DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status
This instant application No. 18/998,586 has claims 1-15 pending.
Priority / Filing Date
Applicant’s claims for priorities of PCT/KR2023/010660 (filed on July 24, 2023) and provisional application No. 63/392,129 are acknowledged. The effective filing date for this instant application is July 26, 2022.
Abstract
The abstract of the disclosure is objected due to the use of implied language. Note that in the abstract, the language should be clear and concise and should not repeat information given in the title. It should avoid using phrases which can be implied, such as, “The disclosure concerns,” “The disclosure defined by this invention,” “The disclosure describes,” etc… See MPEP § 608.01(b).
Note that in the abstract, Applicant cites “The present invention relates to an artificial intelligence apparatus and chemical material search method thereof…” on lines 1-2. This citation clearly provokes the use of implied language and repeats the title. Correction and/or revision are required. One example is as follows:
“An apparatus and system capable of… ”
Drawings
The drawings filed on January 27, 2025 are accepted for examination purposes.
Claim Objections
Claim 13 is objected since the term hamming should be capitalized as Hamming for consistency compared to claim 14.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
Claims 13-14 are each rejected under 35 U.S.C. 112(b) for citing small hamming distances and large hamming distances in each claim since both the terms small and large are indefinite because these units are not quantifiable (e.g., how small is small and how large is large). Revision and/or corrections are required.
Information Disclosure Statement
As required by M.P.E.P. 609(C), the Applicant’s submission of the Information Disclosure Statement filed on January 27, 2025 is acknowledged by the Examiner and the cited references have been considered in the examination of the claims now pending. As required by M.P.E.P. 609 C(2), a copy of each of the PTOL-1449s initialed and dated by the Examiner is attached to the instant Office action.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
The claimed invention in claims 1-15 are directed to a judicial exception (i.e., an abstract idea) without significantly more.
Claims 1-15 pass step 1 of the 35 U.S.C. 101 analysis since each claim is either directed to a method, or an apparatus comprising a database and a processor implemented by hardware (e.g., quantum processor and/or learning processor embedded within an AI server per [64]-[91] of instant specification).
Claim 1 recites each, in part, elements that are directed to an abstract idea (“Courts have examined claims that required the use of a computer and still found that the underlying, patent-ineligible invention could be performed via pen and paper or in a person’s mind.” Versata Dev. Group v. SAP Am., Inc., 793 F.3d 1306, 1335, 115 USPQ2d 1681, 1702 (Fed. Cir. 2015)). The claim recites …searching for a target material from the database; predicting a fingerprint…by…extracting sample data…; evaluating a feature importance…; selecting high-level features…; and searching for the target material… which can be implemented in a human mind and/or with the aid of pen/paper (e.g., mentally searching and predicting by writing down the sample data on paper based on visually available dataset; mentally evaluating a feature importance; and mentally or visually searching for the target material). The limitations, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitations in the mind but for the recitation of generic computer components. That is, other than reciting the computer components (i.e., processor, executable instructions, and pre-trained neural network model), nothing in the claim precludes the limitations from being performed in the human mind per step 2A – prong 1 of the Abstract Idea Analysis. Thus, the limitations are parts of a mental process. Further, the claims recite additional steps of …storing datasets of chemical materials; and …inputting fingerprints of the datasets into a pre-trained neural network model which are extra-solution activities (per step 2A – prong 2 of the Abstract Idea Analysis) that cannot be integrated into a practical application (e.g., the elements recite trivial elements that occurred or would occur after the mental process such as generic data storing and data inputting techniques). Each of the additional limitation(s) is no more than mere instructions to apply the exception using a generic computer component (e.g., processor, memory, computer-executable instructions and AI model). The extra-solution activity in step 2A - prong 2 are reevaluated in step 2B to determining if each limitation is more than what is well-understood, routine, conventional activity in the field. The background of the limitations does not provide any indication that the computer components (e.g., processor, memory, and computer-executable instructions) are not off-the-shelf computer components. The Symantec, TLI, and OOP Techs court decisions cited in MPEP 2106.05(d)(II) indicate that mere receiving, generating, storing, determining, identifying, and transmitting of data over a network are a well-understood, routine, and conventional functions when claimed in a merely generic manner (as it is here). Accordingly, a conclusion that the claims are well-understood, routine, conventional (WURC) activity is supported under Berkheimer Option 2. For these reasons, there is no inventive concept in each claims, thus, the claims are ineligible.
