DETAILED ACTION
Applicant’s response, filed 4/16/2026, has been fully considered. The following rejections and/or objections are either reiterated or newly applied. They constitute the complete set presently being applied to the instant application.
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 .
Claim Status
Claims 1-20 are currently pending and examined on the merits.
Claims 1-20 are rejected.
Priority
The instant application claims priority to U.S. Provisional Application 63/181,772 filed on 29 April 2021 and European Application EP 21315150.9 filed on 27 August 2021. At this point in examination, the effective filing date of claims 1-20 is 29 April 2021.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 3 May 2022, 19 August 2022, and 28 November 2023 are in compliance with the provisions of 37 CFR 1.97. A signed copy of the corresponding 1449 form has been included with this Office Action.
Drawings
The drawings submitted 4/16/2026 have been accepted by the Examiner.
Specification
The objections to the specification are withdrawn, in view of the amendments to the specification submitted 4/16/2026.
Claim Rejections - 35 USC § 112
The previous rejections to claims 3-4, 13, and 19 under 35 U.S.C. 112(b) are withdrawn in view of the claim amendments.
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.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claims recite: (a) mathematical concepts, (e.g., mathematical relationships, formulas or equations, mathematical calculations); and (b) mental processes, i.e., concepts performed in the human mind, (e.g., observation, evaluation, judgement, opinion).
Subject matter eligibility evaluation in accordance with MPEP 2106:
Eligibility Step 1: Claims 1-10 are directed to a method (process) for providing the identification or classification of the target molecule for presentation to a user. Claims 11-16 are directed to a non-transitory computer-readable storage medium (machine). Claims 17-20 are directed to a system (machine). Therefore, these claims are encompassed by the categories of statutory subject matter, and thus satisfy the subject matter eligibility requirements under Step 1.
[Step 1: YES]
Eligibility Step 2A: First, it is determined in Prong One whether a claim recites a judicial exception, and if so, then it is determined in Prong Two whether the recited judicial exception is integrated into a practical application of that exception.
Eligibility Step 2A, Prong One: In determining whether a claim is directed to a judicial exception, examination is performed that analyzes whether the claim recites a judicial exception, i.e., whether a law of nature, natural phenomenon, or abstract idea is set forth or described in the claim.
Claims 1-6, 9, and 11-20 recite the following steps which fall within the mental processes and/or mathematical concepts groups of abstract ideas, as noted below.
Independent claims 1, 11, and 17 further recite:
identifying a surface mesh that defines a surface of the target molecule, the surface mesh comprising a plurality of vertices (i.e., mental processes);
identifying a plurality of surface patches by associating each vertex of the surface mesh with a respective patch (i.e., mental processes);
generating, by the system, a respective real space ID for each patch in the surface patches, wherein generating the respective real space IDs comprises: determining, based on the feature distribution of the patch, a respective feature similarity score for each of the feature distributions of the plurality of known patches (i.e., mental processes);
generating, by the system, a respective real space ID for each patch in the surface patches, wherein generating the respective real space IDs comprises: determining, based on the respective feature similarity scores, an averaged feature similarity score (i.e., mental processes);
using the latent space IDs and the real space IDs to identify at least one candidate item that includes a surface resembling a surface region of the target molecule, wherein the surface region comprises multiple patches in the plurality of surface patches of the target molecule (i.e., mental processes);
using the at least one candidate item to determine an identification or a classification of the target molecule (i.e., mental processes).
Dependent claims 2, 12, and 18 further recite:
wherein identifying a first candidate item that includes a first surface resembling the surface region of the target molecule (i.e., mental processes);
mapping vertices of the target molecule to vertices of the first candidate item, wherein a first vertex on the target molecule is mapped to a second vertex on the first candidate item when a difference between at least one feature at the first vertex and at the second vertex is within a predetermined threshold (i.e., mental processes, mathematical concepts);
identifying a cluster of vertices on the target molecule that are each within a predetermined threshold distance from at least one of the mapped vertices on the target molecule (i.e., mental processes);
aligning the cluster on the target molecule with multiple vertices of the first candidate item by using gradient descent, the multiple vertices being within the first surface on the first candidate item (i.e., mathematical concepts);
identifying, as the surface region, surface patches associated with the vertices of the cluster on the target molecule (i.e., mental processes).
Dependent claims 3, 13, and 19 further recite:
determining a spatial similarity score for the cluster based on a 3D distance between the vertices of the cluster and vertices on the first surface of the first candidate item (i.e., mental processes, mathematical concepts);
wherein the first candidate item is provided in response to determining that the spatial similarity score is greater than a threshold score (i.e., mental process).
