Systems and methods for predicting molecular olfactory properties using machine learning

Abstract

Systems and methods for predicting molecular olfactory properties are provided. An example method includes obtaining a machine-learned graph neural network trained to predict an olfactory property of a molecule based at least in part on chemical structure data associated with the molecule. The method includes obtaining a graph that graphically describes a chemical structure of a selected molecule. The method includes providing the graph as input to the machine-learned graph neural network. The method includes receiving, as output of the machine-learned graph neural network, prediction data describing one or more predicted olfactory properties of the selected molecule. The method includes providing the prediction data describing the one or more predicted olfactory properties of the selected molecule as output.

Description

Technical Field

[0001] The present invention relates generally to machine learning and, more particularly, to using machine learning models to predict the olfactory properties of molecules. Background Art

[0002] The relationship between the structure of a molecule and its olfactory perceptual properties (e.g., the molecule's flavor as observed by humans) is complex and, to date, generally poorly understood. For example, the flavor and fragrance industry typically relies on trial and error, heuristics, and / or mining of natural products to provide commercial products with desired olfactory properties. Meaningful principles for organizing the olfactory environment are generally lacking, although it is known that the mapping between molecular structure and flavor can be very nonlinear, so that small changes in a molecule can produce large changes in olfactory quality. Furthermore, the reverse may also be true, and different families of molecules can all smell the same. Summary of the Invention

[0003] Various aspects and advantages of embodiments of the invention will be set forth in part in the following description, or may be learned from the description, or may be learned through practice of the embodiments.

[0004] One exemplary aspect of the present invention is directed to a computer-implemented method for predicting olfactory properties of molecules. The method includes obtaining, by one or more computing devices, a machine-learned graph neural network trained to predict olfactory properties of molecules based at least in part on chemical structure data associated with the molecules. The method includes obtaining, by one or more computing devices, a graph that graphically describes the chemical structure of a selected molecule. The method includes providing, by one or more computing devices, the graph that graphically describes the chemical structure of the selected molecule as input to the machine-learned graph neural network. The method includes receiving, by one or more computing devices, as output of the machine-learned graph neural network, prediction data describing one or more predicted olfactory properties of the selected molecule. The method includes providing, by one or more computing devices, prediction data describing one or more predicted olfactory properties of the selected molecule as output.

[0005] Another exemplary aspect of the present invention is directed to a computing device. The computing device includes one or more processors; and one or more non-transitory computer-readable media that store instructions. When the instructions are executed by the one or more processors, the computing device is caused to perform operations. The operations include: obtaining a machine-learned graph neural network that is trained to predict one or more olfactory properties of a molecule based at least in part on chemical structure data associated with the molecule. The operations include obtaining graph data representing the chemical structure of a selected molecule. The operations include providing the graph data representing the chemical structure as input to the machine-learned graph neural network. The operations include receiving, as output of the machine-learned graph neural network, predicted data describing one or more olfactory properties associated with the selected molecule. The operations include providing, as output, predicted data describing one or more predicted olfactory properties of the selected molecule.

[0006] Other aspects of the present invention are directed to various systems, apparatuses, non-transitory computer-readable media, user interfaces, and electronic devices.

[0007] These and other features, aspects and advantages of various embodiments of the present invention will be better understood with reference to the following description and appended claims.The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate example embodiments of the present invention and, together with the description, serve to explain the relevant principles. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The present invention provides a detailed description of embodiments for persons skilled in the art with reference to the accompanying drawings, in which:

[0009] Figure 1A depicts a block diagram of an example computing system according to an example embodiment of the present invention;

[0010] Figure 1B depicts a block diagram of an example computing device according to an example embodiment of the present invention;

[0011] Figure 1C depicts a block diagram of an example computing device according to an example embodiment of the present invention;

[0012] Figure 2 depicts a block diagram of an example prediction model according to an example embodiment of the present invention;

[0013] Figure 3 depicts a block diagram of an example prediction model according to an example embodiment of the present invention;

[0014] Figure 4 A flowchart depicting example operations for predicting olfactory properties of molecules according to an example embodiment of the present invention; and

[0015] Figure 5Depicted are example diagrams for visualizing structural contributions associated with predicting olfactory properties, according to an example embodiment of the invention.

[0016] Figure 6 An example model diagram and data flow according to an example embodiment of the present invention are shown.

[0017] FIG7 shows the global structure of the embedding space of example learning according to an example embodiment of the present invention.

[0018] Reference numerals repeated across multiple figures are intended to identify like features across the various implementations. DETAILED DESCRIPTION

[0019] Overview

[0020] Example aspects of the present invention are directed to systems and methods that include or otherwise utilize machine learning models (e.g., graph neural networks) in combination with molecular chemical structure data to predict one or more sensory (e.g., smell, taste, touch, etc.) properties of a molecule. Specifically, the systems and methods of the present invention can predict the olfactory properties of a single molecule based on the molecule's chemical structure (e.g., human-perceived odors expressed using labels such as "sweet," "pine," "pear," "rotten," etc.). According to aspects of the present invention, in some implementations, machine learning graph neural networks can be trained and used to process graphs that graphically describe the chemical structure of molecules to predict the olfactory properties of molecules. Specifically, the graph neural network can directly operate on the graph representation of the chemical structure of the molecule (e.g., perform convolution in the graph space) to predict the olfactory properties of the molecule. As an example, the graph can include nodes corresponding to atoms and edges corresponding to chemical bonds between atoms. Therefore, the systems and methods of the present invention can provide predictive data for predicting the odor of previously unevaluated molecules by using machine learning models. A machine learning model can be trained using, for example, training data comprising descriptions of molecules (e.g., structural descriptions of molecules, graph-based descriptions of the chemical structure of molecules, etc.) that have been labeled (e.g., manually by an expert) with descriptions of olfactory properties that have been assessed for the molecules (e.g., textual descriptions of odor categories such as "sweet," "pine," "pear," "rotten," etc.).