Claim 2 further configures the processor to …convert molecular structures…by encoding…into binary numbers… which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally converting and/or writing down the molecular structures in forms of binary numbers on paper). Further, the claim is a fundamental data formatting operation which does not integrate the idea into a practical application, and converting data to binary is WURC activity which adds nothing significantly more. Thus, the claim is ineligible.
Claim 3 further configures the processor to …pre-train the neural network model to predict a data characteristic…before predicting the fingerprint… which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally learning/analyzing data based on a sequence of mathematical operations such as taking an input condition and updating the model’s parameters before running the final prediction). Further, this step appears to be an extra-solution and WURC activity of pre-training and tuning data model before the actual prediction. Thus, the claim is ineligible.
Claim 4 merely provides definition for the characteristic condition for the chemical materials. Thus, the claim is ineligible.
Claim 5 further configures the processor to …calculate a distance between a characteristic value… and a preset target characteristic value, and predict a fingerprint…based on the calculated distance which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally calculating the distance and mentally predicting the result based on the distance). The claim is pure mathematical application which does not integrate the math into a practical application, and such prediction based on a calculated distance is a WURC mathematical operation (e.g., predicting how far point A is from point B based on the calculated distance). Thus, the claim is ineligible.
Claim 6 further configures the processor to …predict a fingerprint closer…, and predict a fingerprint farther… which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally calculating the distance based on how close or how far each data point is from another). The claim is pure mathematical application which does not integrate the math into a practical application, and such prediction based on a calculated distance is a WURC mathematical operation (e.g., predicting a fingerprint based on how close or how far point A is from point B). Thus, the claim is ineligible.
Claim 7 further configures the processor to …extract the sample data using a quantum annealing method which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally extracting data using a quantum-related algorithm). The quantum annealing is inherently a mathematical optimization process used to find the global minimum of a given objective function over a given set of candidate solution. The claim is pure mathematical application which does not integrate the math into a practical application, and such extracting step based on global minimum is a WURC mathematical operation. Thus, the claim is ineligible.
Claim 8 further configures the processor to …optimize the fingerprint…based on a cost function…, and extract the sample data by calculating an optimized cost function…. which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally optimizing the fingerprint based on a cost function, mentally searching for desired data, and writing down on paper the data using an optimized cost function). The claim is pure mathematical application which does not integrate the math into a practical application, and such optimizing or extracting step based on global minimum is a WURC mathematical operation. Thus, the claim is ineligible.
Claim 9 further configures the processor to …evaluate a feature importance …from frequency obtained… which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally evaluating a feature importance based on observed frequency). The claim is basic statistical method of data analysis which does not integrate the math into a practical application, and such statistical analysis is a WURC mathematical operation. Thus, the claim is ineligible.
Claim 10 further configures the processor to …determine whether a reference value…is previously set, and…select high-level features… which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally determining whether a value is previously set and then mentally selecting high-level features based on the mental determination). The claim act as logical gates (e.g., if reference set, do X – otherwise, do Y) which is basic computer logic and does not integrate the math into a practical application, and such logic is a WURC data filtering operation. Thus, the claim is ineligible.
Claim 11 further configures the processor to …select a preset number of features…if the refence value…is not set which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally selecting a preset number of features when mentally determining that the reference value is not set). The claim act as logical gates (e.g., if reference is not set, do X – otherwise, do Y) which is basic computer logic and does not integrate the math into a practical application, and such logic is a WURC data filtering operation. Thus, the claim is ineligible.
Claim 12 further configures the processor to …search for the target material to extract molecular structures… which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally searching for the target material and writing down the molecular structures on paper). The claim recites extracting of data which is basic computer operation and does not integrate the math into a practical application, and such data retrieval operation is a WURC. Thus, the claim is ineligible.