Dependent claim 4 further recites:
identifying multiple clusters of vertices on the target molecule with vertices mapped to vertices of one or more candidate items (i.e., mental processes);
for each of the clusters, identifying one or more surfaces on the one or more candidate items as resembling respective surface regions of the target molecule (i.e., mental processes).
Dependent claims 5 and 15 further recite:
filtering out, from the multiple clusters, clusters that have less than a predetermined number of vertices (i.e., mental processes).
Dependent claims 6 and 16 further recite:
ranking each cluster in the multiple clusters based on one or more of (i) number of mapped vertices in the cluster, (ii) a ratio of the number of mapped vertices in the cluster to a total number of vertices in the cluster, (iii) the number of vertices on the at least one candidate item mapped to one or more vertices of the cluster, and (iv) a ratio of the number of vertices on the at least one candidate item mapped to one or more vertices of the cluster, to a total number of vertices on the candidate item (i.e., mental processes, mathematical concepts); and
filtering out from multiple clusters, clusters that are ranked lower than a specific threshold rank (i.e., mental processes).
Dependent claim 9 further recites:
using the identification or the classification of the antigen to design or identify an antibody for the antigen based on the epitope (i.e., mental processes).
Dependent claims 14 and 20 further recite:
identifying multiple clusters of vertices on the target molecule with vertices mapped to vertices of one or more candidate items (i.e., mental processes);
performing the operations for each of the clusters to identify one or more surfaces on the one or more candidate items as resembling respective surface regions of the target molecule (i.e., mental processes).
The abstract ideas recited in the claims are evaluated under the broadest reasonable interpretation (BRI) of the claim limitations when read in light of and consistent with the specification. As noted in the foregoing section, the claims are determined to contain limitations that can practically be performed in the human mind with the aid of a pencil and paper, and therefore recite judicial exceptions from the mental process grouping of abstract ideas. Additionally, the recited limitations that are identified as judicial exceptions from the mathematical concepts grouping of abstract ideas are abstract ideas irrespective of whether or not the limitations are practical to perform in the human mind.
Therefore, claims 1-6, 9, and 11-20 recite an abstract idea.
[Step 2a, Prong One: YES]
Eligibility Step 2A, Prong Two: In determining whether a claim is directed to a judicial exception, further examination is performed that analyzes if the claim recites additional elements that, when examined as a whole, integrates the judicial exception(s) into a practical application (MPEP 2106.04(d)). A claim that integrates a judicial exception into a practical application will apply, rely on, or use the judicial exception in a manner that imposes a meaningful limit on the judicial exception. The claimed additional elements are analyzed to determine if the abstract idea is integrated into a practical application (MPEP 2106.04(d)(I); MPEP 2106.05(a-h)). If the claim contains no additional elements beyond the abstract idea, the claim fails to integrate the abstract idea into a practical application (MPEP 2106.04(d)(III)).
The judicial exceptions identified in Eligibility Step 2A, Prong One are not integrated into a practical application because of the reasons noted below.
Claims 5 and 6 do not recite any elements in addition to the judicial exception, and thus are part of the judicial exception.
Claims 1, 11, and 17 recite generating a respective latent space ID for each of the surface patches by using a neural network. The limitation recites “using a neural network”, which provide nothing more than mere instructions to implement an abstract idea on a generic computer. See MPEP 2106.05(f). Therefore, the claimed additional elements do not integrate the abstract ideas into a practical application.
Claims 1-4, 11-14, and 17-20 recite the additional non-abstract elements of data gathering:
receiving a target molecule to be identified or classified (claims 1, 11, and 17);
generating, by the system, a respective real space ID for each patch in the surface patches, wherein generating the respective real space IDs comprises: for each patch in the surface patches, obtaining a plurality of known patches on a plurality of known surfaces with feature distributions similar to a feature distribution of the patch (claims 1, 11, and 17);
generating, by the system, a respective real space ID for each patch in the surface patches, wherein generating the respective real space IDs comprises: generating the respective real space ID for the patch based on the averaged feature similarity score (claims 1, 11, and 17);
obtaining one or more candidate items with known surfaces (claims 1, 11, and 17);
providing the identification or the classification of the target molecule for presentation to a user (claims 1, 11, and 17);
providing the first candidate item as an item that includes the first surface resembling the surface region of the target molecule (claims 2, 12, and 18);
wherein the first candidate item is provided in response to determining that the spatial similarity score is within a threshold score (claims 3, 13, and 19);
performing the method for each of the clusters to identify one or more surfaces on the one or more candidate items as resembling respective surface regions of the target molecule (claims 4, 14, and 20).
which are each a data gathering step, or a description of the data gathered.