[0021] Therefore, aspects of the present invention are directed to using graph neural networks for modeling quantitative structure-odor relationships (QSOR). On a new dataset labeled by olfactory experts, example implementations of the systems and methods described herein significantly outperform previous approaches. Additional analysis demonstrates that the embeddings learned from the graph neural network capture a meaningful representation of the odor space underlying the relationships between structure and odor.

[0022] More specifically, the relationship between the structure of a molecule and its olfactory perceptual properties (e.g., the smell of the molecule as observed by humans) is complex and, to date, poorly understood. Thus, the systems and methods of the present invention provide for the use of deep learning and underutilized data sources to obtain predictions of olfactory perceptual properties of unseen molecules, thereby allowing for improved identification and development of molecules with desired perceptual properties, for example, allowing for the development of new compounds that can be used in commercial flavoring, fragrance, or cosmetic products, improving expertise in predicting the psychoactive effects of drugs from single molecules, and the like. The improved systems for predicting the olfactory perceptual properties of molecules described herein can provide significant improvements in the identification and development of molecules with desired perceptual properties, as well as in the development of new useful compounds.

[0023] More specifically, according to one aspect of the present invention, a machine learning model (such as a graph neural network model) can be trained to provide predictions of perceptual properties of a molecule (e.g., olfactory properties, taste properties, tactile properties) based on an input graph of the molecule's chemical structure. For example, the machine learning model can be provided with an input graph structure of the molecule's chemical structure, such as a standardized description based on the molecule's chemical structure (e.g., a Simplified Molecular Linear Input System (SMILES) string, etc.). The machine learning model can provide an output including a description of the predicted perceptual properties of the molecule, for example, a list of olfactory perceptual properties that describes what the molecule smells like to humans. For example, a SMILES string can be provided, such as the SMILES string "O=C(OCCC(C)C)C" for the chemical structure of isoamyl acetate, and the machine learning model can provide as an output a description of what the molecule smells like to humans, for example, a description of the molecule's odor properties such as "fruit, banana, apple." Specifically, in some embodiments, in response to receiving a SMILES string or other description of a chemical structure, the systems and methods of the present invention can convert the string into a graph structure that graphically describes the two-dimensional structure of the molecule, and can provide the graph structure to a machine learning model that can predict the olfactory properties of the molecule, whether from the graph structure or features derived from the graph structure. In addition to or as an alternative to the two-dimensional graph, the systems and methods can provide for creating a three-dimensional graph representation of the molecule, for example using quantum chemical calculations, for input to the machine learning model.

[0024] In some examples, the prediction can indicate whether the molecule has a specific desired olfactory perception quality (e.g., a target taste perception, etc.). In some embodiments, the prediction data can include one or more types of information associated with the predicted olfactory properties of the molecule. For example, the prediction data for a molecule can provide information for classifying the molecule into an olfactory property class and / or multiple olfactory property classes. In some instances, the class can include a human-provided (e.g., expert) text label (e.g., sour, cherry, pine, etc.). In some instances, the class can include a non-textual representation of the taste / smell, such as a position on a taste continuum, etc. In some instances, the prediction data for a molecule can include an intensity value describing the intensity of the predicted taste / smell. In some instances, the prediction data can include a confidence value associated with the predicted olfactory perception property.

[0025] In addition to or in lieu of specific classifications of molecules, the prediction data can include numerical embeddings that allow for similarity searches, clustering, or other comparisons between two or more molecules based on a measure of distance between two or more embeddings. For example, in some implementations, a machine learning model can be trained to output embeddings that can be used to measure similarity by training the machine learning model using a triplet training scheme, wherein the model is trained to output embeddings that are closer in embedding space for a pair of similar chemical structures (e.g., an anchor example and a positive example), and to output embeddings that are farther in embedding space for a pair of dissimilar chemical structures (e.g., an anchor example and a negative example).

[0026] Therefore, in some implementations, the systems and methods of the present invention may not require the generation of feature vectors describing molecules for input to machine learning models. Instead, the machine learning model can be directly fed with input in the form of graphs of the original chemical structure, thereby reducing the resources required to make predictions about olfactory properties. For example, by providing the graph structure of a molecule as input to a machine learning model, new molecular structures can be conceptualized and evaluated without the need to experimentally generate such molecular structures to determine their perceptual properties, thereby greatly accelerating the ability to evaluate new molecular structures and saving significant resources.

[0027] According to another aspect of the present invention, training data comprising a plurality of known molecules may be obtained to provide training for one or more machine learning models (e.g., graph convolutional neural networks, other types of machine learning models) to provide predictions of molecular olfactory properties. For example, in some embodiments, a machine learning model may be trained using one or more molecular data sets, wherein the data sets include a text description of the chemical structure and perceptual properties of each molecule (e.g., a description of the molecular odor provided by a human expert, etc.). As an example, the training data may be derived from an industry list, such as a list of chemical structures and their corresponding odors for the perfume industry. In some embodiments, due to the fact that some perceptual properties are rare, steps may be taken to balance common perceptual properties and rare perceptual properties when training (multiple) machine learning models.

