Abstract

The present invention relates to a computer-implemented artificial intelligence (Al)-based system and method for predicting at least one impact on a desired tissue by at least one compound, including a database containing information associated at least with compounds; and an Al-based engine which computes for each compound an impact on the desired tissue by an activity predicting module which applies to the database and to the compounds an algorithm based on a machine learning model and provides a predicted activity of the compound based on a structure of the compound, and an impact predicting module which applies to the predicted activity an algorithm utilizing an Al-based tissue model of a reference tissue to predict the at least one impact on the desired tissue. The present invention further provides additional modules and methods of implementing the system.

Description

[0001] AN AI-BASED PLATFORM FOR PREDICTING IMPACTS OF COMPOUNDS AND COMBINATIONS THEREOF ON A TISSUE

[0002] FIELD OF THE INVENTION

[0003] The present disclosure generally relates to a computer implemented method and system for prediction of impacts of compounds and combinations thereof on skin, in particular an artificial intelligence (Al)-based method and system for prediction of impacts of compounds and combinations thereof on skin.

[0004] BACKGROUND OF THE INVENTION

[0005] The cosmetics and dermatological industries are constantly evolving, with new ingredients and new products being launched to the market every year. There is a special interest in innovative dermo-cosmetics and hair care ingredients of natural origins. However, identifying new natural extracts, botanical complexes or compounds of natural origins is like looking for a needle in a haystack. Indeed, there are over 300,000 plants in the world (not including algae, mushroom, fungi and other potentially relevant organisms) and over two million different molecular compounds.

[0006] Studying each one of them in the lab to identify their potential for treating and helping with various skin and hair conditions is unrealistic and new approaches are needed in order to automatically identify extracts and compounds with potential benefits on the skin and. On the other end, there is a similar need to predict potential adverse impacts of such compounds.

[0007] SUMMARY OF INVENTION

[0008] The following embodiments and aspects thereof are described and illustrated in conjunction with compositions and methods which are meant to be exemplary and illustrative, not limiting in scope. In various embodiments, one or more of the above-described problems have been reduced or eliminated, while other embodiments are directed to other advantages or improvements.

[0009] According to some embodiments, there is provided a computer- implemented artificial intelligence (Al)-based system for predicting at least one impact on a desired tissue by at least one compound, the system including: a) an input module, configured to receive at least one compound identifier of the at least one compound, and optionally a name of a desired tissue; b) a database including information associated with at least one category including a compounds category; c) an Al-based engine configured to compute for each compound of the at least one compound received by the input module at least one impact on the desired tissue by: i. an activity predicting module configured to apply to the database and to each one of the at least one compound an algorithm based on a machine learning model and to provide a predicted activity of the compound based on a structure of the compound, and ii. an impact predicting module configured to apply to the activity predicted by the activity predicting module an algorithm utilizing an Al-based tissue model of a reference tissue to predict the at least one impact on the desired tissue, and to provide the predicted impact to the output module; and d) an output module, configured to provide the at least one predicted impact of the at least one compound on the desired tissue.

[0010] In some embodiments, the desired tissue and / or the reference tissue is a mammalian tissue. In some embodiments, the desired tissue and / or the reference tissue is selected from skin, hair, oral cavity tissues, and combinations thereof.

[0011] In some embodiments, the at least one impact is selected from a scent, a taste, a color, a texture, and combinations thereof. In some embodiments, the at least one impact includes an adverse effect selected from a toxic effect, a phototoxic effect, an endocrine effect, a carcinogenic effect, and combinations thereof. In some embodiments, the at least one impact includes a positive, a negative, or a neutral effect on at least one condition associated with skin, hair, and / or the digestive tract. In some embodiments, the at least one impact includes a positive, a negative, or a neutral effect on at least one condition selected from: acne, aging, dandruff, dermatitis, dryness, elasticity, inflammation, UV filtering, perspiration, microbiome profile, oiliness, photoaging, pigmentation, redness, rosacea, scalp-related issues, scarring, seborrhea, sunburns, would healing, wrinkles, alopecia, hair strengthening, hair sharpening, hair growth, gingivitis, breath-freshening, gum health, cavity prevention, tooth whitening, tartar prevention, general well-being, antioxidation, a skin disease or condition, and combinations thereof.

[0012] In some embodiments, the at least one compound includes a natural compound. In some embodiments, the at least one compound includes a synthetic compound. In some embodiments, the at least one compound includes at least one compound having a known activity. In some embodiments, the at least one compound includes at least one compound having an unknown activity.

[0013] In some embodiments, the information associated with the compounds category in the database includes, for each compound in the compounds category, at least one compound identifier and at least one structure of the compound, and for at least one known compound in the compounds category also at least one known activity of the known compound and / or at least one known impact of the known compound on a tissue.

[0014] In some embodiments, the database includes information derived from at least one database or information source selected from a chemical structure database, a cosmetics database, a medical database, a natural compounds database, a plant database, an omics databases, clinical trials, scientific literature, blogs, and social media.

[0015] In some embodiments, the machine learning model used by the activity predicting module is based on a model selected from a Quantitative Structure Activity Relationship (QSAR) model, Quantitative Structure Property Relationship (QSPR) model, molecular docking, pharmacophore modeling, Monte Carlo simulations, quantum chemical calculations, time-dependent density functional theory (TDDF), configurational interaction (CI) models, electronic structure computations, molecular mechanics, ligand-receptor interaction prediction, 3D molecular reconstruction, fingerprint extraction methodologies, drug design algorithms, molecular dynamic (MD) simulations, and combinations thereof.

[0016] In some embodiments, the Al-based tissue model used by the impact predicting module is based on a model selected from a Bayesian network, a Graph Neural Network (GNN), a Boolean network, a model based on ODE (Ordinary Differential Equations), a multi-agent system, and combinations thereof. In some embodiments, the Al-based tissue model is based on a Bayesian network, and is trained with transcriptomic data related to the reference tissue. In some embodiments, the Al-based tissue model is a skin model, and is trained with transcriptomic data related to skin.

[0017] In some embodiments, the impact predicting module applies the algorithm utilizing the AI- based tissue model in (c)(ii) in two steps: step (A), in which the algorithm is applied to the predicted activity and predicts the mechanism of action (MOA) of the at least one compound; and step (B), in which the algorithm is applied to the MOA predicted in step (A) to thereby predict the at least one impact on the desired tissue.

[0018] In some embodiments, the impact predicting module further includes an integration module configured to receive identifiers of a plurality of compounds and an activity for each compound of the plurality of compounds from the activity predicting module; and to provide an integrated impact of the plurality of compounds on the desired tissue. In some embodiments, the integration module is further configured to take into account synergistic and / or antagonistic impacts between the plurality of compounds. In some embodiments, the integration module is further configured to calculate synergy or antagonism for a plurality of compounds based on an in silico multicompounds adaptation of the isobole method. In some embodiments, the calculation of synergy via the integration module for a plurality of compounds includes comparing the computed additive impacts based on the impacts predicted for each one of the plurality of compounds separately to the integrated impact predicted for the plurality of compounds.

[0019] In some embodiments, the integration modules is further configured to consider relative amounts of least two compounds of the plurality of compounds when providing the integrated impact.

[0020] In some embodiments, the system further includes a precomputing module configured to provide a plurality of compound identifiers and a desired tissue to the input module; to obtain corresponding predicted impacts on the tissue for each compound of the plurality of compounds; and to save in the database the impacts with a reference to the corresponding compounds for which they were obtained.

[0021] In some embodiments, the system further includes a compound-to-impact module, configured to receive at least one query compound identifier and a desired tissue, and to provide a predicted precalculated activity, MOA, impact, and / or integrated impact for the at least one query compound, based on querying precomputed data in the database.

[0022] In some embodiments, the system further includes a user interface. In some embodiments, the user interface is configured to receive an input including at least one compound identifier and optionally a desired tissue.

[0023] In some embodiments, the system further includes a compound listing module configured to receive as input an identifier of at least one of an organism, a part of an organism, an extract from the organism, a formulation, and combinations thereof; and to provide as output identifiers of at least one compound included in the input, and wherein the database further includes at least one category selected from an organism category and a formulation category.

[0024] In some embodiments, the organism is selected from a plant, am insect, an alga, a fungus, bacterial, and combinations thereof. In some embodiments, the information associated with the organisms category in the database includes, for each organism in the organisms category, at least one organism identifier and at least one compound included in the organism, and optionally at least one organism part, at least one extract of the organism or part thereof, and / or at least one known impact of the organism on a tissue. In some embodiments, the information associated with the formulation category in the database includes, for each formulation in the formulation category, at least one formulation identifier and at least one compound included in the formulation, and optionally at least one known impact of the formulation on a tissue.

[0025] In some embodiments, the system further includes an impact-to-compound module, configured to receive a desired impact and / or MOA and to provide at least one compound, organism or organism part, extract, and / or formulation, for which the predicted impact, activity, or MOA is similar to the desired impact, activity, or MOA. In some embodiments, the desired impact includes ameliorating symptoms of a disease or a condition. In some embodiments, the desired impact includes a scent or a taste.

[0026] In some embodiments, there is provided a computer-implemented Al-based method for predicting at least one impact of at least one compound on a desired tissue, the method including: a) inputting, by an input module, at least one compound identifier of the at least one compound, and optionally a name of a desired tissue; b) providing a database including information associated with at least one category including a compounds category; c) computing, for each compound of the at least one compound inputted by the input module, at least one impact on the desired tissue by: i. applying, via an activity predicting module, to the database and to each one of the at least one compound an algorithm based on a machine learning model and providing a predicted activity of the compound based on a structure of the compound, and ii. applying, via an impact predicting module, to the predicted activity an algorithm utilizing an Al-based tissue model of a reference tissue to predict the at least one impact on the desired tissue, and providing the predicted impact to the output module; and d) outputting, by an output module, the at least one predicted impact of the at least one compound on the desired tissue.

[0027] In some embodiments, the at least one compound is all compounds in the database, or a subset of compounds in the database.

[0028] In some embodiments, the method further includes prior to step (a), inputting the at least one compound identifier to the input module via a user interface.