Data gathering steps are not an abstract idea, they are extra-solution activity, as they collect the data needed to carry out the JE. The data gathering does not impose any meaningful limitation on the JE, or how the JE is performed. The additional limitation (data gathering) must have more than a nominal or insignificant relationship to the identified judicial exception. (MPEP 2106.04/.05, citing Intellectual Ventures LLC v. Symantee Corp, McRO, TLI communications, OIP Techs. Inc. v. Amason.com Inc., Electric Power Group LLC v. Alstrom S.A.).
Claims 11 and 17 recite the additional non-abstract elements (EIA) of a general-purpose computer system or parts thereof:
a non-transitory, computer-readable medium storing one or more instructions executable by a computer system (claims 11 and 17);
a system, comprising one or more processors (claim 17);
The EIA do not provide any details of how specific structures of the computer elements are used to implement the JE. The claims require nothing more than a general-purpose computer to perform the functions that constitute the judicial exceptions. The computer elements of the claims do not provide improvements to the functioning of the computer itself (as in DDR Holdings, LLC v. Hotels.com LP); they do not provide improvements to any other technology or technical field (as in Diamond v. Diehr); nor do they utilize a particular machine (as in Eibel Process Co. v. Minn. & Ont. Paper Co.). Hence, these are mere instructions to apply the JE using a computer, and therefore the claim does not recite integrate that JE into a practical application.
Thus, the additionally recited elements merely invoke a computer as a tool, and/or amount to insignificant extra-solution data gathering activity, and as such, when all limitations in claims 1-20 have been considered as a whole, the claims are deemed to not recite any additional elements that would integrate a judicial exception into a practical application. Claims 1-4, 11-14, and 17-20 contain additional elements that would not integrate a judicial exception into a practical application and are further probed for inventive concept in Step 2B.
[Step 2A Prong Two: NO]
Eligibility Step 2B: Because the claims recite an abstract idea, and do not integrate that abstract idea into a practical application, the claims are probed for a specific inventive concept. The judicial exception alone cannot provide that inventive concept or practical application (MPEP 2106.05). Identifying whether the additional elements beyond the abstract idea amount to such an inventive concept requires considering the additional elements individually and in combination to determine if they amount to significantly more than the judicial exception (MPEP 2106.05A i-vi).
The claims do not include any additional elements that are sufficient to amount to significantly more than the judicial exception(s) because of the reasons noted below.
With respect to claims 1-4, 11-14, and 17-20: The limitations identified above as non-abstract elements (EIA) related to data gathering do not rise to the level of significantly more than the judicial exception. Activities such as data gathering do not improve the functioning of a computer, or comprise an improvement to any other technical field. The limitations do not require or set forth a particular machine, they do not affect a transformation of matter, nor do they provide an unconventional step (citing McRO and Trading Technologies Int’l v. IBG). Data gathering steps constitute a general link to a technological environment. Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception are insufficient to provide significantly more (as discussed in Alice Corp.,).
With respect to claims 11 and 17: The limitations identified above as non-abstract elements (EIA) related to general-purpose computer systems do not rise to the level of significantly more than the judicial exception. These elements do not improve the functioning of the computer itself, or comprise an improvement to any other technical field (Trading Technologies Int’l v. IBG, TLI Communications). They do not require or set forth a particular machine (Ultramercial v. Hulu, LLC., Alice Corp. Pty. Ltd v. CLS Bank Int’l), they do not affect a transformation of matter, nor do they provide an unconventional step. Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception are insufficient to provide significantly more (as discussed in Alice Corp., CyberSource v. Retail Decisions, Parker v. Flook, Versata Development Group v. SAP America).
The additional element of generating a respective latent space ID for each of the surface patches by using a neural network (claims 1, 11, and 17) is conventional. Evidence for conventionality is shown by Khan et al. (Applied Artificial Intelligence, 2018, 33(1), 87-100). Khan et al. reviews “a patch-based technique for segmentation of latent fingerprint images, which uses Convolutional Neural Network (CNN) to classify patches” (Abstract, lines 6-8). This shows that latent spaces per surface patch are being classified using neural networks, which makes it a conventional practice in the art.
[Step 2B: NO]
Therefore, claims 1-20 are patent ineligible under 35 U.S.C. § 101.
Response to Arguments
Applicant's arguments, see pages 1-4, filed 4/16/2026, with respect to claims 1, 11, and 17, have been fully considered but they are not persuasive.
Applicant asserts that each of the amended claims 1, 11, and 17, as a whole, integrates into a practical application for improving the technical fields of identifying/classifying target molecules as well as the functioning of a computer performing operations to identify/classify target molecules (pg. 2-3, para. 2 of Applicant’s Remarks). This argument is not persuasive as claims 1, 11, and 17 do not encompass additional elements that integrate into a practical application, but rather data gathering steps.