Claim 13 further configures the processor to …extract molecular structures with small hamming distances…and…large hamming distances…and obtain specific key structure information…based on the obtained… which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally determining the small and/or large hamming distances, mentally searching for associated molecular structures and writing down the results on paper). The claim recites using Hamming distance as a mathematical calculation to find common structures which is a set comparison operation and does not integrate the math into a practical application, and such data comparison/retrieval using Hamming distance is WURC based on pure math. Thus, the claim is ineligible.
Claim 14 further configures the processor to …extract s preset number of molecular structures with small Hamming distances…; and extract a preset number of molecular structures with large Hamming distances… which can be implemented in a human mind with the aid of pen/paper as presented above (e.g., mentally determining the small and/or large hamming distances, mentally searching for associated molecular structures and writing down the results on paper). The claim recites using Hamming distance as a mathematical calculation to find common structures which is a set comparison operation and does not integrate the math into a practical application, and such data comparison/retrieval using Hamming distance is WURC based on pure math. Thus, the claim is ineligible.
Claim 15 recites each, in part, elements that are directed to an abstract idea (“Courts have examined claims that required the use of a computer and still found that the underlying, patent-ineligible invention could be performed via pen and paper or in a person’s mind.” Versata Dev. Group v. SAP Am., Inc., 793 F.3d 1306, 1335, 115 USPQ2d 1681, 1702 (Fed. Cir. 2015)). The claim recites converting datasets…into fingerprints; extracting sample data…; evaluating a feature importance…; selecting high-level features…; and searching for the target material… which can be implemented in a human mind and/or with the aid of pen/paper (e.g., writing down on paper the fingerprints of observed chemical structures; mentally searching and predicting by writing down the sample data on paper based on visually available dataset; mentally evaluating a feature importance; and mentally or visually searching for the target material). The limitations, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitations in the mind but for the recitation of generic computer components. That is, other than reciting the computer components (i.e., processor, executable instructions, and pre-trained neural network model), nothing in the claim precludes the limitations from being performed in the human mind per step 2A – prong 1 of the Abstract Idea Analysis. Thus, the limitations are parts of a mental process. Further, the claims recite an additional step of predicting…by inputting fingerprints of the datasets into a pre-trained neural network model which is an extra-solution activity (per step 2A – prong 2 of the Abstract Idea Analysis) that cannot be integrated into a practical application (e.g., the elements recite trivial elements that occurred or would occur after the mental process such as generic data storing and data inputting techniques). Each of the additional limitation(s) is no more than mere instructions to apply the exception using a generic computer component (e.g., processor, memory, computer-executable instructions and AI model). The extra-solution activity in step 2A - prong 2 are reevaluated in step 2B to determining if each limitation is more than what is well-understood, routine, conventional activity in the field. The background of the limitations does not provide any indication that the computer components (e.g., processor, memory, and computer-executable instructions) are not off-the-shelf computer components. The Symantec, TLI, and OOP Techs court decisions cited in MPEP 2106.05(d)(II) indicate that mere receiving, generating, storing, determining, identifying, and transmitting of data over a network are a well-understood, routine, and conventional functions when claimed in a merely generic manner (as it is here). Accordingly, a conclusion that the claims are well-understood, routine, conventional (WURC) activity is supported under Berkheimer Option 2. For these reasons, there is no inventive concept in the claim, thus, the claim is ineligible.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-2, 5-6, 10, 12, and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Hara et al. (Pub. No. US 2021/0110240, published on April 15, 2021; hereinafter Hara) in view of McCarthy et al. (Pub. No. US 2021/0264110, published on August 26, 2021; hereinafter McCarthy).