[0029] In some embodiments, the method further includes prior to step (a), inputting at least one identifier of an organism, a formulation, or a combination thereof to a compound listing module, and providing at least one identifier of a compound included in the organism, formulation, or combination thereof to the input module.

[0030] In some embodiments, the method further includes following step (d), inputting at least one identifier of a query compound into a compound-to-impact module, and obtaining at least one predicted impact for the at least one query compound based on precomputed data in the database. In some embodiments, the method further includes following step (d), inputting a name of a desired impact on the desired tissue to an impact-to-compound module, and obtaining an identifier for at least one compound, organism, formulation, or combination thereof, for which an impact similar to the desired impact was predicted, based on precomputed data in the database.

[0031] In some embodiments, the method is a method further for designing a formulation having a desired impact on a desired tissue, and further includes following step (d): e) inputting at least one identifier of a compound and at least one name of a desired impact; f) receiving at least one predicted impact for the at least one compound on the desired tissue; and g) repeating steps (e) and (f) while slightly changing the compound at each iteration, until the at least one predicted impact is similar to the desired impact

[0032] In some embodiments, the desired impact is a scent.

[0033] In some embodiments, the slightly changing is done by using SMILE.

[0034] In addition to the exemplary aspects and embodiments described above, further aspects and embodiments will become apparent by reference to the figures and by study of the following detailed descriptions.

[0035] BRIEF DESCRIPTION OF DRAWINGS

[0036] The invention will now be described in relation to certain examples and embodiments with reference to the following illustrative figures.

[0037] Fig. 1 presents a schematic representation of the system of the invention, in some embodiments.

[0038] Fig. 2 presents a schematic representation of the system of the invention, in some other embodiments.

[0039] Fig. 3 presents a schematic representation database query options, in some embodiments.

[0040] Fig. 4 shows an example of scent prediction based on scent of known compounds in the database, and similarity of a compound with an unknown scent to compounds having a known scent.

[0041] Fig. 5A-5B explain the evaluation of synergistic effects. Fig. 5A is a schematic graph demonstrating the isobole method, taken from Ebrahimi et al., 2022, Front. Nutr. 9:1047026. Fig. 5B is a schematic diagram of the concept of synergy.

[0042] Fig. 6 presents a screen shot of the system user interface, showing results for synergy.

[0043] Fig. 7 presents a screen shot of the system user interface, showing results for a plant extract query. Fig. 8 presents a screen shot of the system user interface, showing results for a molecular compound query.

[0044] Fig. 9 shows an example for types of information included in the database.

[0045] DETAILED DESCRIPTION OF THE INVENTION

[0046] In the following description, various aspects of the disclosure will be described. For the purpose of explanation, specific configurations and details are set forth in order to provide a thorough understanding of the different aspects of the disclosure. However, it will also be apparent to one skilled in the art that the disclosure may be practiced without specific details being presented herein. Furthermore, well-known features may be omitted or simplified in order not to obscure the disclosure.

[0047] The present invention is directed to an artificial intelligence (Al)-based platform that enables predicting the impacts of products on tissues such as skin, hair, teeth, etc., quickly, safely, and accurately. For example, this platform may be used by cosmetic companies for testing their products in silico prior to testing on people.

[0048] To facilitate this, the Al-based platform utilizes a unique combination of a machine learning model, such as a Quantitative Structure Activity Relationship (QSAR), an Al-based virtual model, of the relevant tissue, constructed for example using a Bayesian network, and a database including information a from a large variety of sources, to predict the impact of various products on the relevant tissue, for example, skin and hair. The ability to predict in silico the impact of a compound, an extract, or a formulation, reduces or eliminates the needs for in vitro and in vivo testing.

[0049] The platform may be used to predict the impact of a variety of compositions including compounds, such as biological organisms (e.g., plants, algae, fungi, etc.), extracts from the organisms, synthetic products, and their combinations. This is carried out by breaking down each composition to the compounds comprised in it, predicting the activity for each compound, and integrating the results to obtain a combined impact of the complete composition.

[0050] As part of the prediction process, the system also predicts synergistic and antagonistic interactions between compounds comprised in a composition.

[0051] Predicting impacts of compounds on a tissue

[0052] In some embodiments, there is provided a computer-implemented Al-based system for predicting at least one impact on a desired tissue by at least one compound, the system including: a) an input module, configured to receive at least one compound identifier of the at least one compound, and optionally a name of a desired tissue; b) a database including information associated with at least one category including a compounds category; c) an Al-based engine configured to compute for each compound of the at least one compound received by the input module at least one impact on the desired tissue by: i. an activity predicting module configured to apply to the database and to each one of the at least one compound an algorithm based on a machine learning model and to provide a predicted activity of the compound based on a structure of the compound, and ii. an impact predicting module configured to apply to the activity predicted by the activity predicting module an algorithm utilizing an Al-based tissue model of a reference tissue to predict the at least one impact on the desired tissue, and to provide the predicted impact to the output module; and d) an output module, configured to provide the at least one predicted impact of the at least one compound on the desired tissue.

[0053] Reference is made to Fig. 1, which shows a schematic representation of system 100, in some embodiments. As seen in Fig. 1, system 100 includes an input module 102, which receives as input at least one compound identifier and optionally a name of a desired tissue, and provides the input to activity predicting module 104, which is part of Al-based engine 105. Activity predicting module 104 applies an algorithm based on a machine learning model to information in database 106 and to each one of the at least one compound, and provides a predicted activity of the compound based on a structure of the compound, which is derived from database 106, to impact predicting module 108. Impact predicting module 108 applies to the activity predicted by the activity predicting module 104 an algorithm utilizing an Al-based virtual tissue model 110 of a reference tissue to predict the at least one impact on the desired tissue, and to provide the predicted impact to the output module 112.

[0054] The term “tissue”, as used herein, is intended to encompass a group of cells having similar structures and functions, or a group of cells which together have a certain function. Accordingly, the definition of a tissue also encompasses groups of cells which together have a certain function (e.g. a skin tissue or a liver tissue) but consist of a variety of cell types, such as keratinocytes, melanocytes, immune cells, fibroblasts, adipocytes, neural cells, etc., which are present in skin. The definition of tissue is further intended to encompass tissues in which the cells are not necessarily in close proximity with each other, such as blood cells or neuronal cells. Systems having a certain function such as sensory systems or perception systems, which may be modelled, are also encompassed by this definition. This definition is further intended to encompass combinations of cells from different organisms which together have a function, and which may be modelled, such as a microbiome, or a combination of gut with gut microbiome, or skin with skin microbiome, etc.

[0055] The term “desired tissue”, as used herein, relates to any tissue for which finding an effect of a compound is desired. This term generally encompasses as nonlimiting examples any mammalian tissue, or more specifically any human tissue, such as, but not limited to, skin, mouth tissues such as gums and teeth, nerve tissue, gastrointestinal tract tissues, mucosal tissues, etc.

[0056] The term “reference tissue”, as used herein, refers specifically to a tissue for which a model is available and is used for the analysis. Generally, the same types of tissues are intended both for the desired tissue and for the reference tissue, including any mammalian tissue, or more specifically any human tissue, such as, but not limited to, skin, mouth tissues such as gums and teeth, nerve tissue, gastrointestinal tract tissues, mucosal tissues, etc.

[0057] It is sometimes useful to use a model of a reference tissue that is similar to the desired tissue because a model of the desired tissue is not available, or in case an impact on the reference tissue also affects the desired tissue (e.g., in the case of skin, which affects hair).

[0058] In some embodiments, the desired tissue is the same as the reference tissue. In some embodiments, the desired tissue is not the same as the reference tissue. When the reference tissue is not the same as the desired tissue, it may be a tissue having similar features (such as pathways, relevant conditions) to the desired tissue, or a tissue which affects the desired tissue (such as skin affecting hair).

[0059] In some embodiments, the desired tissue and / or the reference tissue is a mammalian tissue. In some embodiments, the desired tissue and / or the reference tissue is a human tissue. In some embodiments, the desired tissue and / or the reference tissue is a tissue of a pet or a farm animal, such as a dog, a cat, a fish, a chicken, a horse, or a cow.

[0060] In some embodiments, the desired tissue and / or the reference tissue is selected from skin tissues, hair tissues, mucosal tissues, teeth, gums, digestive tract tissues, gut tissues, neuronal tissues, epithelial tissues, connective tissues, fat tissues, blood system tissues, muscular tissues, liver tissues, lymph tissues, secretory tissues, cancerous tissues, reproductive tissues, sensory tissues and combinations thereof. In some embodiments, the desired tissue and / or the reference tissue is skin or hair tissue.

[0061] In some embodiments, the desired tissue and / or the reference tissue is more than one tissue.

[0062] It is appreciated that entering a name of a tissue is optional, since although the system requires a desired tissue as input, it is understood that the tissue may be predefined and thus a desired tissue may be embedded or encoded in the system rather than provided at each use of the system. It is further understood that “providing a tissue” may also be understood as providing a tissue by the system.

[0063] The term “impact”, as used herein, relates to an outcome, or an effect, obtained by applying the at least one compound to the desired tissue. This term also encompasses a neutral impact, i.e., no effect on the desired tissue is predicted to occur following application of the compound.

[0064] In some embodiments, the impact is selected from a scent, a taste, a color, a texture, and combinations thereof.

[0065] In some embodiments, the impact includes an adverse effect selected from a toxic effect, a phototoxic effect, an endocrine effect, a carcinogenic effect, and combinations thereof.

[0066] In some embodiments, the impact includes a positive, a negative, or a neutral effect on at least one condition related to skin, hair, the digestive tract, and / or a tissue involved with sensory perception (such as smell (olfaction), taste (gustation), and touch (somatosensation)).

[0067] In some embodiments, the impact includes an impact on the microbiome of the tissue.

[0068] In some embodiments, the impact includes an impact on a parasite or a pest such as a fungus, a mite, a cockroach, bacteria, or a mosquito, etc.

