DETAILED ACTION
This Final Office action is responsive to the claims filed on February 12, 2026.
Claims 1-20 are under examination.
Claims 1-20 are rejected under 35 USC 101.
Claims 1-20 are rejected under 35 USC 102 as anticipated by Dhakal.
Response To Arguments/Amendments
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
35 USC 112 Rejections: The Applicants arguments and amendments have been considered and are persuasive. The rejections are herein withdrawn.
35 USC 101 Rejections: Regarding the software per se rejection, the Applicant’s arguments and amendments have been considered and are persuasive. This rejection has been withdrawn. With regard to the issue of subject matter eligibility, the Applicants arguments and amendments have been considered but are NOT persuasive. The Applicant’s arguments will be addressed in the order presented in the Applicant’s response.
The claims are allegedly not directed to the abstract idea, and, even if they were, they allegedly integrate the abstract idea into a practical application in the field of computer technology: The improvement to computer technology is not forthcoming as all elements of the computing are recited at a high level, indicating that the computing recited is merely generic, with no contribution to computing technology. Also, the Applicant has misstated the standards. Whether or not the claim is directed to an abstract idea includes whether or not the claim integrates the abstract idea into a practical application, not a separate inquiry as indicated in the response. The Applicant has also failed to assert what the improvement to computing is and has not stated the improvement in the claim. The Applicant also resorts to the Desjardines, which is limited by the recent Recentive decision and also concerns whether machine learning is math, not a mental process. The Applicant asserts that Desjardines expressly states that improvements to how a machine learning model itself operates, including training of a machine learning model, represent improvements to computer functionality. However, even if, arguendo, this assertion were true, the Applicant’s claim does not recite a machine learning model, let alone, a machine learning model training. Further, even if the claim recited either, there is no teaching in the claim that would represent an improvement to the machine learning technological field, whether it be an improved data structure or an improved training method. The claims do not even necessarily invoke machine learning at all, merely stating that a system for machine learning is used, but this could, under the broadest reasonable interpretation, include conventional methods of analysis that prepare data for eventual use in a machine learning model.
Contrary to the assertions of the Applicant, the determinations of binding sites based on the structures of proteins and their amino acid pairs was conducted using the mind and simple aids long before the advent of computational assembly of protein/amino complexes and the application of machine learning to the assembly. It is for this reason that the steps indicated in the rejection are mental processes AND ARE PRACTICALLY PERFORMABLE IN THE MIND OR WITH SIMPLE AIDS. The generic statement that machine learning is tangentially involved does not modify this determination but merely includes additional limitations that fail to confer eligibility under MPEP 2106.05(f).
The Applicant asserts that the methods improve machine learning by increasing the computational efficiency of machine learning, but the claim does not recite any machine learning. The claim merely states that these steps are conducted in the context of a machine learning system. Not only does the claim fail to recite this alleged improvement, but the Applicant is silent as to how the recitations of the claim confer the alleged improvement. The Applicant also conflates the Step 2B determination with integration into a practical application, but that is not the test at Step 2B. The question at Step 2B is whether the claim provides significantly more than the abstract idea. However, the claims recite no machine learning model, let alone any particular structure or elements of training of a machine learning model, that would represent an inventive concept that relies at least in some way on the additional limitations to provide significantly more than the abstract idea. The Applicant is attempting to claim any system that tangentially involves machine learning to monopolize the use of amino acid structures to determine potential binding sites, something that is a long-standing practice (See e.g., the Hill (1956) and the Bull (1968) references made of record), something that merely applies the mental processes on generic computing components under MPEP 2106.05(f), something that merely invokes a particular field when describing the input data under 2106.05(h), and something that is well-understood, routine, and conventional under MPEP 2106.05(d) (see e.g., the Djakal, the Joseph, the Gervasoni, the Mortazavi, the Veit Acosta, the Hill, and the Bull references made of record).
For a more detailed description of why the claims do not represent any improvement in computing, please see the substance of the rejection. Accordingly, the rejections are maintained.