Regarding claim 1, Hara clearly shows and discloses an artificial intelligence apparatus (Figure 1) comprising:
a database configured to store datasets of chemical materials (The chemical compound database 110 may be a database that stores information about chemical compounds. The chemical compounds registered in the chemical compound database 110 may include various chemical compounds, each of which may be categorized into molecules, ionic compounds, intermetallic compounds and complexes, and may be used in a variety of fields including organic chemistry, inorganic chemistry, biochemistry, pharmacy, etc., [0018]); and
a processor configured to search for a target material from the database (The target set of the chemical compounds may be extracted from the chemical compound database 110 by using a query condition on the property information. For example, chemical compounds with a value of a specific property in a certain range (e.g. a chemical compounds that have boiling points above a specific threshold) may be extracted as target. Alternatively, chemical compounds associated with a specific label or tag may be extracted as target, [0023]), and
wherein the processor is configured to
predict a fingerprint relating to a target characteristic by inputting fingerprints of the datasets (At step S102, the processing unit may read the expression describing a structure of the chemical compound from the chemical compounds database 110. As shown in FIG. 3, the chemical structure of certain chemical compound (that is allyl cyanide or 3-butenenitrile in the example shown in FIG. 3A) can be expressed as a SMILES expression shown in FIG. 3B, [0043]) At step S104, the processing unit may enumerate one or more combinations of the position and the type of the structural element appeared in the expression. As shown in FIG. 3C, the cells corresponding to the enumerated combinations {(1, “C”), (2, “=”), (3, “C”), (4, “C”), (5, “C”), (6, “#”), (7, “N”)} may be set with the first specific value (“1”)), [0048]) into a pre-trained neural network model (At step S111, the processing unit may train the neural network 150 by the training module 140 based on the original and additional training data that is obtained for the target set of the chemical compounds. Parameters of the neural network 150, which may include weights between each units and biases of each unit, are optimized by appropriate training algorithm, [0057]),
extract sample data by optimizing the fingerprint relating to the target characteristic (The converting module 120 may be further configured to generate the training data based on the one or more enumerated combinations for each chemical compound in the target set, [0024]. The analyzing module 160 may be configured to extract information related to the structural feature specific to the target set of the chemical compounds by finding one or more input units strongly connected to a predetermined intermediate unit in the trained neural network 150, [0029]),
evaluate a feature importance from the extracted sample data (the trained neural network 150 can acquire an ability to capture the structural feature specific to the target set. By analyzing parameters of the trained neural network 150, one or more specific input units, each of which has at least one connection to any of one or more predetermined intermediate units stronger than other input units, can be found, [0059]),
select high-level features based on the feature importance (if the neural network 150 has merely one hidden layer, the one or more predetermined intermediate hidden units used for analyzing may be hidden units in the hidden layer. Strongest k-connections for each intermediate hidden unit can be enumerated and the one or more input units involved in any of the enumerated connections can be simply extracted as the information describing the specific structural feature, [0061]), and
search for the target material based on the selected high-level features (The target set of the chemical compounds may be extracted from the chemical compound database 110 by using a query condition on the property information. For example, chemical compounds with a value of a specific property in a certain range (e.g. a chemical compounds that have boiling points above a specific threshold) may be extracted as target. Alternatively, chemical compounds associated with a specific label or tag may be extracted as target, [0023]. If the neural network classifier 150B is used, the processing unit may train the neural network classifier 150B in a supervised manner by using the label information assigned to each group of the chemical compounds so as to enable it to discriminate the chemical compounds into appropriate groups well, [0030], [0035]-[0036], and [0058]).
McCarthy then alternatively or additionally discloses:
predicting a fingerprint relating to a target characteristic (molecules may be represented using a simplified molecular-input line-entry system (SMILES) string. For example, Aspirin (Acetylsalicylic Acid) may be represented by a SMILES string of CC(═O)OC1═CC═CC═C1C(═O)O. SMILES has a grammar that defines syntactically valid strings, and semantic structure may be defined elsewhere. In some implementations, the grammar may specify how ring bonds are written, but does not ensure that they always come in pairs, [0039]) by inputting fingerprints of the datasets into a pre-trained neural network model (the VAE 320 may include a class of generative models that produces a latent space whose encoded distribution may be traversed, such that generating a new sample that is close to the latent encoding of a known sample may be similar to the known sample. In some implementation, a grammar VAE may be used to generate samples with a discrete structure, e.g. molecules, arithmetic expressions, etc., [0045]),
searching for the target material based on the selected high-level features (Referring to step 460, the method 400 may include selecting one or more points in the latent space based on the query for the at least one candidate molecule. In one implementation, one selected point in the latent space may be a centroid/medoid of a first target label cluster, and another selected point in the latent space may be a centroid/medoid of a second target label cluster. For example, a first target label cluster may be a cluster of molecules for treating asthma; and a second target label cluster may be a cluster of molecules for treating diabetes, [0057]-[0059]).