[0069] In some embodiments, the impact includes a positive, a negative, or a neutral effect on at least one condition selected from: acne, aging, dandruff, dermatitis, dryness, elasticity, inflammation, UV filtering, perspiration, microbiome profile, oiliness, photoaging, pigmentation, redness, rosacea, scalp-related issues, scarring, seborrhea, sunburns, would healing, wrinkles, alopecia, hair strengthening, hair sharpening, hair growth, gingivitis, breath-freshening, gum health, cavity prevention, tooth whitening, tartar prevention, general well-being, anti-oxidation, a skin disease or condition, and any combination thereof.

[0070] The term “compound”, as used herein, encompasses small molecule compounds, ions, large molecules, e.g., lipids, natural and synthetic polymers, including nucleic acids and proteins. This term also encompasses a group of compounds having a similar structure, and not only specifically- defined single compounds. For example, this term may encompass a basic compound structure (e.g. a backbone) having optional or different substituents.

[0071] In some embodiments, the at least one compound includes a natural compound. In some embodiments, the at least one compound includes a synthetic compound. In some embodiments, the at least one compound includes a small molecule compound. In some embodiments, the at least one compound includes a polymeric compound. In some embodiments, the at least one compound includes a compound having a known activity. In some embodiments, the at least one compound includes a novel compound (e.g., having an unknown activity).

[0072] In some embodiments, the at least one compound is a plurality of compounds. The term “plurality”, as used herein, means at least 2. In some embodiments, this term means at least 3, 4, 5, 10, 20, 50, or 100.

[0073] In some embodiments, the at least one compound includes a combination of any of the above compound definitions.

[0074] The input and output modules (such as input module 102 and output module 112) may be internal system modules (which may receive inputs internally from the system), or modules which receive user input and / or provide user output, respectively, via a user interface. The input module receives as input, and provides to the Al-based engine (such as Al-based engine 105) described in item (c), a single compound or a plurality of compounds, and the output module returns a predicted impact of the compound or the plurality of compounds, as further described below. Accordingly, the system may be, for example, operated by a user inputting, via a user interface, the compound or a plurality of compounds into the input module, and receiving the predicted impact. Alternatively, the system may be operated by a computer software which accesses the input and output modules as part of predicting impacts of a large plurality of compounds not via a user interface. An example for such a process is when the system is accessed by a software for predicting activities for all compounds in the database, or for all compounds which are part of a certain subset of compounds.

[0075] To facilitate calculating activity of compounds by the activity predicting module, database 106 includes information associated at least with compounds, which is a category in the database.

[0076] The term “category”, as used herein, describes different elements in the database, which are associated with certain information types. For example, the compound category is associated with information such as various available names or identifiers (e.g. accessions in various databases) of the compounds, and the chemical structures, or structural features of the compounds. Additional categories are discussed in more details below, for example an organism category, which is associated with organism identifiers, compounds included in the organism, parts of the organism, extracts of the organism or of its parts, etc.

[0077] In some embodiments, the information associated with the compounds category in the database includes, for each compound in the compounds category, at least one compound identifier and at least one chemical structure of the compound, and for at least one known compound in the compounds category also at least one known activity of the known compound and / or at least one known impact of the known compound on a tissue.

[0078] The known activities and / or impacts of some of the compounds in the database facilitate predicting activities and impacts for compounds having a similar structure, and for which this information is not known. In some embodiments, the information associated with the compounds category in the database further includes at least one of alternative compound names or identifiers, chemical properties, associated activities and / or pathways, scent, taste, regulatory information, and results from clinical trials.

[0079] The term “compound identifier”, as used herein, means a name or another identifier, such as a database accession / name (e.g. an International Union of Pure and Applied Chemistry (IUPAC) name or an International Nomenclature of Cosmetic Ingredients (INCI) name, a Simplified Molecular-Input Line-Entry System (SMILES) name), which unambiguously identifies a specific compound, or a group of compounds. The group of compounds may be grouped by a structural similarity.

[0080] The term “chemical structure”, as used herein, means a molecular (chemical) structure of the compound. The structure may be obtained from an appropriate database, such as PubChem, or inferred based on SMILES. In some embodiments, the structure has some ambiguity, as explained above for the definition of a compound, and relates to a group of compounds having a common structure with some variation.

[0081] The database 106 included in the system is in fact an integration of various databases, and may further include information from additional sources, such as scientific literature and public websites.

[0082] The data is aggregated by using a procedure based on molecular fingerprints, which are specific structural features, and similarity scoring. An aggregation algorithm is used to identify the relevant fingerprints (those changes depending on the molecule: lipid, protein, organic compound and so on ...) and uses similarities scoring to determine if the molecules are identical or not.

[0083] The database architecture may be selected from a noSQL architecture (such as MongoDB), a relational architecture (such as SQL), a distributed architecture, an object-oriented architecture, a network architecture and a hierarchical architecture. In some embodiments, the database is SQLbased. In some embodiments, the database is not SQL-based.

[0084] Once data is collected, the data is cleaned. For example, compounds may be compared by preparing a similarity index based on structural information such as that obtained from SMILES by any acceptable methodology. If organisms are further included in the database, they may be compared by comparing taxonomical information. For any information added to the database, a confidence score may be built, in order to assess the reliability of the information, and so whether it should be incorporated into the database. An example for a confidence score is a Bayesian score, computed based on the information sources, number of appearances, etc.

[0085] In some embodiments, the database includes information derived from at least one database or information source selected from a chemical structure database; a cosmetics database; a medical database; a natural compounds database; an organism database, such as a plant database; an omics database (such as microbiome, metabolomics, transcriptomics); clinical trials information; scientific literature; popular databases; blogs, and social media.

[0086] The activity predicting module (such as activity predicting module 104), which is part of the Al-based engine (such as Al-based engine 105), applies an algorithm based on a machine learning model to the database and to each one of the compounds received from the input module. This algorithm provides a prediction for an activity of the compound based on a structure of the compound, that is in the database.

[0087] It is noted that if the compound provided is not present in the database (e.g., a novel compound), the structure used by the algorithm is a structure similar to the structure of the compound, based on SMILES.

[0088] The term “activity”, as used herein, encompasses various properties and activities of the compound. Nonlimiting examples for activities that may be predicted include binding and interaction with biological target molecules, receptor binding, activation or inhibition of a protein (e.g., an enzyme), upregulation or downregulation of a gene, specific effect on a physical property of the tissue, hygroscopic effects, sensory effects, etc.

[0089] In some embodiments, the machine learning model used by the activity predicting module is selected from a classification model and a regression model.

[0090] In some embodiments, the machine learning model is based on a combination of more than one model.

[0091] In some embodiments, the machine learning model is selected from a Quantitative Structure Activity Relationship (QSAR) model, Quantitative Structure Property Relationship (QSPR) model, molecular docking, pharmacophore modeling, Monte Carlo simulations, quantum chemical calculations, time-dependent density functional theory (TDDF), configurational interaction (CI) models, electronic structure computations, molecular mechanics, ligand-receptor interaction prediction, 3D molecular reconstruction, fingerprint extraction methodologies, drug design algorithms, molecular dynamic (MD) simulations, and combinations thereof.

[0092] In some embodiments, the machine learning model includes a Quantitative Structure Activity Relationship (QSAR) model.

[0093] In some embodiments, the implementation of the QSAR model includes an algorithm selected from an XGBoost (Extreme Gradient Boosting) algorithm, random forest algorithm, ANN (Artificial Neural Networks) algorithm, deep learning algorithm, and SVM (Support Vector Machine) algorithm. In some embodiments, molecular features used for the QSAR include physicochemical descriptors, topological descriptors, geometrical descriptors, electronic descriptors, constitutional descriptors, fragment-based descriptors, quantum chemical descriptors, fingerprints, and / or spectral descriptors.

[0094] The impact predicting module (such as impact predicting module 108), which is also part of the Al-based engine (such as Al-based engine 105), applies an Al-based tissue model to the activity predicted by the activity predicting module.

[0095] Optionally, the Al-based tissue model is further applied to the at least one compound.

[0096] In some embodiments, the Al-based tissue model used by the impact predicting module is based on a model selected from a Bayesian network, a Graph Neural Network (GNN), a Boolean network, a model based on ODE (Ordinary Differential Equations), a multi-agent system, and combinations thereof.

[0097] In some embodiments, the Al-based tissue model used by the impact predicting module is based on a Bayesian network.

[0098] In some embodiments, the Al-based tissue model includes nodes which contain properties of the tissue. Nonlimiting examples for properties of a tissue include proteins for which activation or inhibition may affect the tissue; genes for which up / downregulation may affect the tissue; tissue physical properties such as water retention, free radical scavenging, etc.

[0099] In some embodiments, the Al-based tissue model further includes grouping of nodes by a mechanism of action (MOA), based on scientific literasture.

[0100] The term “mechanism of action” or “MOA”, as used herein, encompasses, for example, biochemical / metabolic pathways affected by the compound; anti-microbial activity, inflammatory or anti-inflammatory activity, antioxidant activity, antipruritic activity, apoptosis induction or prevention, cytostatic effect, cell regeneration, hypo or hyperhidrosis, cell penetration, pain reduction, comedogenicity, etc.

[0101] The prediction by the impact predicting module may be done in two steps: in the first step, the tissue model is applied to the activity (or activities) predicted by the activity prediction module in order to predict an MOA, and in the second step, an impact is predicted for the at least one compound based on the MOA.

[0102] In some embodiments, the impact predicting module applies the algorithm in (c)(ii) in two steps: step (A), in which the impact predicting module applies to the activity predicted via the activity predicting module the algorithm utilizing the Al-based tissue model which predicts the MOA of the at least one compound; and step (B), in which the impact predicting module applies to the MOA predicted in step A the algorithm utilizing the Al-based tissue model, to thereby predict the at least one impact on the desired tissue.

[0103] In some embodiments applying the algorithm utilizing the tissue model in step (A) includes checking the predicted activity with respect to nodes of the tissue model to provide a predicted MOA by computing the conditional probability of the MOA based on the predicted value of connected nodes; and applying the algorithm utilizing the tissue model in step (B) further includes computing the conditional probability of a target impact based on both the computed value of the MOA and the values of each node in the tissue model.