35 USC 102/3: The Applicant’s arguments and amendments have been considered but are not persuasive. Contrary to the Applicant’s assertion, the Applicant cannot meaningfully assert that the amendment is not to overcome the art as presented in the rejection. There is a presumption that the Applicant has amended to overcome the art, and the Applicant has provided a mere assertion without evidence or arguments to rebut this presumption. The rebuttal is not accepted. The single Dhakal reference teaches all of the limitations of the claims arranged or combined in the same way as recited in the claim. Responsive to the amendment that was clearly made to overcome the art and that presents matter not presented in the claim set prior to the amendment, the existing 35 USC 102 rejection is not only maintained for the pre-amendment and post-amendment claim set, but a further 35 USC 103 rejection is additionally/alternatively applied to the subject matter of the claims as amended. Should the 35 USC 103 rejection be necessary, it was necessitated by the amendment. The specific mapping of the elements is provided in the body of the rejection. Accordingly, the 35 USC 102 rejection is maintained.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on December 3, 2025 was filed prior to this action. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
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.
Subject Matter Eligibility
Claims 1-20 are rejected under 35 U.S.C. 101 for being directed to a judicial exception without significantly more. The Applicant is attempting to claim any system that tangentially involves machine learning to monopolize the use of amino acid structures to determine potential binding sites, something that is a long-standing practice (See e.g., the Hill (1956) and the Bull (1968) references made of record), something that merely applies the mental processes on generic computing components under MPEP 2106.05(f), something that merely invokes a particular field when describing the input data under 2106.05(h), and something that is well-understood, routine, and conventional under MPEP 2106.05(d) (see e.g., the Djakal, the Joseph, the Gervasoni, the Mortazavi, the Veit Acosta, the Hill, and the Bull references made of record). The following is a detailed (in)eligibility analysis.
Independent Claims
Step 2A, Prong 1
Independent claims 1, 9, and 17 recite a mental process.
Claim 9
Claim 9 recites (claim limitations in bold italic, references to the Applicant’s specification):
[…] execute actions comprising:
determining […] a first feature representation characterizing a first molecule and a second feature representation characterizing a second molecule; (Mental Evaluation, Mental Process – This determination can be practically performed mentally or with the aid of pen, paper, or a calculator. See Page 7, Line 12- Page 9, Line 2)
determining […] a third feature representation characterizing linking information indicating the which of the plurality of amino acid units in the first molecule are able to be linked to which of the plurality of amino acids in the second molecule, the linking information comprising multiple pairs of amino acid units; (Mental Evaluation, Mental Process – This determination can be practically performed mentally or with the aid of pen, paper, or a calculator. See Page 7, Line 12- Page 9, Line 2)
determining […] a candidate region for the first molecule based at least on the first feature representation, the second feature representation, and the third feature representation, the candidate region comprising multiple candidate positions for docking the first molecule with the second molecule; and (Mental Evaluation, Mental Process – This determination can be practically performed mentally or with the aid of pen, paper, or a calculator. See Page 9, Line 3- Page 10, Line 20)
for each candidate position of the multiple candidate positions, determining […] a result of docking the first molecule with the second molecule at the candidate position. (Mental Evaluation, Mental Process – This determination can be practically performed mentally or with the aid of pen, paper, or a calculator. See Page 10, Line 21- Page 11, Line 15)
Claim 9 recites mental processes and, hence, under MPEP 2106.04(a)(2)(III), an abstract idea.
Claim 9 recites an abstract idea.
Step 2A, Prong 2
The claims fail to recite additional limitations that integrate the abstract idea into a practical application.
Claim 9
Claim 9 recites the following additional limitations:
An electronic device, comprising: at least one processor; and a memory coupled to the at least one processor, wherein the memory has instructions stored therein, and the instructions, when executed by the at least one processor, cause the electronic device to execute actions […]
[…] in a/the processor-based machine learning system […]
These are generic computing elements recited at a high level of generality, which, under MPEP 2106.05(f), fail to integrate the abstract idea into a practical application at Step 2A, Prong 2.
Also, the real-world quantities the data represents, if any, merely limit the abstract idea to a field of technology and, under MPEP 2106.05(h), fail to integrate the abstract idea into a practical application.
Claim 9 fails to recite any additional limitations that integrate the abstract idea into a practical application.
Claim 9 is directed to the abstract idea.
Step 2B
The claims fail to recite additional limitations that combine with the other elements of the claim to provide significantly more than the abstract idea that would confer an inventive concept.