[0028] According to another aspect of the present invention, in some embodiments, systems and methods can provide an indication of how changes in molecular structure can affect predicted perceptual properties. For example, systems and methods can provide an indication of how changes in molecular structure may affect the intensity of a particular perceptual property, how disastrous a change in the structure of a molecule would be to a desired perceptual quality, and the like. In some embodiments, systems and methods can provide an indication of how one or more atoms and / or groups of atoms are likely to be added and / or removed from the structure of a molecule to determine the effect of such additions / removals on one or more desired perceptual properties. For example, iterations and different changes can be performed on a chemical structure, and the results can then be evaluated to understand how such changes will affect the perceptual properties of the molecule. As another example, the gradient of a classification function of a machine learning model can be evaluated (e.g., with respect to a particular label) at each node and / or edge of an input graph (e.g., via backpropagation through the machine learning model) to generate a sensitivity map (e.g., indicating the importance of each node and / or edge of the input graph to the output of such a particular label). Furthermore, in some implementations, a graph of interest can be obtained, similar graphs can be sampled by adding noise to the graph, and the average of the resulting sensitivity maps for each sampled graph can then be used as the sensitivity map for the graph of interest. Similar techniques can be performed to determine perceptual differences between different molecular structures.

[0029] According to another aspect, the systems and methods of the present invention can be used to explain and / or visualize which aspects of a molecule's structure contribute most to its predicted odor quality. For example, in some embodiments, a heat map can be generated to overlay the molecular structure, which provides an indication of which parts of the molecule's structure are most important to the molecule's perceptual properties and / or which parts of the molecule's structure are less important to the molecule's perceptual properties. In some implementations, data indicating how changes in the molecular structure affect olfactory perception can be used to generate a visualization of how the structure contributes to predicting olfactory quality. For example, as described above, iterative changes in the structure of a molecule (e.g., knock-down techniques, etc.) and their corresponding outputs can be used to evaluate which parts of a chemical structure contribute most to olfactory perception. As another example, as described above, gradient techniques can be used to generate a sensitivity map of a chemical structure, which can then be used to produce a visualization (e.g., in the form of a heat map).

[0030] According to another aspect of the invention, in some embodiments, (multiple) machine learning models can be trained to generate predictions of molecular chemical structures that will provide one or more desired perceptual properties (e.g., generating molecular chemical structures that will produce a specific taste quality, etc.). For example, in some implementations, an iterative search can be performed to identify (multiple) proposed molecules that are predicted to exhibit one or more desired perceptual properties (e.g., a target taste quality, intensity, etc.). For example, the iterative search can propose many candidate molecular chemical structures that can be evaluated by (multiple) machine learning models. In one example, the candidate molecular structures can be generated by an evolutionary or genetic process. As another example, the candidate molecular structures can be generated by a reinforcement learning agent (e.g., a recurrent neural network) that seeks to learn a strategy that maximizes a reward that is a function of whether the generated candidate molecular structures exhibit one or more desired perceptual properties.

[0031] Therefore, in some implementations, multiple candidate molecule graph structures describing the chemical structure of each candidate molecule can be generated (e.g., iteratively generated) and used as input to the machine learning model. The graph structure of each candidate molecule can be input to the machine learning model for evaluation. The machine learning model can generate prediction data for each candidate molecule, which describes one or more perceptual properties of the candidate molecule. The candidate molecule prediction data can then be compared with one or more desired perceptual properties to determine whether the candidate molecule will exhibit the desired perceptual properties (e.g., a feasible molecule candidate, etc.). For example, a comparison can be performed to generate a reward (e.g., in a reinforcement learning scheme) or to determine whether to retain or discard a candidate molecule (e.g., in an evolutionary learning scheme). A brute force search method can also be used. In a further implementation, it may or may not have the above-mentioned evolutionary or reinforcement learning structure, and the search for candidate molecules that exhibit one or more desired perceptual properties can be constructed as a constrained multi-parameter optimization problem based on the optimization defined for each desired property.

[0032] According to another aspect of the present invention, systems and methods can provide for predicting, identifying, and / or optimizing other properties associated with molecular structures as well as desired olfactory properties. For example, (multiple) machine learning models can predict or identify properties of molecular structures such as optical properties (e.g., transparency, reflectivity, color, etc.), taste properties (e.g., tastes like "banana," "sour," "spicy," etc.), shelf stability, stability at specific pH levels, biodegradability, toxicity, industrial applicability, etc.

[0033] According to another aspect of the present invention, the machine learning model described herein can be used in active learning technology to narrow down a wide range of candidate fields into smaller sets of molecules, which are then manually evaluated. According to other aspects of the present invention, systems and methods can allow for the synthesis of molecules with specific properties in an iterative design-test-refinement process. For example, based on the predicted data from the machine learning model, molecules can be proposed for development. These molecules can then be synthesized and then subjected to specialized testing. Feedback from the testing can then be provided back to the design phase to refine the molecules, thereby better achieving desired properties, etc.

[0034] The system and method of the present invention provide many technical effects and benefits. As an example, the system and method described herein can allow to reduce the time and resources required to determine whether a molecule will provide the desired perceived quality. For example, the system and method described herein allow to use a graph structure describing the chemical structure of the molecule, rather than having to generate a feature vector describing the molecule to provide model input. Therefore, the system and method provide technical improvements in obtaining and analyzing model input and generating model prediction output required resources. In addition, using a machine learning model to predict olfactory properties represents the integration of machine learning into practical applications (e.g., predicting olfactory properties). That is, the machine learning model is suitable for predicting the specific technical implementation of olfactory properties.

[0035] Referring now to the accompanying drawings, example embodiments of the present invention will be discussed in further detail.