[0104] In a nonlimiting example, the tissue model may be built by using a seeded Bayesian approach. The network is then trained using transcriptomic data from the relevant tissue and additional evidence such as metabolomic data.

[0105] In some embodiments, the Al-based tissue model is based on a Bayesian network, and is trained with transcriptomic data related to the reference tissue. In some embodiments, the Al-based tissue model is further trained with metabolomic data related to the reference tissue.

[0106] In some embodiments, the Al-based tissue model is a skin model, and is trained with skin transcriptomic data. In some embodiments, the skin model is further trained with metabolomic data related to skin.

[0107] In some embodiments, the Al-based tissue model is a human skin model, and is trained with human skin transcriptomic data. In some embodiments, the human skin model is further trained with human metabolomic data related to skin.

[0108] In some embodiments, more than one tissue model is used by the system.

[0109] In some embodiments, the impact predicting module further includes an integration module (such as integration module 114) configured to receive identifiers of a plurality of compounds and an activity for each compound of the plurality of compounds from the activity predicting module; and to provide an integrated impact of the plurality of compounds on the desired tissue.

[0110] Measuring combined efficacy of ingredients often involves distinguishing between synergistic and additive impacts, a challenge traditionally tackled in vitro using the isobole approach. For instance, in a scenario involving two ingredients, the impactive concentration (EC50) required to achieve 50% of the maximal impact is determined for each ingredient individually. Using this data, additive impacts are extrapolated through models such as Loewe or Bliss. Subsequently, the EC50 for the combination of these ingredients is measured in the laboratory, and the difference between the observed EC50 and the extrapolated EC50 is calculated to estimate the synergistic impacts.

[0111] In some embodiments, the integration module is further configured to take into account synergistic and / or antagonistic impacts between the plurality of compounds.

[0112] Synergy refers to the phenomenon where the combined impact of multiple components is greater than the sum of their individual impacts. In the context of skincare, oral care, and hair care, synergy can lead to enhanced benefits such as improved hydration, anti-aging impacts, stronger teeth, and healthier hair. However, the vast amount of available compositions is not conducive to identifying such synergistic combinations.

[0113] A “synergistic impact”, as used herein, relates to a predicted impact for at least two compounds that is significantly higher than the additive impact of the same compounds.

[0114] An “antagonistic impact”, as used herein, relates to a predicted impact for at least two compounds that is significantly lower than the additive impact of the same compounds.

[0115] Accordingly, a synergistic impact and / or an antagonistic impact may be calculated by the same method, and synergy or antagonism is defined based on the result. Further explanations of synergy calculation according to the invention are provided in Fig. 5, below.

[0116] In some embodiments, the integration module is further configured to calculate synergy or antagonism for a plurality of compounds based on an in silico multi-compounds adaptation of the isobole method.

[0117] In some embodiments, the calculation of synergy and / or antagonism via the integration module for a plurality of compounds includes comparing the computed additive impacts based on the impacts predicted for each one of the plurality of compounds separately, to the integrated impact predicted for the plurality of compounds.

[0118] In some embodiments, the integration modules is further configured to consider relative amounts of least two compounds of the plurality of compounds when providing the integrated impact. For example, ratios between certain compounds, or percentages of compounds, in the plurality of compounds may further be provided by the input module.

[0119] The system may be used by querying ad hoc, i.e. inputting at least one query compound and computing for the at least one query compound a predicted impact.

[0120] The term “query compound”, as used herein, or the term “query” prepended to other terms such as compound, organism, formulation, etc., is only intended to indicate that the query compound (organism or formulation) is entered into the system as input, e.g., a query, as opposed to originating in the database. However, the scope of “query” terms is generally intended to be the same as the scope of the terms themselves.

[0121] However, the system may also be used by precomputing the impact for compounds in the system, and then querying the database to find at least one impact which matches a query compound. In more detail, the latter is achieved by the precomputing, as described above (without a user input), for all compounds in the system, or for a subset of compounds in the system, predicted activities of the compounds, as well as predicted MOAs, and impacts of the compounds on tissues.

[0122] After the predicted activities, MOAs, and impacts are computed by the system, they may be saved as part of the database, together with a reference to the compounds for which they were predicted.

[0123] Reference is made to Fig. 2, which shows a schematic representation of system 200, in some embodiments. As seen in Fig. 2, system 200, similar to system 100 depicted in Fig. 1, includes an input module 202, which receives as input at least one compound identifier and optionally a name of a desired tissue, and provides the input to activity predicting module 204, which is part of AI- based engine 205. Activity predicting module 204 applies an algorithm based on a machine learning model to information in database 206 and to each one of the at least one compound and provides a predicted activity of the compound based on a structure of the compound, which is derived from database 206, to impact predicting module 208. impact predicting module 208 applies to the activity predicted by the activity predicting module 204 an algorithm utilizing an Al-based tissue model of a reference tissue to predict the at least one impact on the desired tissue, and to provide the predicted impact to the output module 212. The above-mentioned elements are similar to the corresponding elements in Fig. 1, and other elements depicted and explained with reference to Fig. 1 may also apply here (such as the integration module 114).

[0124] As shown in Fig. 2, the system further includes a precomputing module 216, which send lists of input compounds to the input module (e.g., all compounds in the database, or a subset of the compounds), and receives output impacts, which include for each compound a predicted impact from the output module. The impact predicted to for each compound is saved in the database, with a reference to the compound for which it was predicted.

[0125] Additionally, combinations of compounds may be defined, and then the precomputing module may obtain from the output module also integrated impacts for the combinations and save the impacts with a reference to the corresponding combinations. Such combinations may be predefined combinations, or combinations relating to specific categories in the database which are further described below, such as organisms and / or formulations, etc.

[0126] Accordingly, the system may further include a precomputing module, configured to provide a plurality of compound identifiers and a desired tissue (which may also be predetermined in the system) to the input module; to obtain corresponding predicted impacts on the tissue for each compound of the plurality of compounds; and to save in the database the predicted impacts with a reference to the corresponding compounds for which they were obtained. In some embodiments, the plurality of compound identifiers include all compound identifiers in the database. In some embodiments, the plurality of compound identifiers include a subset of compound identifiers in the database. The subset may be predefined, defined by a user, or correspond to another entity in the database, as further discussed below.

[0127] In some embodiments, the precomputing module further defines at least one combination of compounds, retrieves predicted integrated impacts for the at least one combination, and saves the predicted integrated impacts with a reference to the at least one combination for which they were predicted.

[0128] It is appreciated that “by-products” obtained by the system in the process of predicting impacts, such as activities and / or MOAs, and other features of compounds, may also be saved in the database with a reference to the corresponding compounds, for which these activities and / or MOAs were obtained.

[0129] Once the precomputed impacts (and optionally also activities and MOAs) are in the databased, the database may be queried in several different ways, described herein. One of the ways is to input at least one query compound, and retrieve predicted impacts or integrated predicted impacts for the at least one query compound.

[0130] Accordingly, in some embodiments, the system further includes an compound-to-impact module (e.g., Compound-impact-module 307 shown in Fig. 3, described below), configured to receive at least one query compound identifier and a desired tissue, and to provide a predicted precalculated activity, MOA, impact, and / or or integrated impact for the at least one query compound, based on querying precomputed data in the database.

[0131] A nonlimiting example for using the system is for designing a new formulation for obtaining a desired impact on a tissue. The compounds included in the formulation are inputted into the system together with a target tissue, and optionally together with relative quantities / concentrations of the compounds in the formulation, and the system predicts the impact of the formulation on the tissue, taking into account any interactions between the compounds, resulting in synergistic or antagonistic effects. Small changes may consequently be made to the formulation, and the cycle is repeated until a satisfactory result is obtained.

[0132] Other examples for uses may include predicting side-effects and / or toxicity or a new product which includes, for example, a combination of a known compounds, or plant extracts.

[0133] Additionally, impact of novel compounds may be predicted, when they have a structure that is defined (e.g., by SMILES), and similar compounds (e.g., by SMILES) are included in the database.

[0134] In some embodiments, the system further includes a user interface. A user interface is intended to allow a user to provide input into the system, and to receive output from the system.

[0135] Various types of inputs may be supported by the system, including at least one identifier of a compound, or of a plurality of compounds, which is clear from the above description. However, as further described below, the input may include further types of input. Nonlimiting examples for such additional input types includes identifiers of organisms and optionally a names of organism parts or extracts of such organisms, names or identifiers of formulations (e.g., of cosmetic / medical products); as well as a desired impact such a desired effect on a tissue or on a condition related to the tissue, etc. The output may be an impact on a tissue, activity, or MOA; or a compound or plurality of compounds, or organisms, products, etc. Modes of using the system are described below.

[0136] Nonlimiting examples for a user interface include a graphic user interface (GUI) and an application programming interface (API).

[0137] In some embodiments, the user interface is configured to receive an input including at least one compound identifier. In some embodiments, the user interface is further configured to receive a desired tissue. In some embodiments, the user interface is configured to output an activity, MOA, and / or impact on the desired tissue, of the at least one compound.

[0138] The core components of the system, as described above, are capable of receiving a compound, or a plurality of compounds, and a relevant tissue, and providing predicted impacts or the compound or compounds on the tissue.

[0139] As explained above, calculations by the system, e.g., calculating the impact for a compound or a group of compounds, may be conducted following suitable input from a user (or a program), or the predicted impacts may be saved in the database, and then queried by a user (or a program).

[0140] Predicting impacts of an organism or a formulation

[0141] As understood from the above, the system facilitates providing as input at least one compound, i.e., a compound or a plurality of compounds. However, the system further facilitates providing as input also organisms, parts of organisms, extracts obtained from organisms, formulations such as cosmetic / medical products, and basically anything which includes a known list of compounds.

[0142] To this end, the system is capable of “breaking down” the input into the list of compounds included in the input. For example, if the input includes a name of an organism, the system finds the compounds included in the organism, based on information in the database. If the input includes a specific part or parts of the organism, then the system finds the compounds included in the organism parts.