An electronic device, comprising: at least one processor; and a memory coupled to the at least one processor, wherein the memory has instructions stored therein, and the instructions, when executed by the at least one processor, cause the electronic device to execute actions […]
[…] in a/the processor-based machine learning system […]
These are generic computing elements recited at a high level of generality, which, under MPEP 2106.05(f), fail to combine with the other elements of the claim to provide significantly more than the abstract idea that would confer an inventive concept.
Also, the real-world quantities the data represents, if any, merely limit the abstract idea to a field of technology and, under MPEP 2106.05(h), fail to combine with the other elements of the claim to provide significantly more than the abstract idea that would confer an inventive concept.
Claim 9 fails to provide additional limitations that combine with the other elements of the claim to provide significantly more than the abstract idea that would confer an inventive concept.
Claim 9 is ineligible.
Claims 1 and 17
Regarding claim 1, claim 1 recites the method claim 9 is configured to execute and is rejected for at least the same reasons as claim 9.
Claim 1 is ineligible.
Regarding claim 17, claim 17 recites the configured operations of claim 9 and is rejected for at least the same reasons as claim 9. Should the Applicant amend claim 17 to positively recite the CRM on which the software is stored, this amended version of claim 17 would merely recite an implementation of the memory recited in claim 9, and would be similarly rejected for at least the same reasons as claim 9.
Claim 17 is ineligible.
Dependent Claims:
The dependent claims are also ineligible for the following reasons.
Claims 2, 10, and 18
wherein determining a first feature representation characterizing a first molecule and a second feature representation characterizing a second molecule comprises: determining the first feature representation based on an amino acid sequence of the first molecule; and determining the second feature representation based on an amino acid sequence of the second molecule.
These determinations are elements of the determination steps of the independent claims and are, therefore, elements of the mental process, abstract idea for the same reasons as the corresponding determination steps of the independent claims.
Claims 2, 10, and 18 fail to provide any additional limitations that confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 2, 10, and 18 are ineligible.
Claims 3, 11, and 19
wherein determining a first feature representation characterizing a first molecule and a second feature representation characterizing a second molecule comprises: determining the first feature representation and the second feature representation using at least one feature extraction module.
The determination is still an mental process, abstract idea for the same reasons as the corresponding determination steps in the independent claims.
The recitation of the at least one feature extraction model and the AlphaFold model are generic computing elements that, under MPEP 2106.05(f), fails to confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 3, 11, and 19 fail to provide any additional limitations that confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 3, 11, and 19 are ineligible.
Claims 4, 12, and 20
The method according to claim 3, wherein the at least one feature extraction module comprises a first feature extraction module and a second feature extraction module, the first feature extraction module is further trained based on a training data set associated with the first molecule, and the second feature extraction module is further trained based on a training data set associated with the second molecule.
The feature extraction modules and the generic training are recited at a high level and are, therefore, generic computing elements that, under MPEP 2106.05(f), fail to confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 4, 12, and 20 fail to provide any additional limitations that confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 4, 12, and 20 are ineligible.
Claims 5 and 13
wherein determining the candidate region for the first molecule comprises further determining the candidate region based on attitude information of the second molecule.
These features merely qualify the determination steps of the independent claims and are elements of the mental process, abstract idea for at least the same reasons.
Claims 5 and 13 fail to provide any additional limitations that confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 5 and 13 are ineligible.
Claims 6 and 14
wherein determining a candidate region for the first molecule comprises: determining the candidate region using a machine learning model.
The determination is a mental process, an abstract idea as demonstrated with respect to the independent claims.
This recitation of the use of a machine learning at a high level is the use of a generic computing component, which, under MPEP 2106.05(f), fails to confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 6 and 14 fail to provide any additional limitations that confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 6 and 14 are ineligible.
Claims 7 and 15
The method according to claim 1, wherein determining a result of docking the first molecule with the second molecule at the candidate position comprises: determining the result using a molecular docking algorithm.
The determination is a mental process, an abstract idea as demonstrated with respect to the independent claims.
This recitation of the use of a machine learning at a high level is the use of a generic computing component, which, under MPEP 2106.05(f), fails to confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 7 and 15 fail to provide any additional limitations that confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 7 and 15 are ineligible.
Claim 8 and 16
wherein the first molecule is a targeted protein, and the second molecule is a ligand.
This describes the nature of the data used in the determinations and, so, is an element of the abstract idea.