It would have been obvious to an ordinary person skilled in the art at the time of the invention was effectively filed to incorporate the teachings of McCarthy with the teachings of Hara for the purpose of selecting predictive biomarkers based on chemical and/or biological data having features that can be described using variational auto-encoders to generate latent spaces for modeling the probability metrics distributions of latent variables that may be explored using interpolation methods and quantitative structure-activity relationship models.
Regarding claim 2, McCarthy then discloses converting molecular structures corresponding to the datasets of the chemical materials into the fingerprints by encoding the molecular structures into binary numbers before predicting the fingerprint relating to the target characteristic (Referring to step 414, the method 400 may include converting the historical drug data into the one-hot encoding format. Referring to FIG. 7A, the method may including converting molecular structure data in SMILES data format 710 to a parse tree based on the set of production rules corresponding to SMILES grammar; converting the parse tree 720 to a corresponding production rules 730; and converting the corresponding production rules 730 to the one-hot encoding format 740 for the molecular structure data, [0053]).
Regarding claim 5, Hara and McCarthy further disclose when predicting the fingerprint relating to the target characteristic, calculating a distance between a characteristic value of a dataset in the database and a preset target characteristic value, and predicting a fingerprint relating to the target characteristic based on the calculated distance (By analyzing parameters of the trained neural network 150, one or more specific input units, each of which has at least one connection to any of one or more predetermined intermediate units stronger than other input units, can be found. The one or more input units may represent a continuous structure or distant structures, [0059] of Hara. Interpolation may be implemented by calculating Jacobian matrix corresponding to first order derivative of change for all points of interest, and building a k-d tree over the resulting distances, [0048], 0059] of McCarthy).
Regarding claim 6, McCarthy further discloses predicting a fingerprint closer to the target characteristic as the distance between the characteristic value of the dataset in the database and the pre-set target characteristic value decreases; and predicting a fingerprint farther from the target characteristic as the distance between the characteristic value of the dataset in the database and the pre-set target characteristic value increases (Interpolation may be implemented by calculating Jacobian matrix corresponding to first order derivative of change for all points of interest, and building a k-d tree over the resulting distances. The method may receive user preference data including domain-specific heuristics such as fingerprint similarity, synthetic accessibility, absolute difference of drug-likeliness, or the like. The method may further determine an optimum path from the source to destination points on the k-dimensional (k-d) tree based on the preference data, [0048], [0059]).
Regarding claim 10, Hara further discloses determining whether a reference value for selection of the feature is previously set (For the first type of the analysis, the target set of the chemical compounds may be designated such that one group of chemical compounds with a similar property are included in the target set. The group may include chemical compounds labeled with same label and/or chemical compounds satisfying a same predetermined condition, [0031]), and if the reference value for selection of the feature is previously set, selecting high-level features having feature importance equal to or higher than the reference value based on the pre-set reference value, when selecting the high-level features (By analyzing parameters of the trained neural network 150, one or more specific input units, each of which has at least one connection to any of one or more predetermined intermediate units stronger than other input units, can be found, [0059]).
Regarding claim 12, Hara further discloses searching for the target material to extract molecular structures corresponding to the target material from the data sets in the database based on the selected high-level features when searching for the target material (The target set of the chemical compounds may be extracted from the chemical compound database 110 by using a query condition on the property information. For example, chemical compounds with a value of a specific property in a certain range (e.g. a chemical compounds that have boiling points above a specific threshold) may be extracted as target. Alternatively, chemical compounds associated with a specific label or tag may be extracted as target, [0023]).