[0036] Example devices and systems

[0037] Figure 1A A block diagram of an example computing system 100 is depicted that can facilitate predicting sensory properties of molecules, such as olfactory sensory properties, according to an example embodiment of the present invention. System 100 is provided as an example only. Other computing systems including different components can be used in addition to or in place of system 100. System 100 includes a user computing device 102, a server computing system 130, and a training computing system 150, which are communicatively coupled via a network 180.

[0038] The user computing device 102 may be any type of computing device, such as, for example, a personal computing device (e.g., a laptop or desktop), a mobile computing device (e.g., a smartphone or tablet), a gaming console or controller, a wearable computing device, an embedded computing device, or any other type of computing device.

[0039] The user computing device 102 includes one or more processors 112 and memory 114. The one or more processors 112 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one or more processors operatively connected. The memory 114 can include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 114 can store data 116 and instructions 118 that are executed by the processor 112 to cause the user computing device 102 to perform operations.

[0040] In some implementations, the user computing device 102 can store or include one or more machine learning models 120, such as the olfactory characteristic prediction machine learning models discussed herein. For example, the machine learning model 120 can be or can include a variety of machine learning models, such as a neural network (e.g., a deep neural network) or other types of machine learning models, including nonlinear models and / or linear models. The neural network can include a feedforward neural network, a recurrent neural network (e.g., a long short-term memory recurrent neural network), a convolutional neural network, or other forms of neural networks. Figure 2 and Figure 3 An example machine learning model 120 is discussed.

[0041] In some implementations, one or more machine-learned models 120 can be received from the server computing system 130 over the network 180, stored in the user computing device memory 114, and then used or otherwise implemented by the one or more processors 112. In some implementations, the user computing device 102 can implement multiple parallel instances of a single machine-learned model 120.

[0042] Additionally or alternatively, one or more machine-learned models 140 may be included in, or stored and implemented by, a server computing system 130 that communicates with the user computing device 102 according to a client-server relationship. For example, the machine-learned models 140 may be implemented as part of a web service by the server computing system 140. Thus, one or more models 120 may be stored and implemented on the user computing device 102 and / or one or more models 140 may be stored and implemented on the server computing system 130.

[0043] The user computing device 102 may also include one or more user input components 122 for receiving user input. For example, the user input component 122 may be a touch-sensitive component (e.g., a touch-sensitive display or touchpad) that is sensitive to the touch of a user input object (e.g., a finger or a stylus). The touch-sensitive component may be used to implement a virtual keyboard. Other example user input components include a microphone, a traditional keyboard, a camera, or other means by which a user can provide user input.

[0044] The server computing system 130 includes one or more processors 132 and memory 134. The one or more processors 132 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller), and can be one processor or multiple processors operatively connected. The memory 134 can include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 134 can store data 136 and instructions 138 that are executed by the processor 132 to cause the server computing system 130 to perform operations.

[0045] In some implementations, server computing system 130 includes or is otherwise implemented by one or more server computing devices. In instances where server computing system 130 includes multiple server computing devices, such server computing devices may operate according to a sequential computing architecture, a parallel computing architecture, or some combination thereof.

[0046] As described above, the server computing system 130 may store or otherwise include one or more machine learning models 140. For example, the model 140 may be or may otherwise include a variety of machine learning models, such as a machine learning model for predicting olfactory characteristics. Example machine learning models include neural networks or other multi-layer nonlinear models. Example neural networks include feedforward neural networks, deep neural networks, recurrent neural networks, and convolutional neural networks. Figures 2 to 4 Discuss example model 140.

[0047] User computing device 102 and / or server computing system 130 may train models 120 and / or 140 via interaction with training computing system 150 communicatively coupled via network 180. Training computing system 150 may be separate from server computing system 130 or may be part of server computing system 130.

[0048] The training computing system 150 includes one or more processors 152 and memory 154. The one or more processors 152 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.), and can be one processor or multiple processors that are operably connected. The memory 154 can include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 154 can store data 156 and instructions 158 that are executed by the processor 152 to cause the training computing system 150 to perform operations. In some implementations, the training computing system 150 includes or is otherwise implemented by one or more server computing devices.

[0049] The training computing system 150 may include a model trainer 160 that trains the machine learning models 120 and / or 140 stored on the user computing device 102 and / or the server computing system 130 using a variety of training or learning techniques (such as, for example, back propagation of errors). In some implementations, performing back propagation of errors may include performing truncated time-based back propagation. The model trainer 160 may perform a number of generalization techniques (e.g., weight decay, dropout, etc.) to improve the generalization ability of the trained model.

[0050] In particular, model trainer 160 can train machine-learned models 120 and / or 140 based on a set of training data 162. Training data 162 can include, for example, molecular descriptions (e.g., graphical descriptions of the chemical structure of a molecule) that have been labeled (e.g., manually by an expert) with descriptions of olfactory properties that have been assessed for the molecule (e.g., textual descriptions of odor categories such as "sweet," "pine," "pear," "rotten," etc.).

[0051] The model trainer 160 includes computer logic that is utilized to provide the desired functionality. The model trainer 160 can be implemented in hardware, firmware, and / or software that controls a general-purpose processor. For example, in some implementations, the model trainer 160 includes a program file stored on a storage device, loaded into memory, and executed by one or more processors. In other implementations, the model trainer 160 includes one or more sets of computer-executable instructions that are stored in a tangible computer-readable storage medium such as a RAM hard disk or optical or magnetic media.

[0052] Network 180 can be any type of communication network, such as a local area network (e.g., an intranet), a wide area network (e.g., the Internet), or some combination thereof, and can include any number of wired or wireless links. In general, communications over network 180 can be carried via any type of wired and / or wireless connection, using a variety of communication protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), and / or protection schemes (e.g., VPN, secure HTTP, SSL).