Should it be found otherwise, specifying that the data represents targeted proteins or ligands merely limits the abstract idea to a particular field, which, under MPEP 2106.05(h), fails to confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 8 and 16 fail to provide any additional limitations that confer eligibility at Step 2A, Prong 2 and Step 2B.
Claims 8 and 16 are ineligible.
Claim Rejections - 35 USC § 102/103
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
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.
Claim(s) 1-20 is/are rejected under 35 U.S.C. 102(a)(1)/(a)(2) as anticipated by or, in the alternative, under 35 U.S.C. 103 as obvious over NPL: “Artificial intelligence in the prediction of protein–ligand interactions: recent advances and future directions” by Dhakal et al.
Claims 1, 9, and 17
Regarding claim 9, Dhakal teaches:
9. An electronic device, comprising: at least one processor; and memory coupled to the at least one processor, wherein the memory has instructions stored therein, and the instructions, when executed by the at least one processor, cause the electronic device to execute actions comprising: (Page 4, Left Column, Third Paragraph “Additionally, Chen’s group [39] summarizes the web servers and databases used in drug–target identification and drug discovery” – Web servers have memory and processors and are used to conduct docking simulations. See the Authors at the top of Page 1 and the footnotes of Page 1. - Also, the authors are computer scientists writing for their computer science departments. See the software used as elements of the methods, e.g., AlphaFold (Page 7, Left Column, Second Paragraph), AutoDock, AutoDock Vina, Glide, GOLD, and ICM (Page 16, Left Column, Last Paragraph). – The teaching of use of software implies the use of a computer that has memory and a processor of some kind.)
determining, in a processor-based machine learning system, a first feature representation characterizing a first molecule and a second feature representation characterizing a second molecule; (Dhakal Page 7, Prediction of the ligan bind sites of proteins “Virtual Screening (VS) requires knowledge of the location of the ligand-binding site (LBS), which in some cases this information is unknown. Accurate prediction of protein– ligand binding sites from a 3D protein structure plays a crucial role in structure-based drug design [50, 51] and can aid in drug side effects prediction [52] as well as understanding a protein’s function [53]. Intermolecular interactions between proteins and ligands occur through amino acid residues at specific positions in the protein, usually located in pocket-like regions. Identification of these key residues is imperative for elucidating protein function, analyzing molecular interactions and facilitating docking computations in virtual screening–based drug design. These specific key amino acid residues in proteins are called the LBSs. Empirical studies show that the actual ligand-binding site correlates to the biggest pocket on the surface of a protein [6, 54]. On a test set of 67 protein structures [55], the SURFNET architecture [56] successfully predicted the ligand-binding site as the largest pocket in 83% of the cases. The findings from LIGSITE [57] also displayed that the ligand-binding site was found in the largest pocket in all 10 proteins tested. Similar was the result yielded from POCKET [58]. Each amino acid (residue) has a distinct impact on the structure and function of a protein. Even if the measured distance between two residues in a protein sequence is long, the spatial distance between them may be short due to protein folding [59–61]. As a result, residues in the sequence that are far from the target residue sequentially, but spatially close, can also have a significant effect on the position of the binding residues. AlphaFold [62] can be considered as one of the major breakthroughs that predicts the tertiary structures of most proteins rather accurately integrating 1D, 2D and 3D protein features. Ultimately, there is a need to consider the spatially neighboring residues for the binding site prediction. Furthermore, the secondary and tertiary structure of the protein also impacts binding, often more significantly than the primary structure.” – The feature representation of a 3-D folded protein molecule is provided for each molecule being considered. The features include pockets that are elements of binding sites. Page 7, Prediction of the ligan bind sites of protein “Unfortunately, influenza is irritatingly good at manufacturing drug-resistant variations driving the need for constant development of new drugs. One such drug to be developed is peramivir also known as BCX-1812 (sketch shown in Figure 1A. Structurally, peramivir inhibits neuraminidase (interaction shown in Figure 6A) by forming numerous electrostatic interactions including salt bridges and H-bonding [63]. When observing the protein monomer, 11 hydrogen bonds can be found holding peramivir in place greatly increasing the binding affinity for the compound through its enthalpic effects on the system (Figure 6B and C) [6, 54–58]. Although the experimental determination provides the most accurate assignment of the binding locations, it is a time- and labor-intensive process. Computational methods for the detection and characterization of functional sites on proteins have grown in popularity, and as a result, numerous methods have been developed in recent decades attempting to address this issue. – This is an example showing that the structure of a protein and binding ligand are both modeled to determine binding. Abstract “Finally, we survey the machine learning (ML) approaches implemented to predict protein-ligand binding sites.” – The reference teaches methods conducted by machine learning systems.)