Regarding claim 15, Hara clearly shows and discloses a method of searching a chemical material in an artificial intelligence apparatus (Abstract), the method comprising:
converting datasets of chemical materials into fingerprints (At step S102, the processing unit may read the expression describing a structure of the chemical compound from the chemical compounds database 110. As shown in FIG. 3, the chemical structure of certain chemical compound (that is allyl cyanide or 3-butenenitrile in the example shown in FIG. 3A) can be expressed as a SMILES expression shown in FIG. 3B, [0043]) At step S104, the processing unit may enumerate one or more combinations of the position and the type of the structural element appeared in the expression. As shown in FIG. 3C, the cells corresponding to the enumerated combinations {(1, “C”), (2, “=”), (3, “C”), (4, “C”), (5, “C”), (6, “#”), (7, “N”)} may be set with the first specific value (“1”)), [0048]);
predicting a fingerprint relating to a target characteristic by inputting fingerprints of the datasets into a pre-trained neural network model (At step S111, the processing unit may train the neural network 150 by the training module 140 based on the original and additional training data that is obtained for the target set of the chemical compounds. Parameters of the neural network 150, which may include weights between each units and biases of each unit, are optimized by appropriate training algorithm, [0057]);
extracting sample data by optimizing the fingerprint relating to the target characteristic (The converting module 120 may be further configured to generate the training data based on the one or more enumerated combinations for each chemical compound in the target set, [0024]. The analyzing module 160 may be configured to extract information related to the structural feature specific to the target set of the chemical compounds by finding one or more input units strongly connected to a predetermined intermediate unit in the trained neural network 150, [0029]),
evaluating a feature importance from the extracted sample data (the trained neural network 150 can acquire an ability to capture the structural feature specific to the target set. By analyzing parameters of the trained neural network 150, one or more specific input units, each of which has at least one connection to any of one or more predetermined intermediate units stronger than other input units, can be found, [0059]);
selecting high-level features based on the feature importance (if the neural network 150 has merely one hidden layer, the one or more predetermined intermediate hidden units used for analyzing may be hidden units in the hidden layer. Strongest k-connections for each intermediate hidden unit can be enumerated and the one or more input units involved in any of the enumerated connections can be simply extracted as the information describing the specific structural feature, [0061]); and
searching for the target material based on the selected high-level features (The target set of the chemical compounds may be extracted from the chemical compound database 110 by using a query condition on the property information. For example, chemical compounds with a value of a specific property in a certain range (e.g. a chemical compounds that have boiling points above a specific threshold) may be extracted as target. Alternatively, chemical compounds associated with a specific label or tag may be extracted as target, [0023]. If the neural network classifier 150B is used, the processing unit may train the neural network classifier 150B in a supervised manner by using the label information assigned to each group of the chemical compounds so as to enable it to discriminate the chemical compounds into appropriate groups well, [0030], [0035]-[0036], and [0058]).
McCarthy then alternatively or additionally discloses:
predicting a fingerprint relating to a target characteristic (molecules may be represented using a simplified molecular-input line-entry system (SMILES) string. For example, Aspirin (Acetylsalicylic Acid) may be represented by a SMILES string of CC(═O)OC1═CC═CC═C1C(═O)O. SMILES has a grammar that defines syntactically valid strings, and semantic structure may be defined elsewhere. In some implementations, the grammar may specify how ring bonds are written, but does not ensure that they always come in pairs, [0039]) by inputting fingerprints of the datasets into a pre-trained neural network model (the VAE 320 may include a class of generative models that produces a latent space whose encoded distribution may be traversed, such that generating a new sample that is close to the latent encoding of a known sample may be similar to the known sample. In some implementation, a grammar VAE may be used to generate samples with a discrete structure, e.g. molecules, arithmetic expressions, etc., [0045]),
searching for the target material based on the selected high-level features (Referring to step 460, the method 400 may include selecting one or more points in the latent space based on the query for the at least one candidate molecule. In one implementation, one selected point in the latent space may be a centroid/medoid of a first target label cluster, and another selected point in the latent space may be a centroid/medoid of a second target label cluster. For example, a first target label cluster may be a cluster of molecules for treating asthma; and a second target label cluster may be a cluster of molecules for treating diabetes, [0057]-[0059]).
It would have been obvious to an ordinary person skilled in the art at the time of the invention was effectively filed to incorporate the teachings of McCarthy with the teachings of Hara for the purpose of selecting predictive biomarkers based on chemical and/or biological data having features that can be described using variational auto-encoders to generate latent spaces for modeling the probability metrics distributions of latent variables that may be explored using interpolation methods and quantitative structure-activity relationship models.