[0143] Accordingly, in some embodiments, the system further includes a compound listing module configured to receive as input an identifier of at least one of an organism, a part of an organism, an extract from the organism, a formulation, and combinations thereof; and to provide as output identifiers of at least one compound included in the input, and wherein the database further includes at least one category selected from an organism category and a formulation category.

[0144] In some embodiments, the input further includes at least one compound.

[0145] It should be clarified that the input to the compound listing module may include combinations of the above, such as a combination of an organism or an extract and a formulation, and may further include an identifier or identifiers of at least one compound.

[0146] In principle, this is done by the information in the database, which is obtained from different sources and relates compounds to organisms, or specific parts of organisms (e.g., leaves or roots of a plant, etc.), or extracts.

[0147] In some embodiments, the organism is selected from any animal from which cosmetic or medical compounds may be obtained. Although most cosmetic or medical products are obtained from plants, alga, and fungi, other animals include compounds which may be used for cosmetics and / or medicine. One example is whale vomit, or ambergris, used in the perfume industry.

[0148] In some embodiments, the organism is selected from a plant, an insect, an alga, a fungus, bacteria, and a combination thereof.

[0149] The term “organism identifier”, or “identifier of an organism”, as used herein, means an identifier of an organism, such as an organism genus, species, or strain.

[0150] The term “formulation”, as used herein, is intended to encompass formulations, such as formulations of known cosmetic products or of known medical products, or new formulations. A formulation may include a list of compounds including the formulation, and optionally their relative amounts, such as ratios or percentages.

[0151] In some embodiments, the compounds listing module is further configured to provide relative amounts, such as ratios or percentages, of the compounds included in the organism / organism part / extract / formulation.

[0152] In some embodiments, the information associated with the organisms category in the database includes, for each organism in the organisms category, at least one organism identifier and at least one compound included in the organism, and optionally at least one organism part, at least one extract of the organism or part thereof, and / or at least one known impact of the organism on a tissue. In some embodiments, the information associated with the organisms category in the database further includes at least one of alternative names / identities, biochemical pathways of the organism or parts thereof; methods of extraction or extracts, and uses of the organism, parts, or extracts thereof.

[0153] In some embodiments, the information associated with the formulation category in the database includes, for each formulation in the formulation category, at least one formulation identifier and at least one compound included in the formulation, and optionally at least one known impact of the formulation on a tissue.

[0154] In some embodiments, the information associated with the formulation category in the database further includes at least one of alternative names / identifiers of products associated with the formulation (such as cosmetic / medical products), methods of its preparation, and uses thereof.

[0155] Additional information in the database regarding any of the categories may include information such as use for various purposes (cosmetic, clinical) including tradition (or popular) use, toxicity information, regulatory information, images (e.g. pictures of organisms or a schematic diagram of a structure), etc., retrieved from various sources including scientific reports, clinical trials, blogs, as well as other sources available on the internet.

[0156] The system may also include a trend prediction module, which is configured to provide existing trends of using a certain compound, extract, or product, or certain names thereof, including prediction of future use and breakdown according to countries. This module is based on aggregated information obtained from various sources including social networks, blog posts, scientific posts and internet searches.

[0157] In some embodiments, the input includes an organism part, and the compound listing module further implements an algorithm which receives the list of compounds included in the organism together with information relates to the compounds, based, e.g., on the SMILES database, and predicts the compound expected to be included in the organism part by computing chemical features based on, e.g., the SMILES information and similarities to other compounds.

[0158] In some embodiments, the input includes an extract of an organism, and the compound listing module further implements an algorithm which receives the list of compounds included in the organism together with information related to the compounds, based, e.g., on the SMILES database, and predicts the compounds expected to be included in the extract, based on calculation of chemical features such as solubility in various solvents, molecular weight, partition coefficients, etc., based on the database information.

[0159] In some embodiments, the output from the compound listing module is provided to the input module of the system. In some embodiments, when the compound listing module provides input including a plurality of compounds to the input module, the impact predicting module implements the integration module in order to provide an integrated impact of the plurality of compounds.

[0160] As explained below, the system may be used by ad hoc querying, or by post hoc querying, i.e., following precomputations of impacts on compounds in the database, as explained above and further below.

[0161] As explained above, the precomputation is conducted by the precomputing module (such as precomputing module 216 in Fig. 2), which may be applied to all compounds in the system, or to a subset of compounds in the database. Subsets may include, for example, compounds of a specific organism or formulation, etc. Additionally, combinations for computing integrated activity may also be defined per organism or part thereof, extract or formulation. As noted above, not only predicted impacts and integrated predicted impacts are saved in the database, but also predicted activities and MO As.

[0162] Accordingly, in some embodiments, the compound-to-impact module (such as compound- to-impact module 307 in Fig. 3), mentioned above, is further configured to receive (in addition to at least one compound identifier) at least one identifier of an query organism, a query formulation, and combinations thereof (or further in combination with a compound), and to provide a predicted precalculated activity, MOA, and / or impact for the compound included. If needed, the compound- to-impact module uses the compound listing module to break down tes organisms and / or formulations into compounds and find matches at the compound level rather than at the organism / formulation level.

[0163] In some embodiments, the user interface is configured to receive an input including an identifier selected from a compound identifier, an organism identifier, a formulation identifier, and combinations thereof. In some embodiments, the input includes an organism, and further includes an organism part and / or extract of the organism or of the organism part. In some embodiments, the user interface is further configured to receive a desired tissue. In some embodiments, the user interface is configured to output an activity, MOA, and / or impact on the desired tissue of compounds included in the input organism or part or extract thereof, or formulation.

[0164] Predicting compounds responsible for a desired impact

[0165] The system further facilitates providing as input a desired outcome, such as an impact on a desired tissue, in order to obtain compounds, or combinations of compounds, as well as organisms or formulations including them, which are predicted to have the desired impact.

[0166] As already explained above, this is facilitated by precomputing the predicted activities, MOAs, and impacts by the precomputing module, and saving them as part of the database, together with a reference to the compounds, organisms, formulations, etc. for which they were predicted.

[0167] Reference is made to Fig. 3, which is a schematic diagram of database 306, which is similar to database 106 and 206 of Fig. 1 and Fig. 2, respectively) and some query options, according to some embodiments. As shown, compound-to-impact module 307 queries the database with at least one query compound identifier (which may also be an query organism, query formulation, etc. as noted above), and receives impacts (which may be integrated impacts, or further include activities and / or MOAs) predicted for the query compounds.

[0168] An alternative option includes using the impact-to-compound module 309 which queries the database including precomputed data with desired impacts (possibly also activities and / or MOAs) and receives at least one compound identifier (which may also be an organism, formulation, etc. as noted above), for which impacts similar to the desired impacts were predicted.

[0169] Accordingly, in some embodiments, the system further includes an impact-to-compound module, configured to receive at least one impact, activity, and / or MOA and to provide at least one compound, organism or organism part, extract, and / or formulation, for which the predicted impact, activity, or MOA is similar to the desired impact, activity, or MOA.

[0170] In some embodiments, the at least one desired impact is at least 2, 3, 4, or 5 desired impacts.

[0171] Therefore, instead of inputting at least one compound, organism or part or extract thereof, formulation, or combination thereof, in order to obtain a predicted impact on a tissue, it is possible to input a desired activity, MOA, and / or impact on a tissue and to obtain at least one compound, organism or part or extract thereof, formulation, or combination thereof, which are predicted to have an activity, MOA, and / or impact similar to the desired activity, MOA, and / or impact.

[0172] To this end, the impact-to-compound module receives as input at least one desired activity, MOA, impact, and / or combinations thereof, compares the input to predicted activities, MOAs, and / or impacts precomputed in the database, and returns at least one compound, organism or part or extract thereof, formulation, or combination thereof, for which the predicted activity, MOA, and / or impact is similar to the desired activity, MOA, and / or impact.

[0173] The term “similar” here is used qualitatively to indicate that a comparison is being made, and a quantitative value thereof may depend on the algorithm used for comparing the input to the precalculated data.

[0174] In some embodiments, the desired impact includes ameliorating symptoms of a disease or a condition. In some embodiments, the desired impact is a scent or a taste.

[0175] In some embodiments, the input includes a desired activity, MOAs, impact, or combinations thereof. In some embodiments, the input includes selecting activities, MOAs, and / or impacts from a predetermined list. In some embodiments, the input further includes inputting desired levels of activities, MOAs, and / or impacts selected from a predetermined list. In some embodiments, the input further includes inputting additional desired features of the compounds, organisms, formulations to be retrieved. Examples for such desired further features include origin (e.g. plant of origin of a compound), country of origine, or other desired physical features, etc.

[0176] This way of using the system may be useful, for example, when designing a new scent, by inputting the desired scent as the impact, and finding combinations of compounds, organisms, and / or formulations, which are predicted to provide the desired scent.

[0177] In some embodiments, the user interface is configured to receive an input including a desired impact on a tissue. In some embodiments, the user interface is configured to receive an input including a desired tissue. In some embodiments, the user interface is configured to output at least one compound, organism or part or extract thereof, formulation, or a combination thereof, for which the desired impact is predicted.

[0178] Finding compounds having similar activities, MOA, or impact as a known compound

[0179] As explained above, the system facilitates, inter alia, providing as input compound, organisms, parts of organisms, extracts obtained from organisms, formulations such as cosmetic / medical products, and basically anything which includes a known list of compounds. The desired output described above for this case is an impact of the input on a tissue.

[0180] However, the information precomputed by the system, as also explained above, includes such impacts, as well as activities and MOAs for compounds in the system, as well as for organisms (including parts and extracts thereof) and formulations including them.

[0181] Accordingly, the system further facilitates inputting at least one compound and obtaining as output “matching” compounds as well as extracts, formulations, and combinations thereof including them, which have similar activities, MOAs, and / or impacts on a desired tissue.