determining, in the processor-based machine learning system, a third feature representation characterizing linking information indicating which of a plurality of amino acid units in the first molecule are able to be linked to which of a plurality of amino acid units in the second molecule, the linking information comprising multiple pairs of amino acid units; (Dhakal Page 6, Left Column, First-Third Paragraphs “The most frequent datasets used in this domain are BioLip, CASPtargets, LigAsite and PDBbind. It is worth noting that many datasets were created using the original protein structure and ligand data in the Protein Data Bank(PDB) […] Starting from 2008, other types of biomolecular complexes in PDB were added into PDBbind. Being updated annually, the latest release (i.e.version2020) contains binding data (Kd, Ki and IC50 values) for 19 443 protein–ligand, 2852protein–protein, 1052protein–nucleic acid and 149 nucleic acid-ligand complexes as shown in Figure 5. Here, all binding data are curated by the authors derived from original literatures.“ – The datasets include amino acid units that are linked between protein molecule. Page 7, Left Column, Last Paragraph “Identification of these key residues is imperative for elucidating protein function, analyzing molecular interactions and facilitating docking computations in virtual screening–based drug design. These specific key amino acid residues in proteins are called the LBSs.” Page 11, Left Column, First Paragraph “They implemented Naïve Bayes classifier with amino acid residue in membrane protein sequence. Here, they predicted whether the given input is a ligand-binding residue or not using only sequence-based information. They opted Bayesian classifiers since they are resistant to real-world noise and missing values.” – Again, the amino acid pairs are considered as input data in a machine learning method.)
determining, in the processor-based machine learning system, a candidate region for the first molecule based at least on the first feature representation, the second feature representation, and the third feature representation, the candidate region comprising multiple candidate positions for docking the first molecule with the second molecule; and (Dhakal Page 7, Prediction of the ligand bind sites of protein “Intermolecular interactions between proteins and ligands occur through amino acid residues at specific positions in the protein, usually located in pocket-like regions.” Page 7, Binding site prediction methods “Many different approaches to predicting the binding site have been established over the last two decades, based on (i) templates, (ii) energy functions, (iii) geometric considerations and (iv) ML.” Page 9, Left Column, Last Paragraph “The growing availability of high-resolution protein structures in various databases has opened up new possibilities for machine learning (ML) applications. The basic workflow of existing ML methods to predict binding sites can be divided into five main steps: data acquisition and preprocessing, feature engineering, model development, training–testing, hyperparameter tuning and evaluation. At first, several sources of known protein–ligand binding data are aggregated, and several significant features are extracted to represent the protein and ligand, which are then normalized. Then, ML models are designed to use the input features to predict binding sites, including shallow supervised learning algorithms, artificial neural network, convolutional neural network and ensemble methods (different approaches are further described in detail in Sections ‘Classical ML methods for binding site prediction’ and ‘Deep learning methods for binding site prediction’). A typical ML workflow is illustrated in Figure 3. We group ML methods into two categories: classical ML methods (non–deep learning methods) and modern deep learning methods to be described separately below. – An ML model can be used to determine multiple potential binding sites from the 3D protein models of the protein and ligand. See Also Table 5 on Page 15, illustrating that essentially all of the studies used the PDBBind data set that includes structures of molecules and specific interactions between amino acid pairs of those molecules as input to a machine learning system.)