[0053] Figure 1A An example computing system that can be used to implement the present invention is shown. Other computing systems may also be used. For example, in some implementations, user computing device 102 may include model trainer 160 and training data 162. In such implementations, model 120 may be trained and used locally on user computing device 102. Any component shown as included in one of device 102, system 130, and / or system 150 may alternatively be included in the other one or both of device 102, system 130, and / or system 150.

[0054] Figure 1B Depicted is a block diagram of an example computing device 10 according to an example embodiment of the present invention. Computing device 10 may be a user computing device or a server computing device.

[0055] Computing device 10 includes a number of applications (e.g., applications 1 through N). Each application includes its own machine learning library and (multiple) machine learning models. For example, each application may include a machine learning model. Example applications include text messaging applications, email applications, dictation applications, virtual keyboard applications, browser applications, and the like.

[0056] like Figure 1B As shown, each application can communicate with many other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In some implementations, each application can use an API (e.g., a public API) to communicate with each device component. In some implementations, the API used by each application is specific to that application.

[0057] Figure 1C Depicted is a block diagram of an example computing device 50 according to an example embodiment of the present invention. Computing device 50 may be a user computing device or a server computing device.

[0058] The computing device 50 includes a number of applications (e.g., applications 1 through N). Each application communicates with a central intelligence layer. Example applications include a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, and the like. In some implementations, each application can communicate with the central intelligence layer (and the model(s) stored therein) using an API (e.g., a public API across all applications).

[0059] The central intelligence layer includes many machine learning models. For example, Figure 1C As shown, a corresponding machine learning model (e.g., model) can be provided for each application and managed by the central intelligence layer. In other implementations, two or more applications can share a single machine learning model. For example, in some implementations, the central intelligence layer can provide a single model (e.g., a single model) for all applications. In some implementations, the central intelligence layer is included in the operating system of the computing device 50 or is otherwise implemented by the operating system of the computing device 50.

[0060] The central intelligence layer can communicate with the central device data layer. The central device data layer can be a centralized data repository for computing devices 50. Figure 1C As shown, the central device data layer can communicate with many other components of the computing device, such as, for example, one or more sensors, context managers, device state components, and / or additional components. In some implementations, the central device data layer can communicate with each device component using an API (e.g., a private API).

[0061] Example model layout

[0062] Figure 2 A block diagram of an example prediction model 202 according to an example embodiment of the present invention is depicted. In some implementations, the prediction model 202 is trained to receive a set of input data 204 (e.g., molecular chemical structure image data, etc.) and, as a result of receiving the input data 204, provide output data 206, e.g., prediction data of olfactory properties of molecules.

[0063] Figure 3 A block diagram of an example machine-learned model 202 is depicted according to an example embodiment of the present invention. The machine-learned model 202 is similar to Figure 2 The prediction model 202, in addition to Figure 3The machine-learned model 202 is an example model that includes an olfactory property prediction model 302 and a molecular structure optimization prediction model 306. In some implementations, the machine-learned prediction model 202 may include an olfactory property prediction model 302 that predicts one or more olfactory perceptual properties of a molecule based on its chemical structure (e.g., provided in the form of a graph structure); and a molecular structure optimization prediction model 306 that predicts how changes in the molecular structure can affect the predicted perceptual properties. Thus, the model may provide output that includes both the olfactory perceptual properties and how the molecular structure affects those predicted olfactory properties.

[0064] Example Method

[0065] Figure 4 A flow chart of an example method 400 for predicting olfactory characteristics according to an example embodiment of the present invention is depicted. Although for purposes of illustration and discussion, Figure 4 The steps are depicted as being performed in a particular order, but the method of the present invention is not limited to the particular order or arrangement shown. The various steps of method 400 may be omitted, rearranged, combined, and / or adjusted in various ways without departing from the scope of the present invention. Method 400 may be implemented by one or more computing devices, such as Figure 1A-1C One or more computing devices depicted in .

[0066] At 402, method 400 may include obtaining, by one or more computing devices, a machine-learned graph neural network trained to predict olfactory properties of molecules based at least in part on chemical structure data associated with the molecules. Specifically, a machine-learned prediction model (e.g., a graph neural network, etc.) may be trained and used to process a graph that graphically describes the chemical structure of a molecule to predict the olfactory properties of the molecule. For example, the trained graph neural network may operate directly on the graph representation of the molecule's chemical structure (e.g., performing convolution in the graph space) to predict the olfactory properties of the molecule. The machine-learned model may be trained using training data that includes descriptions of molecules (e.g., graphical descriptions of the molecule's chemical structure, etc.) that have been labeled (e.g., manually by an expert) with descriptions of olfactory properties that have been assessed for the molecule (e.g., textual descriptions of odor categories such as "sweet," "pine," "pear," "rotten," etc.). The trained machine-learned prediction model may provide prediction data that predicts the odor of previously unassessed molecules.

[0067] More specifically, most machine learning models require regularly shaped inputs (e.g., pixel grids or vectors of numbers) as input. However, GNNs enable the use of irregularly shaped inputs, such as graphs, to be directly used in machine learning applications. Therefore, according to aspects of the present invention, molecules can be interpreted as graphs by viewing atoms as nodes and bonds as edges. An example GNN is a learnable permutation-invariant transformation on nodes and edges that produces a fixed-length vector that is further processed by a fully connected neural network. Compared to general features carefully designed by experts, GNNs can be considered as learnable featurizers specifically for a task.