[0182] This may be achieved by modules already described above or by similar modules. For example, input may be provided directly to the input module (if the input is at least one compound), or to the compound listing module (if the input includes organisms or parts or extracts thereof, formulations, etc.), activities, MOAs, and impacts for input compounds in the input are computed, as described above, and the calculated information is used to query the precalculated data in the database by the impact- to-compound module, in order to obtain “matching” compounds, organisms, formulations, and combinations thereof, having predicted activities, MOAs, and / or impacts, which are similar to the input compounds, organisms, formulations, and combinations thereof. A nonlimiting example may be looking for compounds of a natural origin that have the same activity as a known synthetic product or molecule, and that may have fewer side effects.

[0183] In some embodiments, the user interface is configured to receive an input including an identifier selected from at least one compound identifier, an organism identifier, a formulation identifier, and combinations thereof. In some embodiments, the compound is a synthetic compound. In some embodiments, the user interface is further configured to receive a desired tissue. In some embodiments, the user interface is configured to output at least one compound, organism or part or extract thereof, formulation, or a combination thereof, having activities, MOAs, and or impacts on the desired tissue, similar to those of the compounds included in the at least one compound, organism, formulation, and / or combinations thereof of the input.

[0184] It is further appreciated that all intermediate features calculated for the compounds, organism, formulations, and combinations thereof, such as activities, MOAs, synergies and antagonism between compounds, etc., may also be retrieved by a user.

[0185] Methods of using the system of the invention

[0186] In some embodiments, there is provided a computer-implemented artificial intelligence (AI)- based method for predicting at least one impact of at least one compound on a desired tissue, the method including: a) inputting, by an input module, at least one compound identifier of the at least one compound, and optionally a name of a desired tissue; b) providing a database including information associated with at least one category including a compounds category; c) computing, for each compound of the at least one compound inputted by the input module, at least one impact on the desired tissue by: i. applying, via an activity predicting module, to the database and to each one of the at least one compound an algorithm based on a machine learning model and providing a predicted activity of the compound based on a structure of the compound, and ii. applying, via an impact predicting module, to the predicted activity an algorithm utilizing an Al-based tissue model of a reference tissue to predict the at least one impact on the desired tissue, and providing the predicted impact to the output module; and d) outputting, by an output module, the at least one predicted impact of the at least one compound on the desired tissue.

[0187] Definitions and embodiments mentioned above and which may be relevant to the methods also apply here, and vice versa. Some particularly relevant embodiments may be pointed out or explicitly repeated. Specifically, any implementation described above for the system, is intended to also encompass the corresponding method.

[0188] For terms used with reference to the methods, unless stated otherwise, their definition and embodiments are intended to be the same as above (mutatis mutandis).

[0189] It should first be clarified that the steps of inputting and outputting may be conducted by a user (via a user interface, as indicated above), or they may be part of a computer program which calls these steps. For example, as discussed above, many parameters, including activities, MOAs, and impacts, may be precalculated, and therefore, a program may be run without user input to calculate these parameters, e.g., for all compounds in the database (or for a subset of compounds in the database), etc., such as by using the precomputing module.

[0190] In some embodiments, the at least one compound is all compounds in the database. In this case, the program runs without a user interface, to precalculate the parameters for later queries.

[0191] In some embodiments, the method further includes a step, prior to step (a), of inputting the at least one compound identifier to the input module via a user interface. In this case, a user may query the system with at least one identifier of a specific compound, and precomputation is not necessarily needed, but it is still possible that the result is precomputed if the compound is already in the database (e.g., to save time).

[0192] In some embodiments, the method further includes, prior to step (a), inputting at least one identifier of an organism, a formulation, or a combination thereof to a compound listing module, and providing at least one identifier of a compound included in the organism, formulation, or combination thereof to the input module. In some embodiments, the input further includes at least one compound identifier. In some embodiments, the input further includes an organism part of extract name.

[0193] As explained above, the compound listing module breaks down the organism or formulation to the at least one compound (or plurality of compounds) included in the organism or formulation by using information provided in the database to find which compounds are included in organisms or formulations.

[0194] The following post hoc querying should be done after steps (a)-(d) had been conducted for a plurality of compounds, such as all compounds in the database, or a subset thereof, so that predicted impacts, activities, and MOAs are already in the database.

[0195] In some embodiments, the method further includes, following step (d), inputting at least one identifier of a query compound into a compound-to-impact module, and obtaining at least one predicted impact for the compound, based on precomputed data in the database. In some embodiments, the method further includes, following step (d), inputting at least one identifier of a query organism (optionally further including a name of an organism part or of an extract), a query formulation, or a combination thereof into a compound-to-impact module, and obtaining at least one predicted impact for compounds in the query organism, query formulation, or combination thereof, based on precomputed data in the database.

[0196] In some embodiments, the method further includes, following step (d), inputting at least one name of a desired impact into an impact-to-compound module, and obtaining at least one identifier of a compound, organism (optionally further including a name of an organism part or of an extract), formulation, or combination thereof, for which an impact similar to the desired impact was predicted, based on precomputed data in the database.

[0197] The system may also be used for designing a formulation having a desired impact on a desired tissue by repeated cycles wherein a compound is slightly modified in each cycle, such as by, after step (d): inputting at least one identifier of a compound and at least one name of a desired impact (and optionally a name of a desired tissue); receiving at least one predicted impact for the at least one compound on the desired tissue; and repeating the steps while slightly changing the compound structure (e.g. by using SMILES and optionally generative Al) at each iteration, until the at least one predicted impact is similar to the desired impact.

[0198] The term “similar” is intended herein as defined above, mutatis mutandis.

[0199] Corresponding methods and systems not requiring a tissue model

[0200] Some of the specific implementations described below, do not require the use of the tissue model in step (c) of the system and method described above.

[0201] Accordingly, in some embodiments, there is provided a computer-implemented artificial intelligence (Al)-based system for finding compounds having a desired activity based on comparison to known compounds, the system including: a) an input module, configured to receive at least one compound identifier of the at least one compound; b) a database including information associated with at least one category including a compounds category; c) an Al-based engine configured to compute for each compound of the at least one compound received by the input module at least one activity by an activity predicting module configured to apply to the database and to each one of the at least one compound an algorithm based on a machine learning model and to provide a predicted activity of the compound based on a structure of the compound, and d) an output module, configured to provide the at least one predicted activity of the at least one compound.

[0202] Additionally, in some embodiments, there is provided a computer-implemented artificial intelligence (Al)-based method for finding compounds having a desired activity based on comparison to known compounds, the method including: a) inputting, by an input module, at least one compound identifier of the at least one compound; b) providing a database including information associated with at least one category including a compounds category; c) computing, for each compound of the at least one compound inputted by the input module, at least one activity by applying, via an activity predicting module, to the database and to each one of the at least one compound an algorithm based on a machine learning model and providing a predicted activity of the compound based on a structure of the compound, and d) outputting, by an output module, the at least one predicted activity of the at least one compound.

[0203] Definitions and embodiments mentioned above and which may be relevant to the methods and systems in this chapter also apply here, and vice versa (mutatis mutandis). Some particularly relevant embodiments may be pointed out or explicitly repeated. For terms used in this chapter, unless stated otherwise, their definition and embodiments are intended to be the same as above (mutatis mutandis).

[0204] The above method and system generally operates in the same was as described above, configured to accept ad hoc and post hoc queries by using a precomputing module, as well as being configured to accept queries including compounds, organisms (or part or extracts thereof), and formulations, by using a compound listing module and integration module, etc. The main difference in this chapter is in that the method and system do not use a tissue model and therefore predicts activities of compounds rather than impacts on tissues, as described in the previous methods above. Accordingly, the system and method described in this chapter are capable of accepting queries including a desired activity by an activity-to-compound module similar to the impact-to-compound module described above.

[0205] In this chapter, the definition of “activity” further includes elements previously included under the definitions of “impact” or “MO A”, when they can be considered as not affecting a specific tissue, such as, e.g., scent, taste, texture, etc.

[0206] The method and system of the invention may be used for modifications of a formulation, extract, or a compound to achieve a specific activity, such as a specific scent, as shown in Fig. 4).

[0207] As shown on the left of the figure, the database includes various types of experimental data for many compounds, which are analyzed by QSSR (a specific implementation of QSAR, specific to scent as the predicted activity). A query compound having an unknown scent may have its fragrance predicted (both quantitatively and qualitatively) based on structural similarity to at least one compound in the database (e.g. by using SMILES). Additionally, the query compound may further be modified (e.g. by using generative Al), and the effect of the modified compound may further be assessed by the system, until a desired effect is obtained.

[0208] In a similar way, it would be possible to search the database for compounds having an activity similar to currently used compounds, but that is natural, or has fewer side effects.

[0209] Another implementation is finding compounds which act in synergy with a given compound.

[0210] Reference is now made to Fig. 5, which explains how synergism is calculated. Fig. 5A explains the isobole method for defining synergistic and antagonistic behaviour of two compounds (Ebrahimi et al., 2022, Front. Nutr. 9:1047026). The figures shows three types of combination effects (synergistic, antagonistic, and additive). The graphs show the combination of concentrations of the two compounds required to reach a specific impact. As can be seen, when the behavior is synergistic, a lower amount of the compounds is needed compared to an additive behavior, and when the behavior is antagonistic, more of the compounds is needed to achieve the same effect compared to other behaviors.

[0211] Additionally, Fig. 5B shows a schematic representation of the method of predicting synergy according to the invention. The extrapolated additive impact is calculated from the impacts of two compounds, and compared to the integrated impact predicted for both compounds together.

[0212] It is therefore possible to use the system in order to find synergistic combinations for a given compound (or within combinations, organisms, formulations, etc.).

[0213] This may be carried out, for example by querying the database with a certain compound, and obtaining from the database all combinations which include the query compound and for which synergy has been calculated.

[0214] It is also possible to calculate synergy for a given combination of compounds ad hoc, as explained above, by the integration module of the impact predicting module.

[0215] Results for such a query are presented in Fig. 6, which is a screen shot of the system user interface, showing for a compound named borneol, which is present in a mixture of extracts, that synergistic effects are found for several conditions including anti-gingivitis and sensitive gum, and for several MO As including inflammatory reduction and antioxidant properties. Figs. 7-9 demonstrate general features of the system.