PNG
media_image1.png
740
1059
media_image1.png
Greyscale
for each candidate position of the multiple candidate positions, determining a result of docking the first molecule with the second molecule at the candidate position. (Dhakal Page 11, Deep learning methods for binding site prediction “In 2020, a 3D fully CNN (based on an architecture called U-Net) was published for finding druggable pockets on protein surface [97]. U-Net [102] is a state-of-the-art neural network architecture that was initially invented to deal with the 2D medical images. In this method, the task of pocket detection was reformulated as a 3D image segmentation problem. Both the input and output are represented as 3D grids of the same dimensions.” Page 12, Prediction of protein-ligand binding affinity “In order to be a lead molecule for drug development, a molecule must be able to bind tightly to a target protein; i.e. it must have a high affinity. The degree of attraction between a receptor (e.g. a protein) and its binding partner (e.g. drug or inhibitor) is measured by binding affinity, which can be expressed by the thermodynamic value of dissociation constant (Kd) or in the case of inhibitors (Ki). Table 3 demonstrates a variability of different inhibitors acting upon different proteins. SmCI N23A is a mutant variant of the SmCI inhibitor, demonstrating how small changes to an inhibitor can greatly affect Ki. Predicting a protein–ligand complex’s binding affinity (such as inhibition constant, dissociation constant and binding energy) is critical for efficient and effective rational drug design. However, experimentally measuring protein–ligand binding affinity is time-consuming and complex, which is one of the major bottlenecks of the drug discovery process.” – Binding affinities for each interaction of sites between the molecules is determined. One quantification of this is the dissociation constant. See Also Table 5 on Page 15, illustrating that essentially all of the studies used the PDBBind data set that includes structures of molecules and specific interactions between amino acid pairs of those molecules as input to a machine learning system to determine likely binding sites and binding affinity of bonds between those sites.)
Regarding claim 1, claim 1 recites the operations that the system of claim 9 is configured to execute, so claim 1 is rejected for at least the same reasons as claim 9.
Regarding claim 17, claim 17 recites the configured operations of claim 9 and is rejected for at least the same reasons as claim 9. Should the Applicant amend claim 17 to positively recite the CRM on which the software is stored, this amended version of claim 17 would merely recite an implementation of the memory recited in claim 9, and would be similarly rejected for at least the same reasons as claim 9.
Claims 2, 10, and 18
Regarding claim 10, Dhakal teaches the features of claim 9, and further teaches:
The device according to claim 9, wherein determining a first feature representation characterizing a first molecule and a second feature representation characterizing a second molecule comprises: determining the first feature representation based on an amino acid sequence of the first molecule; and determining the second feature representation based on an amino acid sequence of the second molecule. (Dhakal Page 7, Prediction of the ligand bind sites of proteins “Accurate prediction of protein– ligand binding sites from a 3D protein structure plays a crucial role in structure-based drug design [50, 51] and can aid in drug side effects prediction [52] as well as understanding a protein’s function [53]. Intermolecular interactions between proteins and ligands occur through amino acid residues at specific positions in the protein, usually located in pocket-like regions. Identification of these key residues is imperative for elucidating protein function, analyzing molecular interactions and facilitating docking computations in virtual screening–based drug design. These specific key amino acid residues in proteins are called the LBSs. Empirical studies show that the actual ligand-binding site correlates to the biggest pocket on the surface of a protein [6, 54].” – Amino sequences are used to determine the binding sites by dictating the protein folding and final conformation of the proteins. This is done for both the ligand and the target protein.)
Regarding claim 2, claim 2 recites the operations that the system of claim 10 is configured to execute, so claim 2 is rejected for at least the same reasons as claim 10.
Regarding claim 18, claim 18 recites the configured operations of claim 10 and is rejected for at least the same reasons as claim 10. Should the Applicant amend claim 18 to positively recite the CRM on which the software is stored, this amended version of claim 18 would merely recite an implementation of the memory recited in claim 10, and would be similarly rejected for at least the same reasons as claim 10.
Claims 3, 11, and 19
Regarding claim 11, Dhakal teaches the features of claim 9, and further teaches:
The device according to claim 9, wherein determining a first feature representation characterizing a first molecule and a second feature representation characterizing a second molecule comprises: determining the first feature representation and the second feature representation using at least one feature extraction module. (Dhakal Page 7, Right Column, Second Paragraph “Each amino acid (residue) has a distinct impact on the structure and function of a protein. Even if the measured distance between two residues in a protein sequence is long, the spatial distance between them may be short due to protein folding [59–61]. As a result, residues in the sequence that are far from the target residue sequentially, but spatially close, can also have a significant effect on the position of the binding residues. AlphaFold [62] can be considered as one of the major breakthroughs that predicts the tertiary structures of most proteins rather accurately integrating 1D, 2D and 3D protein features. Ultimately, there is a need to consider the spatially neighboring residues for the binding site prediction. Furthermore, the secondary and tertiary structure of the protein also impacts binding, often more significantly than the primary structure.” – Dhakal teaches the use of AlphaFold to determine the 3D structure and determined pockets that are binding site condidates.)