[0068] Some example GNNSs include one or more message passing layers, each followed by a reduce-sum operation, followed by several fully connected layers. The example final fully connected layer has a number of outputs equal to the number of predicted odor descriptors. Figure 6 An example model is shown in , which shows an example model schematic and data flow. Figure 6 In the example shown, each molecule is first characterized by its constituent atoms, bonds, and connectivity. Each graph neural network (GNN) layer transforms the features from the previous layer. The output from the final GNN layer is reduced to a vector, which is then used to predict odor descriptors via a fully connected neural network. In some example implementations, the graph embedding can be retrieved from the penultimate layer of the model. An example of the embedding space representation of four odor descriptors is shown in the lower right corner.

[0069] Reference again Figure 4 At 404, method 400 may include obtaining, by one or more computing devices, a graph that graphically describes the chemical structure of the selected molecule. For example, an input graph structure of the chemical structure of a molecule (e.g., a previously unevaluated molecule, etc.) may be obtained for predicting one or more sensory (e.g., olfactory) properties of the molecule. For example, in some embodiments, the graph structure may be obtained based on a standardized description of the chemical structure of the molecule, such as a Simplified Molecular Linear Input System (SMILES) string. In some embodiments, in response to receiving the SMILES string or other description of the chemical structure, the one or more computing devices may convert the string into a graph structure that graphically describes the two-dimensional structure of the molecule. Additionally or alternatively, the one or more computing devices may provide for creating a three-dimensional representation of the molecule, such as using quantum chemical calculations, to input into a machine learning model.

[0070] At 406, method 400 may include providing, by one or more computing devices, a graph describing the chemical structure of the selected molecule as input to a graph neural network for machine learning. For example, the graph structure describing the chemical structure of the molecule obtained at 404 may be provided to a machine learning model (e.g., a trained graph convolutional neural network and / or other type of machine learning model), which may predict the olfactory properties of the molecule from the graph structure or features derived from the graph structure.

[0071] At 408, method 400 may include receiving, by one or more computing devices, prediction data describing one or more predicted olfactory properties of the selected molecule as an output of the machine-learned graph neural network. Specifically, the machine-learned model may provide output prediction data comprising a description of the predicted perceptual properties of the molecule, such as, for example, a list of olfactory perceptual properties that describe what the molecule smells like to a human. For example, a SMILES string may be provided, such as the SMILES string "O=C(OCCC(C)C)C" for the chemical structure of isoamyl acetate, and the machine-learned model may provide as output a description of what the molecule smells like to a human (e.g., a description of the molecule's odor properties), such as "fruit, banana, apple."

[0072] In some example embodiments, the prediction data may indicate whether the molecule has a specific desired olfactory perception quality (e.g., target taste perception, etc.). In some example embodiments, the prediction data may include one or more types of information associated with the predicted molecule's olfactory properties. For example, the prediction data for a molecule may provide information for classifying the molecule into an olfactory property class and / or multiple olfactory property classes. In some instances, the class may include a human-provided (e.g., expert) text label (e.g., sour, cherry, pine, etc.). In some instances, the class may include a non-text representation of the taste / smell, such as a position on a taste continuum, etc. In some example embodiments, the prediction data for a molecule may include an intensity value describing the intensity of the predicted taste / smell. In some example embodiments, the prediction data may include a confidence value associated with the predicted olfactory perception property. In some example embodiments, in addition to or in lieu of the specific classification of the molecule, the prediction data may include a numerical embedding that allows similarity searches or other comparisons to be performed between two molecules based on a measure of the distance between the two embeddings.

[0073] At 410 , method 400 may include providing, by one or more computing devices, as output, prediction data describing one or more predicted olfactory properties of the selected molecule.

[0074] Figure 5 Depicted is an example diagram for visualizing structural contributions associated with predicting olfactory properties, according to an example embodiment of the present invention. Figure 5As shown, in some embodiments, the systems and methods of the present invention can provide output data to facilitate interpretation and / or visualization of which aspects of a molecule's structure contribute most to its predicted odor quality. For example, in some embodiments, a heat map can be generated to overlay the molecular structure, such as visualizations 502, 510, and 520, that provides an indication of which parts of the molecule's structure are most important and / or which parts of the molecule's structure are less important to the molecule's perceived properties. As an example, a heat map visualization (such as visualization 502) can provide an indication that atom / bond 504 is likely to be most important for predicting the perceived property, atom / bond 506 is likely to be moderately important for predicting the perceived property, and atom / bond 508 is likely to be less important for predicting the perceived property. In another example, visualization 510 can provide an indication that atom / bond 512 is likely to be most important for predicting the perceived property, atom / bond 514 is likely to be moderately important for predicting the perceived property, and atoms / bonds 516 and 518 are likely to be less important for predicting the perceived property. In some implementations, data indicating how changes in the molecular structure affect olfactory perception can be used to generate a visualization of how the structure contributes to predicting olfactory quality. For example, iterative changes in the structure of a molecule (e.g., knockdown techniques, etc.) and their corresponding outputs can be used to evaluate which parts of the chemical structure contribute most to olfactory perception.

[0075] Graph Neural Network Embeddings for Example Learning

[0076] Some of the example neural network architectures described herein can be configured to construct representations of input data in their intermediate layers. The success of deep neural networks in prediction tasks relies on the quality of their learned representations (often called embeddings). The structure of the learned embeddings can even lead to insights into the task or problem domain, and the embeddings themselves can even become objects of study.

[0077] Some example computing systems can save the activations of the penultimate fully connected layer as a fixed-dimensional "odor embedding." GNN models can transform the graph structure of molecules into a fixed-length representation useful for classification. The learned GNN embeddings for odor prediction tasks may include semantically meaningful and useful organization of odor molecules.