[0216] Fig. 7 presents a screen shot of the system user interface, showing example results for a plant extract query. As shown, the information presented includes effects on various parameters of the skin, links to molecules included in the extract, pathways affected by the extract, as well as trends of use over time and place.

[0217] The user can access detailed information regarding the extract, in this example a plant extract but similar information is available when looking for extracts from different organisms such as fungi, bacteria or algae.

[0218] The information may include predicted impacts on a variety of skin and hair conditions ranging from sensitive skin and dandruff to acne and atopic dermatitis; potential adverse impacts, carcinogenic impacts, and / or phototoxic impacts on the skin; a list of molecules that are known to be present in the searched organism or extract, with links to detailed information for each one of these molecules; a list of the predicted MOAs of the plant on the skin; current and predicted trends for the compound or extract, including the evolution of the trend through time, a break-down per region of the world, the different possible names for the extract / compound and the predicted popularity in five years of time for each one of them.

[0219] Fig. 8 presents a screen shot of the system user interface, showing example results for a molecular compound query. As shown, the information presented includes effects on various parameters of the skin, links to plants containing the compound, and pathways affected by the compound.

[0220] The information may include the chemical formula and the predicted chemical structure; predicted impacts of the compound on the skin and hair, including adverse impacts (e.g., a negative score for photoaging as it is predicted to have phototoxic impact); a list of organisms that are known to contain this compound (origins), with links to information on each of those organisms; a list of detailed bioactivities and molecular targets of the compound as well as the predicted MOAs; current and predicted trends of the compound, including breakdown by world region.

[0221] Fig. 9 shows an example for types of information included in the database. As shown such information includes information related to nucleic acids, microbiome, molecular structures, target genes, extracts and cosmetic ingredients, consumer reviews and posts, etc.

[0222] The system of the invention can provide insights into the mechanisms of action. It does this by predicting interactions between biological pathways activated by different ingredients. For example, the system might reveal that two ingredients (e.g. compounds, organisms, extracts, formulations, etc.) activate complementary biological pathways, or that one ingredient inhibits an inhibitor of the targeted pathway of the other. This capability allows for a deeper understanding of how ingredients work together synergistically.

[0223] By providing indications of the mechanisms of action, the system of the invention enables easier validation of synergy without relying solely on isobole analysis. This is particularly advantageous for newly identified complexes, as it provides detailed insights into how the combination works at a biological level.

[0224] Moreover, this methodology allows for the massive screening of thousands of possible combinations to identify the most impactive active ingredient combinations for use in formulations. This large-scale screening capability accelerates the discovery process, making it possible to pinpoint the best synergistic blends efficiently and accurately.

[0225] When selecting an active ingredient for a cosmetics formulation, two main aspects must be considered to ensure an efficient formula: the ingredient's efficacy for the target condition and its ability to combine with other ingredients to create a synergistic impact. Some ingredients are highly impactive on their own but do not necessarily contribute significantly to increasing the overall synergy potential of the formula. Conversely, there are ingredients that may not be the most potent individually but become highly valuable in a blend due to their ability to synergistically enhance the impacts of other ingredients. This ability is defined herein as synergy potential.

[0226] Synergy potential is crucial for creating optimal blends. Being able to predict not only the individual benefits of an ingredient but also its synergy potential allows formulators to design more impactive and innovative products. The Al-based approach predicts the synergy potential of cosmetic ingredients by identifying key biological activities that enhance the action of other compounds on a target condition. This process is done by the system and methods described above.

[0227] By integrating these insights, the system of the invention can predict which ingredients will work best together to enhance overall efficacy. For instance, an ingredient that moderately affects skin hydration might significantly boost the anti-aging impacts of another ingredient by activating complementary biological pathways. This predictive capability enables the creation of formulations that are not only impactive individually but also optimized for synergistic interactions, providing superior results compared to conventional formulations. This predictive capability can also be used to create extracts with strong internal synergistic interactions.

[0228] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention pertains.

[0229] Unless indicated otherwise, the term "a" and "an" refers to one or to more than one (i.e., to at least one, or to one or more) of the grammatical object of the article. By way of example, “an element” means one element or more than one element.

[0230] The term "about" when referring to a measurable value such as an amount, a ratio, and the like, is meant to encompass variations of ±10% of the indicated value, as such variations are also suitable to perform the disclosed invention. Any numerical values appearing in the application are intended to be construed as if preceded by “about”, unless indicated otherwise.

[0231] While certain embodiments of the invention have been illustrated and described, it will be clear that the invention is not limited to the embodiments described herein. Numerous modifications, changes, variations, substitutions, and equivalents will be apparent to those skilled in the art without departing from the spirit and scope of the present invention as described by the claims, which follow.

[0232] The following examples are presented in order to more fully illustrate some embodiments of the invention. They should in no way be construed, however, as limiting the broad scope of the invention. One skilled in the art can readily devise many variations and modifications of the principles disclosed herein without departing from the scope of the invention.

[0233] EXAMPLES

[0234] The below examples provide some proof of concept for the in silico predictions.

[0235] Example 1: Predicting Synergistic Impacts in Green Mountain's Botaniplex™ CALM4

[0236] A study was conducted to evaluate Green Mountain's Botaniplex™ CALM4, which is marketed for its soothing properties. This complex combines four plants from the traditional Chinese pharmacopeia: Scutellaria baicalensis, Rheum palmatum, Cnidium monnieri, and Glycyrrhiza glabra. Using the Al-based platform described herein, a strong synergistic impact for protection against UV damage was predicted, together with a suggested mechanism of action.

[0237] The system identified and quantified the key biological activities of the individual ingredients in the plants, and their combined impacts. The predictions indicated that Calm 4 would exhibit a potent and synergistic antioxidative impact, significantly enhancing the skin's defense against oxidative stress induced by UV radiation. Additionally, the combination was predicted to decrease the secretion of pro -inflammatory factors, thus mitigating inflammation and enhancing soothing properties under UV exposure.

[0238] The Al-driven predictions were subsequently validated through in vitro anti-inflammation experiments. The results showed that Botaniplex™ CALM 4 indeed decreases inflammation by decreasing secretion of pro-inflammatory mediators (IL8 and PGE2) in keratinocytes subjected to UVB (inflammation stimulus), as predicted by the system. Further, it was found that the antioxidative capacity of Botaniplex™ CALM 4 was higher than “anti oxi dative beverages” such as orange juice, green tea, red wine, and coffee, as predicted by the system.

[0239] This validation underscores the robustness of the Al-driven predictions in identifying potent botanical combinations with enhanced antioxidant capabilities. Additionally, the observed inhibition of pro-inflammatory factor secretion under UVB light further validates the efficacy of Calm 4 in mitigating UV-induced inflammatory responses, aligning with the anticipated antiinflammatory properties predicted by the algorithm.

[0240] Specific predictions regarding the four strains in the complex:

[0241] • Rheum palmatum: potential to inhibit cyclophilin D, a protein implicated in mitochondrial dysfunction triggered by UV radiation. This suggests a mechanism for maintaining cellular integrity and resilience against UV-induced stress.

[0242] • Glycyrrhiza glabra: potential inhibitory impacts on nitric oxide synthase (NOS) activity and NOS2 expression. By targeting these pathways, Glycyrrhiza glabra may mitigate UV damage, as nitric oxide can react with UV-generated reactive oxygen species to produce highly reactive radicals.

[0243] • Cnidium monnieri: ability to stimulate histone deacetylase SIRT1 activity, facilitating cellular repair mechanisms. Additionally, it may inhibit NADPH oxidase, thereby reducing the secondary generation of reactive oxygen species following UV exposure.

[0244] • Scutellaria baicalensis: Predicted to inhibit the IL1 pathways, particularly crucial in mitigating UV-induced inflammatory responses. This suggests a potential role in alleviating skin inflammation triggered by UV radiation.

[0245] This case study demonstrates the practical application of the Al-based system in the development and validation of innovative cosmetic formulations.

[0246] Example 2: Prediction of the biological activity of Angelica sinensis

[0247] The biological activity of Angelica sinensis has been predicted, indicating a significant impact on inflammaging for this plant. Inflammaging is characterized by the activation of inflammatory pathways in keratinocytes during the aging process, exacerbated by exposure to UVB radiation.

[0248] The in-vitro showed inhibition of proinflammatory agents (IL8 and PGE2) secretion from UVB -stimulated keratinocytes by 0.05% angelica extract. This validation confirmed the predicted anti-inflammaging impact of Angelica sinensis. This validation underscores the plant's potential as a promising anti-aging agent, particularly in addressing the underlying inflammatory processes associated with skin aging induced by UV radiation.

[0249] Moreover, the Al-based analysis identified several bioactive compounds within Angelica sinensis with remarkable properties:

[0250] • Chlorogenic add (Heriguard): Demonstrated potent anti -pigmentation activity by inhibiting monophenol monooxygenase, while also acting as a scavenger of free radicals and inhibiting lipid peroxidation.

[0251] • Ferulic acid: Not only exhibited anti-inflammatory properties but also protected skin DNA from UV damage. Its regulation of CTGF (Connective Tissue Growth Factor) presents an intriguing avenue for enhancing collagen production, essential for maintaining skin elasticity and firmness.

[0252] • Z-ligustilide: Shown to protect against senescence, a critical aspect in counteracting the aging process.

[0253] • Sucrose: With its potential to induce angiogenesis, offers novel opportunities for skincare innovation, particularly in promoting skin rejuvenation and wound healing processes.

[0254] • Senkyunolide I: Holds strong promise in preventing photo-aging and serving as an apoptosis agonist, potentially delaying cellular aging and enhancing skin resilience against environmental stressors.

[0255] The molecular activity for those bioactives was also confirmed via in vitro testing.