Regarding claim 3, claim 3 recites the operations that the system of claim 11 is configured to execute, so claim 3 is rejected for at least the same reasons as claim 11.
Regarding claim 19, claim 19 recites the configured operations of claim 11 and is rejected for at least the same reasons as claim 11. Should the Applicant amend claim 19 to positively recite the CRM on which the software is stored, this amended version of claim 19 would merely recite an implementation of the memory recited in claim 11, and would be similarly rejected for at least the same reasons as claim 11.
Claims 4, 12, and 20
Regarding claim 12, Dhakal teaches the features of claim 11, and further teaches:
wherein the at least one feature extraction module comprises a first feature extraction module and a second feature extraction module, the first feature extraction module is further trained based on a training data set associated with the first molecule, and the second feature extraction module is further trained based on a training data set associated with the second molecule. (Dhakal Page 7, Right Column, Second Paragraph “Each amino acid (residue) has a distinct impact on the structure and function of a protein. Even if the measured distance between two residues in a protein sequence is long, the spatial distance between them may be short due to protein folding [59–61]. As a result, residues in the sequence that are far from the target residue sequentially, but spatially close, can also have a significant effect on the position of the binding residues. AlphaFold [62] can be considered as one of the major breakthroughs that predicts the tertiary structures of most proteins rather accurately integrating 1D, 2D and 3D protein features. Ultimately, there is a need to consider the spatially neighboring residues for the binding site prediction. Furthermore, the secondary and tertiary structure of the protein also impacts binding, often more significantly than the primary structure.” – Dhakal teaches the use of AlphaFold to determine the 3D structure and determined pockets that are binding site candidates for each of the molecules based on the amino acid sequences. Each protein is modeled from its different amino acid sequence, which is a different training set.)
Regarding claim 4, claim 4 recites the operations that the system of claim 12 is configured to execute, so claim 4 is rejected for at least the same reasons as claim 12.
Regarding claim 20, claim 20 recites the configured operations of claim 12 and is rejected for at least the same reasons as claim 12. Should the Applicant amend claim 20 to positively recite the CRM on which the software is stored, this amended version of claim 20 would merely recite an implementation of the memory recited in claim 12, and would be similarly rejected for at least the same reasons as claim 12.
Claims 5 and 13
Regarding claim 13, Dhakal teaches the features of claim 9, and further teaches:
wherein determining the candidate region for the first molecule comprises further determining the candidate region based on at least one of the following: linking information about amino acid units of the first molecule and amino acid units of the second molecule; attitude information of the second molecule. (Dhakal Page 17-18, Interconnection of protein-ligand binding site, binding affinity and binding pose “Like all matter, proteins are influenced and shaped by thermodynamic principles that include protein–ligand interactions. Often when characterizing these interactions, intermolecular forces such as hydrogen bonding, van der Waals forces, ion-induced dipoles, desolvation and electrostatic forces are described. All of which directly impact the enthalpy ΔH of the system. Interactions such as H-bonding can net approximately 20 kJ/mol assuming optimal geometry and distance. The intermolecular interactions of the ligand can also dictate the pose of the ligand within the binding pocket. Poses can vary greatly from ligand to ligand due to adjustments of the ligand to bend or twist to accommodate the attractive and repulsive forces involved. It is important to remember molecules are dynamic in nature with varying degrees of flexibility and not static and stiff. Ligands will attempt to orient themselves in the lowest energy conformation possible. Of course, entropy also plays a big role as hydrophobicity of the compounds and the environment in which it is docking can greatly shift what kinds of ligands can bind and how the compound can be accommodated within the binding pocket. The hydrophobicity of the pocket and the ligand can drastically alter the pose of the prospective ligand or outright prevent docking even if the compound is capable of accommodating some of the intermolecular forces needed for specific binding. The entropic penalty of the protein–ligand complex must also not be ignored as it will require most of the free energy that would result from the stabilizing interactions. – Dhakal accounts for linking (e.g., hydrogen bonding, van der Waals forces, ion-induced dipoles, desolvation and electrostatic forces) of amino acides in the proteins and the attitude (pose) of the ligand.)
Regarding claim 5, claim 5 recites the operations that the system of claim 13 is configured to execute, so claim 5 is rejected for at least the same reasons as claim 13.