Claims 3-4 are rejected under 35 U.S.C. 103 as being unpatentable over Hara in view of McCarthy and further in view of Wang et al. (Pub. No. US 2022/0215899, filed on December 21, 2021; hereinafter Wang I).
Regarding claim 3, Wang I then discloses when a characteristic condition of a chemical material to be explored is input (The test data set corresponding to the preset target obtained by the screening operation in the present embodiment may also be used in the training process of the affinity prediction model in the embodiment shown in FIG. 1 or FIG. 2, thus effectively guaranteeing the accuracy of the test data set of the preset target in the training sample, and then further improving the precision of the trained affinity prediction model, [0089]), pre-training the neural network model to predict a data characteristic corresponding to the characteristic condition based on fingerprints of training data and test data, before predicting the fingerprint relating to the target characteristic (since the pre-trained affinity prediction model is first adopted to screen the information of the several drugs, in the test data set finally obtained based on the information of the several drugs, the preset target and the drugs have higher affinities; however, in the test data set in the training sample in the embodiment shown in FIG. 1 or FIG. 2, the training target and the test drug may have a low affinity, as long as it is obtained through experiments, [0090]).
It would have been obvious to an ordinary person skilled in the art at the time of the invention was effectively filed to incorporate the teachings of Wang I with the teachings of Hara, as modified by McCarthy, for the purpose of using artificial intelligence to effectively improve accuracy and a training effect of the trained affinity prediction model associated with a target to be detected with a drug to be detected based on acquired a test data set corresponding to the target to be detected to participate in the prediction.
Regarding claim 4, Wang I further discloses the characteristic condition for the chemical materials includes the target characteristic of the chemical material to be explored and a target value of the target characteristic (The test data set corresponding to the preset target obtained by the screening operation in the present embodiment may also be used in the training process of the affinity prediction model in the embodiment shown in FIG. 1 or FIG. 2, thus effectively guaranteeing the accuracy of the test data set of the preset target in the training sample, and then further improving the precision of the trained affinity prediction model, [0089]).
Claims 7-8 are rejected under 35 U.S.C. 103 as being unpatentable over Hara in view of McCarthy and further in view of Miyazawa et al. (Pub. No. US 2021/0248186, published on August 12, 2021; hereinafter Miyazawa).
Regarding claim 7, Miyazawa then discloses extracting the sample data using a quantum annealing method (The extraction unit 11, the generation unit 12, and the operation unit 13 may be implemented by using program modules executed by a processor such as a central processing unit (CPU) or a digital signal processor (DSP). The operation unit 13 may be hardware that executes the simulated annealing method or the replica exchange method by using a digital circuit or may be hardware that performs quantum annealing, [0036]).
It would have been obvious to an ordinary person skilled in the art at the time of the invention was effectively filed to incorporate the teachings of Miyazawa with the teachings of Hara, as modified by McCarthy, for the purpose of executing a search for a ground state for a generated evaluation function based on problem data indicating the combinatorial optimization problem and an extracted set of variables satisfying a part of constraint conditions of a combinatorial optimization problem.
Regarding claim 8, Miyazawa further discloses optimizing the fingerprint relating to the target characteristic based on a cost function of a fingerprint prediction model, and extracting the sample data by calculating an optimized cost function based on the optimized fingerprint, when extracting the sample data (The extraction unit 32 extracts a set of variables satisfying a part of the constraint conditions of the combinatorial optimization problem based on the problem data acquired by the input unit 31 and stored in the storage unit 36. The extracted set of variables is stored in the storage unit 36, [0065]. See further Figure 1 and texts).
Claims 9, and 11 are rejected under 35 U.S.C. 103 as being unpatentable over Hara in view of McCarthy and further in view of Miller et al. (Pub. No. US 2020/0303071, published on September 24, 2020; hereinafter Miller).
Regarding claim 9, Miller then discloses when evaluating the feature importance, evaluating a feature importance of each fingerprint from frequency obtained from the sample data (…by weighing features of the data... The program code determines the frequency of a code and represents this frequency with a number between 0 and 1. The program code utilizes these frequency codes to perform binning based on how often each item occurs within the data set, [0061]).