[0078] Odor embeddings that reflect commonsense relationships between odors should exhibit both global and local structure. Specifically, for global structure, perceptually similar odors should be nearby in the embedding. For local structure, individual molecules with similar odor perceptions should be clustered together and thus nearby in the embedding.

[0079] The example embedding representation of each data point can be generated from the penultimate layer output of the GNN model trained on the example. For example, each molecule can be mapped to a 63-dimensional vector. Qualitatively, in order to visualize this space in 2D, principal component analysis (PCA) can optionally be used to reduce its dimensionality. Using kernel density estimation (KDE) can highlight the distribution of all molecules sharing similar labels.

[0080] An example global structure of the embedding space is shown in Figure 7. In this example, we found that individual odor descriptors (such as musk, cabbage, lily, and grape) tend to cluster within their respective specific ranges. For odor descriptors that frequently co-occur, we found that the embedding space captures the implicit hierarchical structure within the odor descriptors. The clusters of odor labels jasmine, lavender, and lily of the valley are found within the cluster of the broader odor label floral.

[0081] Figure 7 shows a 2D representation of the GNN model embedding as the learned odor space. Molecules are represented as single points. The shaded and outlined regions are kernel density estimates of the labeled data distribution. A. Four odor descriptors with low co-occurrence have low overlap in the embedding space. B. Three general odor descriptors (floral, meaty, alcohol), each of which contains more specific labels within its boundaries. Example experiments show that the generated embeddings can be used to retrieve molecules that are perceptually similar to the source molecules (e.g., using nearest neighbor search on the embeddings).

[0082] Transfer Learning with Examples

[0083] Odor descriptors may be newly invented or refined (e.g., a molecule with a pear descriptor may later be attributed to the more specific pear skin, pear stem, pear flesh, pear core descriptors). A useful odor embedding would be able to perform transfer learning to this new descriptor using only limited data. To simulate this situation, the example experiment removes one odor descriptor at a time from the dataset. Using the embeddings trained from (N-1) odor descriptors as featurization, a random forest is trained to predict the previously held-out odor descriptors. We use cFP and Mordred features as baselines for comparison. In this task, the GNN embedding significantly outperforms Morgan fingerprints and Mordred features, but as expected, it still performs slightly worse than the GNN trained on the target odor. This shows that GNN-based embeddings can generalize to predict new but related odors.

[0084] In another example, the proposed QSOR modeling approach can generalize to adjacent perception tasks and capture meaningful and useful structures about human olfactory perception, even when measured using different methods in different contexts.

[0085] Additional inventions

[0086] The technology discussed herein relates to servers, databases, software applications, and other computer-based systems, as well as the actions taken and information sent to and from such systems. The inherent flexibility of computer-based systems allows for a variety of possible configurations, combinations, and partitioning of tasks and functions between and among components. For example, the processing discussed herein can be implemented using a single device or component or multiple devices or components working in combination. Databases and applications can be implemented on a single system or distributed across multiple systems. Distributed components can operate sequentially or in parallel.

[0087] Although the present subject matter has been described in detail with respect to a variety of specific example embodiments thereof, each example is provided by way of explanation, not limitation, of the present invention. Those skilled in the art, upon understanding the foregoing, can readily produce substitutions, variations, and equivalents to these embodiments. Therefore, the present subject matter invention does not exclude modifications, variations, and / or additions to the present subject matter that would be apparent to one of ordinary skill in the art. For example, a feature shown or described as part of one embodiment may be used in conjunction with another embodiment to yield a further embodiment. Therefore, the present invention is intended to encompass such substitutions, variations, and equivalents.

Claims

1. A computer-implemented method, comprising: Obtaining, by one or more computing devices, a machine-learned graph neural network trained to predict olfactory properties of molecules based at least in part on chemical structure data associated with the molecules; obtaining, by the one or more computing devices, a diagram graphically depicting a chemical structure of the selected molecule; providing, by the one or more computing devices, the graph describing the chemical structure of the selected molecule in a graphical manner as input to a graph neural network for machine learning; receiving, by the one or more computing devices, prediction data describing one or more predicted olfactory properties of the selected molecule as an output of a machine-learned graph neural network, wherein the prediction data indicative of the one or more predicted olfactory properties of the selected molecule comprises a numerical embedding; providing, by the one or more computing devices, as output, prediction data describing the one or more predicted olfactory properties of the selected molecule; and The one or more computing devices identify other molecules having olfactory properties similar to the predicted olfactory properties of the selected molecule by comparing the numerical embeddings with other numerical embeddings output by the machine-learned graph neural network for the other molecules.

2. The computer-implemented method of claim 1 , wherein: Obtaining, by the one or more computing devices, a graph neural network for machine learning includes: obtaining, by the one or more computing devices, training data comprising a plurality of example chemical structures, each example chemical structure being labeled with one or more olfactory property labels describing an olfactory property of the example chemical structure; and A machine-learned graph neural network is trained, by the one or more computing devices, to predict olfactory properties of molecules based in part on the obtained training data.

3. The computer-implemented method of claim 1 , further comprising: generating, by the one or more computing devices, visualization data depicting the relative importance of one or more structural units of the chemical structure of the selected molecule with respect to a predicted olfactory property associated with the selected molecule; as well as providing, by the one or more computing devices, visualization data associated with the predicted data indicative of the one or more olfactory characteristics; The visualization data includes an indication of which portion of the chemical structure of the selected molecule is more important for a perceived property of the molecule.

4. The computer-implemented method of claim 1 , further comprising: Data is generated by the one or more computing devices indicating how structural changes to the chemical structure of the selected molecule affect predicted olfactory properties associated with the selected molecule.