Claims 6 and 14
Regarding claim 14, Dhakal teaches the features of claim 9, and further teaches:
The device according to claim 9, wherein determining a candidate region for the first molecule comprises: determining the candidate region using a machine learning model. (Dhakal Page 11, Deep learning methods for binding site prediction “Deep learning methods have grown in popularity in recent years due to their potential in capturing complicated relationships hidden within the data. Several deep learning methods for binding site prediction are summarized in Table 2.” See Also Table 2 on Page 11 – Dhakal teaches several machine learning methods for determining the candidate regions.)
Regarding claim 6, claim 6 recites the operations that the system of claim 14 is configured to execute, so claim 6 is rejected for at least the same reasons as claim 14.
Claims 7 and 15
Regarding claim 15, Dhakal teaches the features of claim 9, and further teaches:
The device according to claim 9, wherein determining a result of docking the first molecule with the second molecule at the candidate position comprises: determining the result using a molecular docking algorithm. (Dhakal Page 16, Left Column, Last Paragraph “The current approaches of predicting protein–ligand binding pose typically have two steps: (i) generating protein–ligand binding poses and (ii) evaluating the poses using a scoring function. As the first step is mostly carried out by some standard/mature docking tools such as AutoDock [142], AutoDock Vina [143] […]” – Dhakal teaches using AutoDock to determine a result of a docking position, a docking score.)
Regarding claim 7, claim 7 recites the operations that the system of claim 15 is configured to execute, so claim 7 is rejected for at least the same reasons as claim 15.
Claims 8 and 16
Regarding claim 16, Dhakal teaches the features of claim 9, and further teaches:
The device according to claim 9, wherein the first molecule is a targeted protein, and the second molecule is a ligand. (Dhakal Pages 1-2, Introduction to protein-ligand interactions “Particularly, intermolecular interactions between proteins and ligands occur at specific positions in the protein, known as ligand-binding sites, which has sparked a lot of interest in the domain of molecular docking and drug design. Binding sites, also referred to as binding pockets, are typically concavities on the surface of proteins. Pockets, where small drug-like ligands bind, are typically located in deep cavities. Ligand-binding sites are typically found in large, deep pockets [5, 6] on the protein surface, while some of them may exist in exposed shallow clefts [7, 8]. In medicinal chemistry, there is an emphasis on identifying key proteins whose biochemical functions can be definitively linked to diseases. Such proteins become targets for drug development. In fact, the binding site is considered druggable if the ligand binds with high affinity at the binding site and has an effective therapeutic action [9].” Page 11, Deep learning methods for binding site prediction “Several convolutional layers were stacked to obtain hierarchical features from input. Binding residues belonging to any selected ligand class were classified as positive samples in the training sets, whereas the remainder were labeled as negative samples. Seven types of features are used for the protein-ligand binding residue prediction: position-specific score matrix, relative solvent accessibility, secondary structure, dihedral angle (predicted by ANGLOR [101]), conservation scores, residue type and position embeddings, which are purely derived from protein sequences.” – The binding site determinations are as between a protein and a ligand.)
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 2022/0323378 A1 to Joseph (Teaches ML and software methods, including AutoDock and AlphaFold, for analyzing binding sites of proteins and ligands.)
NPL: “A Comprehensive Mapping of the Druggable Cavities within the SARS-CoV-2 Therapeutically Relevant Proteins by Combining Pocket and Docking Searches as Implemented in Pockets 2.0” by Gervasoni et al. (Teaches ML and software methods, including AutoDock and AlphaFold, for analyzing binding sites of proteins and ligands.)
NPL: “Bioinformatics Aanalysis of Lampyridae Family Luciferases by Homology Modeling and Substrate Docking Studies” by Mortazavi et al. (Teaches ML and software methods, including AutoDock and AlphaFold, for analyzing binding sites of proteins and ligands.)
NPL: “The Impact of Crystallographic Data for the Development of Machine Learning Models to Predict Protein-Ligand Binding Affinity” by Veit-Acosta et al. (Teaches ML and software methods, including AutoDock and AlphaFold, for analyzing binding sites of proteins and ligands.)
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JAY MICHAEL WHITE whose telephone number is (571)272-7073. The examiner can normally be reached Mon-Fri 11:00-7:00 EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Ryan Pitaro can be reached at (571) 272-4071. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/J.M.W./Examiner, Art Unit 2188 /RYAN F PITARO/Supervisory Patent Examiner, Art Unit 2188