It would have been obvious to an ordinary person skilled in the art at the time of the invention was effectively filed to incorporate the teachings of Miller with the teachings of Hara, as modified by McCarthy, for the purpose of determining the feature importance based on statistical significance of each feature in the data with respect to its effect on the generated model to accomplish a certainty function associated with the feature importance.
Regarding claim 11, Miller then discloses selecting a preset number of features belonging to high levels from features arranged in descending order of feature importance if the reference value for selection of the feature is not set (The program code selects top features (i.e., features with largest values of mutual information, down to the level of significance) from each of the categories and orders them in descending order (according to the values of mutual information). By removing data that does not include top features, the program code focuses the analysis and increases the efficiency in later identifications, [0061]).
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Hara in view of McCarthy and further in view of Wang et al. (Pub. No. US 2018/0150785, published on May 31, 2018; hereinafter Wang II).
Regarding claim 13, Wang II then discloses when extracting molecular structures corresponding to the target material, extracting molecular structures with small hamming distances and molecular structures with large hamming distances, obtain a first common molecular structure from the molecular structures with small hamming distances, obtain a second common molecular structure from the molecular structures with large hamming distances (the recommendation component 240 compares a set of common elements to the specified member profile to determine a hidden similarity value. The first hidden feature vector comprises a first set of common elements for a first portion of members of the second set of members associated with the set of first-level interactions. The second hidden feature vector comprises a second set of common elements for a second portion of members of the second set of members associated with the set of second-level interactions. The hidden similarity value may also be determined as a Hamming distance indicating a similarity between the set of common elements and the attributes of the specified member profile, [0067]), and obtain specific key structure information of the target characteristic for searching for the target material based on the obtained first common molecular structure and the obtained second common molecular structure (the recommendation component 240 receives the first hidden feature vector and the second hidden feature vector, of each job profile, from the respective vector model. Upon identifying attributes associated with the specified member profile as matching at least a portion of the first and second hidden feature vectors within or associated with one or more job profiles of the set of job profiles, the recommendation component 240 selects the one or more job profiles as a recommendation set, [0059]).
It would have been obvious to an ordinary person skilled in the art at the time of the invention was effectively filed to incorporate the teachings of Wang II with the teachings of Hara, as modified by McCarthy, for the purpose of enhancing searching and retrieval of desired objects with high level of similarity based on common elements associated with a set of feature vectors matching the desired objects.
Allowable Subject Matter
Claim 14 is objected for being dependent on a base rejected claim but would be allowable over the prior art (with the above 35 U.S.C. 101 rejection being resolved) if rewritten in independent form to incorporate the limitations of the base claim and all intervening claim(s).
Relevant Prior Art
The following prior art are deemed relevant to the claims:
Kumar et al. (Pub. No. US 2024/0020320) teaches obtaining at least one query related to one or more database operations; obtaining information pertaining to a set of multiple storage resources; classifying the at least one query as associated with at least one of multiple subsets of storage resources among the set of multiple storage resources by processing at least a portion of the query and at least a portion of the information pertaining to the set of multiple storage resources using one or more artificial intelligence techniques; and performing one or more automated actions based at least in part on the classifying of the at least one query.
Hammond et al. (Pub. No. US 2017/0213155) teaches a training an AI model can take place in one or more training cycles to yield a trained state of the AI model. A learner module or the predictor can elicit a prediction from the trained AI model and send the prediction to the instructor module. The instructor module, in turn, can send the prediction to the training data source for updated training data based upon the prediction and, optionally, instruct the learner module in additional training cycles. When the one or more training cycles are complete, the learner module can save the trained state of the network of processing nodes in the trained AI model.
Contact Information
Any inquiry concerning this communication or earlier communications from the Examiner should be directed to Son Hoang whose telephone number is (571) 270-1752. The Examiner can normally be reached on Monday – Friday (7:00 AM – 4:00 PM).
If attempts to reach the Examiner by telephone are unsuccessful, the Examiner’s supervisor, Sherief Badawi can be reached on (571) 272-9782. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/SON T HOANG/
Primary Examiner, Art Unit 2169 December 10, 2025