5. The computer-implemented method of claim 1 , wherein: The predicted data indicative of the one or more olfactory properties of the selected molecule comprises an intensity of the olfactory property.

6. The computer-implemented method of claim 1 , further comprising: obtaining, by the one or more computing devices, a second diagram graphically depicting a second chemical structure of a second selected molecule; providing, by the one or more computing devices, a second graph graphically describing a second chemical structure of a second selected molecule as input to a graph neural network for machine learning; receiving, by the one or more computing devices, second prediction data describing one or more second olfactory properties associated with a second selected molecule as an output of the machine-learned graph neural network; as well as One or more olfactory differences between the selected molecule and the second selected molecule are determined, by the one or more computing devices, based on a comparison of the predicted data for the selected molecule and second predicted data for the second selected molecule.

7. The computer-implemented method of claim 1 , further comprising determining, by the one or more computing devices, data indicating one or more of the following by inputting a graph describing the chemical structure of the selected molecule into a machine-learned graph neural network or an additional machine-learned graph neural network: optical properties of the selected molecules; the taste properties of the selected molecules; biodegradability of the selected molecules; the stability of the selected molecule; or Toxicity of the selected molecules.

8. The computer-implemented method of claim 1 , wherein: The graph graphically depicting the chemical structure of the selected molecule includes a two-dimensional graph structure indicating a two-dimensional representation of the chemical structure of the selected molecule.

9. The computer-implemented method of claim 1 , wherein: The graph graphically depicting the chemical structure of the selected molecule includes a three-dimensional graph structure indicating a three-dimensional representation of the chemical structure of the selected molecule, and wherein the method further comprises performing, by the one or more computing devices, one or more quantum chemical calculations to identify the three-dimensional representation of the chemical structure of the selected molecule.

10. The computer-implemented method of claim 1 , further comprising: performing, by the one or more computing devices, an iterative search process to identify additional molecules exhibiting one or more desired olfactory properties, wherein the iterative search process comprises, for each of a plurality of iterations: generating, by the one or more computing devices, a candidate molecule graph that graphically describes candidate chemical structures of candidate molecules; Providing, by the one or more computing devices, a candidate molecule graph that graphically describes a candidate chemical structure of a candidate molecule as an input to a graph neural network for machine learning; receiving, by the one or more computing devices, prediction data describing one or more predicted olfactory properties of the candidate molecule as output of the machine-learned graph neural network; and The one or more predicted olfactory properties of the candidate molecule are compared to the one or more expected olfactory properties by the one or more computing devices.

11. A computing device comprising: one or more processors; as well as One or more non-transitory computer-readable media storing instructions that, when executed by the one or more processors, cause a computing device to perform operations comprising: obtaining a machine-learned graph neural network trained to predict one or more olfactory properties of a molecule based at least in part on chemical structure data associated with the molecule; obtaining graph data representing the chemical structure of the selected molecule; Provide graph data representing chemical structures as input to graph neural networks for machine learning; receiving, as an output of a machine-learned graph neural network, prediction data describing one or more olfactory properties associated with a selected molecule, wherein the prediction data indicative of the one or more predicted olfactory properties of the selected molecule comprises a numerical embedding; providing as output prediction data describing the one or more predicted olfactory properties of the selected molecule; and The one or more computing devices identify other molecules having olfactory properties similar to the predicted olfactory properties of the selected molecule by comparing the numerical embeddings with other numerical embeddings output by the machine-learned graph neural network for the other molecules.

12. The computing device of claim 11, wherein: Obtaining a machine learning graph neural network trained to predict one or more olfactory properties of a molecule also includes: obtaining training data comprising a plurality of example chemical structures, each example chemical structure being labeled with one or more olfactory property labels describing the olfactory properties of the example chemical structure; and A machine learning graph neural network is trained to predict olfactory properties based in part on the obtained training data.

13. The computing device of claim 11, the operations further comprising: Data is generated that indicates how structural changes to the chemical structure of the selected molecule affect predicted olfactory properties associated with the selected molecule.

14. The computing device of claim 11, the operations further comprising: generating visualization data depicting the relative importance of one or more structural units of a selected molecule with respect to a predicted olfactory property associated with the selected molecule; as well as providing visualization data associated with the predicted data describing one or more olfactory characteristics; The visualization data includes an indication of which portion of the chemical structure of the selected molecule is more important for a perceived property of the molecule.

15. The computing device of claim 11, wherein: The predicted data indicative of the one or more olfactory properties of the selected molecule comprises an intensity of the olfactory property.

16. The computing device of claim 11, the operations further comprising: obtaining graph data representing a chemical structure of a second selected molecule; providing the graph data representing the chemical structure of a second selected molecule as input to a graph neural network for machine learning; receiving, as output of the machine-learned predictive model, prediction data describing one or more olfactory properties associated with a second selected molecule; as well as One or more perceived differences between the selected molecule and a second selected molecule are determined.

17. The computing device of claim 11, the operations further comprising determining, based at least in part on the graph data representing the chemical structure, data indicating one or more of: optical properties of the selected molecules; the taste properties of the selected molecules; biodegradability of the selected molecules; the stability of the selected molecule; or Toxicity of the selected molecules.

18. The computing device of claim 11, wherein: The graph data representing the chemical structure of the selected molecule includes a graph structure indicating a two-dimensional structure of the selected molecule.

19. The computing device of claim 11, wherein: The graph data representing the chemical structure of the selected molecule includes a three-dimensional graph structure indicating a three-dimensional representation of the chemical structure of the selected molecule, wherein the operations further include performing one or more quantum chemical calculations to identify the three-dimensional representation of the chemical structure of the selected molecule.

Images