Claims

CLAIMSWhat is claimed is:

1. A computer-implemented artificial intelligence (Al)-based system for predicting at least one impact on a desired tissue by at least one compound, the system comprising: a) an input module, configured to receive at least one compound identifier of the at least one compound, and optionally a name of a desired tissue; b) a database comprising information associated with at least one category comprising a compounds category; c) an Al-based engine configured to compute for each compound of the at least one compound received by the input module at least one impact on the desired tissue by: i. an activity predicting module configured to apply to the database and to each one of the at least one compound an algorithm based on a machine learning model and to provide a predicted activity of the compound based on a structure of the compound, and ii. an impact predicting module configured to apply to the activity predicted by the activity predicting module an algorithm utilizing an Al-based tissue model of a reference tissue to predict the at least one impact on the desired tissue, and to provide the predicted impact to the output module; and d) an output module, configured to provide the at least one predicted impact of the at least one compound on the desired tissue.

2. The Al-based system of claim 1, wherein the desired tissue and / or the reference tissue is a mammalian tissue.

3. The Al-based system of claim 1 or 2, wherein the desired tissue and / or the reference tissue is selected from skin, hair, oral cavity tissues, and combinations thereof.

4. The Al-based system of any one of claims 1-3, wherein the at least one impact is selected from a scent, a taste, a color, a texture, and combinations thereof.

5. The Al-based system of any one of claims 1-4, wherein the at least one impact comprises an adverse effect selected from a toxic effect, a phototoxic effect, an endocrine effect, a carcinogenic effect, and combinations thereof.

6. The Al-based system of any one of claims 1-5, wherein the at least one impact comprises a positive, a negative, or a neutral effect on at least one condition associated with skin, hair, and / or the digestive tract.

7. The Al-based system of any one of claims 1-6, wherein the at least one impact comprises a positive, a negative, or a neutral effect on at least one condition selected from: acne, aging, dandruff, dermatitis, dryness, elasticity, inflammation, UV filtering, perspiration, microbiome profile, oiliness, photoaging, pigmentation, redness, rosacea, scalp-related issues, scarring, seborrhea, sunburns, would healing, wrinkles, alopecia, hair strengthening, hair sharpening, hair growth, gingivitis, breath-freshening, gum health, cavity prevention, tooth whitening, tartar prevention, general well-being, anti-oxidation, a skin disease or condition, and combinations thereof.

8. The Al-based system of any one of claims 1-7, wherein the at least one compound comprises a natural compound.

9. The Al-based system of any one of claims 1-8, wherein the at least one compound comprises a synthetic compound.

10. The Al-based system of any one of claims 1-9, wherein the at least one compound comprises at least one compound having a known activity.

11. The Al-based system of any one of claims 1-10, wherein the at least one compound comprises at least one compound having an unknown activity.

12. The Al-based system of any one of claims 1-11, wherein the information associated with the compounds category in the database comprises, for each compound in the compounds category, at least one compound identifier and at least one structure of the compound, and for at least one known compound in the compounds category also at least one known activity of the known compound and / or at least one known impact of the known compound on a tissue.

13. The Al-based system of any one of claims 1-12, wherein the database comprises information derived from at least one database or information source selected from a chemical structure database, a cosmetics database, a medical database, a natural compounds database, a plant database, an omics databases, clinical trials, scientific literature, blogs, and social media.

14. The Al-based system of any one of claims 1-13, wherein the machine learning model used by the activity predicting module is based on a model selected from a Quantitative StructureActivity Relationship (QSAR) model, Quantitative Structure Property Relationship (QSPR) model, molecular docking, pharmacophore modeling, Monte Carlo simulations, quantum chemical calculations, time-dependent density functional theory (TDDF), configurational interaction (CI) models, electronic structure computations, molecular mechanics, ligandreceptor interaction prediction, 3D molecular reconstruction, fingerprint extraction methodologies, drug design algorithms, molecular dynamic (MD) simulations, and combinations thereof.

15. The Al-based system of any one of claims 1-14, wherein the Al-based tissue model used by the impact predicting module is based on a model selected from a Bayesian network, a Graph Neural Network (GNN), a Boolean network, a model based on ODE (Ordinary Differential Equations), a multi-agent system, and combinations thereof.

16. The Al-based system of claim 15, wherein the Al-based tissue model is based on a Bayesian network, and is trained with transcriptomic data related to the reference tissue.

17. The Al-based system of any one of claims 1-16, wherein the Al-based tissue model is a skin model, and is trained with transcriptomic data related to skin.

18. The Al-based system of any one of claims 1-17, wherein the impact predicting module applies the algorithm utilizing the Al-based tissue model in (c)(ii) in two steps: step (A), in which the algorithm is applied to the predicted activity and predicts the mechanism of action (MOA) of the at least one compound; and step (B), in which the algorithm is applied to the MOA predicted in step (A) to thereby predict the at least one impact on the desired tissue.

19. The Al-based system of any one of claims 1-18, wherein the impact predicting module further comprises an integration module configured to receive identifiers of a plurality of compounds and an activity for each compound of the plurality of compounds from the activity predicting module; and to provide an integrated impact of the plurality of compounds on the desired tissue.

20. The Al-based system of claim 19, wherein the integration module is further configured to take into account synergistic and / or antagonistic impacts between the plurality of compounds.

21. The Al-based system of claim 20, wherein the integration module is further configured to calculate synergy or antagonism for a plurality of compounds based on an in silico multicompounds adaptation of the isobole method.

22. The Al-based system of claim 20 or 21, wherein the calculation of synergy via the integration module for a plurality of compounds comprises comparing the computed additive impacts based on the impacts predicted for each one of the plurality of compounds separately to the integrated impact predicted for the plurality of compounds.

23. The Al-based system of any one of claims 20-22, wherein the integration modules is further configured to consider relative amounts of least two compounds of the plurality of compounds when providing the integrated impact.

24. The Al-based system of any one of claims 1-23, further comprising a precomputing module configured to provide a plurality of compound identifiers and a desired tissue to the input module; to obtain corresponding predicted impacts on the tissue for each compound of the plurality of compounds; and to save in the database the impacts with a reference to the corresponding compounds for which they were obtained.

25. The Al-based system of claim 24, further comprising a compound-to-impact module, configured to receive at least one query compound identifier and a desired tissue, and to provide a predicted precalculated activity, MOA, impact, and / or integrated impact for the at least one query compound, based on querying precomputed data in the database.

26. The Al-based system of any one of claims 1-25, further comprising a user interface.

27. The Al-based system of claim 26, wherein the user interface is configured to receive an input comprising at least one compound identifier and optionally a desired tissue.

28. The Al-based system of any one of claims 1-27, further comprising a compound listing module configured to receive as input an identifier of at least one of an organism, a part of an organism, an extract from the organism, a formulation, and combinations thereof; and to provide as output identifiers of at least one compound included in the input, and wherein the database further comprises at least one category selected from an organism category and a formulation category.

29. The Al-based system of claim 28, wherein the organism is selected from a plant, am insect, an alga, a fungus, bacterial, and combinations thereof.

30. The Al-based system of claim 28 or 29, wherein the information associated with the organisms category in the database comprises, for each organism in the organisms category, at least one organism identifier and at least one compound comprised in the organism, and optionally atleast one organism part, at least one extract of the organism or part thereof, and / or at least one known impact of the organism on a tissue.

31. The Al-based system of any one of claims 28-30, wherein the information associated with the formulation category in the database comprises, for each formulation in the formulation category, at least one formulation identifier and at least one compound comprised in the formulation, and optionally at least one known impact of the formulation on a tissue.

32. The Al-based system of any one of claims 1-31, further comprising an impact-to-compound module, configured to receive a desired impact and / or MOA and to provide at least one compound, organism or organism part, extract, and / or formulation, for which the predicted impact, activity, or MOA is similar to the desired impact, activity, or MOA.

33. The Al-based system of claim 32, wherein the desired impact comprises ameliorating symptoms of a disease or a condition.

34. The Al-based system of claim 32, wherein the desired impact comprises a scent or a taste.

35. A computer-implemented artificial intelligence (Al) -based method for predicting at least one impact of at least one compound on a desired tissue, the method comprising: a) inputting, by an input module, at least one compound identifier of the at least one compound, and optionally a name of a desired tissue; b) providing a database comprising information associated with at least one category comprising a compounds category; c) computing, for each compound of the at least one compound inputted by the input module, at least one impact on the desired tissue by: i. applying, via an activity predicting module, to the database and to each one of the at least one compound an algorithm based on a machine learning model and providing a predicted activity of the compound based on a structure of the compound, and ii. applying, via an impact predicting module, to the predicted activity an algorithm utilizing an Al-based tissue model of a reference tissue to predict the at least one impact on the desired tissue, and providing the predicted impact to the output module; and d) outputting, by an output module, the at least one predicted impact of the at least one compound on the desired tissue.

36. The Al-based method of claim 35, wherein the at least one compound is all compounds in the database, or a subset of compounds in the database.

37. The Al-based method of claim 35, further comprising, prior to step (a), inputting the at least one compound identifier to the input module via a user interface.

38. The Al-based method of claim 35, further comprising, prior to step (a), inputting at least one identifier of an organism, a formulation, or a combination thereof to a compound listing module, and providing at least one identifier of a compound included in the organism, formulation, or combination thereof to the input module.

39. The Al-based method of claim 36, further comprising, following step (d) of the method of claim 35, inputting at least one identifier of a query compound into a compound-to-impact module, and obtaining at least one predicted impact for the at least one query compound based on precomputed data in the database.

40. The Al-based method of claim 36, further comprising, following step (d) of the method of claim 35, inputting a name of a desired impact on the desired tissue to an impact-to-compound module, and obtaining an identifier for at least one compound, organism, formulation, or combination thereof, for which an impact similar to the desired impact was predicted, based on precomputed data in the database.

41. The Al-based method of claim 35, for designing a formulation having a desired impact on a desired tissue, further comprising, following step (d) of claim 37: e) inputting at least one identifier of a compound and at least one name of a desired impact; f) receiving at least one predicted impact for the at least one compound on the desired tissue; and g) repeating steps (e) and (f) while slightly changing the compound at each iteration, until the at least one predicted impact is similar to the desired impact.

42. The method of claim 41, wherein the desired impact is a scent.

43. The method of claim 41 or 42, wherein the slightly changing is done by using SMILE.

Images