With respect to the Applicant’s argument that using both “latent space IDs” and “real space IDs” provides improved efficiency and accuracy for classifying target molecules (pg. 4, para. 1 of Applicant’s Remarks), this argument is not persuasive because using the latent space IDs and the real space IDs to identify candidate items as recited in claims 1, 11, and 17 is a judicial exception and cannot be considered an additional element that integrates into a practical application. Furthermore, the claims reciting a particular way of generating real space IDs to classify target molecules encompasses limitations such as obtaining known surface patches and determining feature similarity scores, which are additional elements of data gathering and mental processes, respectively.
Therefore, the rejection to claims 1-20 under 35 USC § 101 is maintained with modifications as necessitated by amendment of the claims, filed 4/16/2026.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1, 8-11, 17, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Gainza et al. (Nature Methods, 2019, 17(2), 184-192), as provided in the IDS filed 9/19/2022, in view of Yin et al. (Proceedings of the National Academy of Sciences, 2009, 106(39), 16622-16626).
With respect to claims 1, 11, and 17:
Regarding the recited receiving a target molecule to be identified or classified, Gainza et al. discloses a “Protein molecular surface” (Fig. 1a). This figure suggests a surface of a target protein molecule to be identified or classified.
Regarding the recited identifying a surface mesh that defines a surface of the target molecule, the surface mesh comprising a plurality of vertices, Gainza et al. discloses “Briefly, from a protein structure we compute a discretized molecular surface (solvent excluded surface) and assign geometric and chemical features to every point (vertex) in the mesh.” (Pages 184-185, col. 1, lines 1-3, Fig. 1 a, b). This describes a discretized molecular surface, which is also the surface mesh of a target protein molecule. This also suggests that there is more than one vertex in the surface mesh, which is confirmed through observation of Fig. 1b.
Regarding the recited identifying a plurality of surface patches by associating each vertex of the surface mesh with a respective path, Gainza et al. discloses “Around each vertex of the mesh, we extract a patch with geodesic radius of r = 9 Å or r = 12 Å.” (Page 185, col. 1, lines 3-5, Fig. 1 a, b). This suggests that with the plurality of vertices in the surface mesh, surface patches are identified by associating each vertex with a respective path or radius, as specified above.
Regarding the recited generating a respective latent space ID for each of the surface patches by using a neural network, Gainza et al. discloses “A convolutional layer with a set of filters is then applied to the output of the soft polar grid layer.” (Page 186, col. 1, lines 6-7). Also, further discloses “The procedure is repeated for different patch locations similar to a sliding window operation on images, producing the surface fingerprint descriptor at each point in the form of a vector that embeds information about the surface patterns of the center point and its neighborhood.” (Page 186, col. 2, lines 4-8). Gainza et al. discloses “With this framework we created descriptors for surface patches that can be further processed in neural network architectures.” (Page 187, col. 1, lines 3-4). This suggests that convolutional layers are being used to compute latent space IDs for each surface patch. Fingerprint descriptors are latent space IDs because they are manually optimized vectors that describe the protein surface features. Furthermore, this is shown in Fig. 1d, where fingerprint descriptors are generated for each patch using application-specific neural network architectures.
Gainza et al. does not disclose generating, by the system, a respective real space ID for each patch in the surface patches, wherein generating the respective real space IDs comprises: for each patch in the surface patches, obtaining a plurality of known patches on a plurality of known surfaces with feature distributions similar to a feature distribution of the patch.
However, Yin et al. discloses “we scan the entire protein surface to locate all possible patches. However, instead of performing explicit comparisons of the patches, we use geometric fingerprints to rapidly judge if patches are similar. Unlikely patches are rejected, and only the patches with the best-scoring fingerprints are explicitly aligned to measure the surface similarity.” (pg. 16623, col. 1, para. 1). Also, further discloses “For each protein structure, we first calculate the DFSS scores of all possible patches as compared to the query patch, and kept the top 10% of the best-scoring (DFSS) patches for more accurate AFSS scoring.” (pg. 16626, col. 2, para. 2, lines 1-3). This teaches obtaining patches from known Protein Data Bank (PDB) surfaces with fingerprints similar to fingerprints of a query patch.
Gainza et al. does not disclose determining, based on the feature distribution of the patch, a respective feature similarity score for each of the feature distributions of the plurality of known patches.
However, Yin et al. discloses “We compare the fingerprints by measuring the root-mean deviations of each fingerprint bin as DFSS =
∑
i
(
x
i
-
y
i
)
2
/
N
, where n = 60 is the total number of bins, and
x
i
and
y
i
are the normalized distributions in bin i for the 2 patches, respectively.” (pg. 16625, col. 2, para. 5). Also, further discloses “each fingerprint is comprised of a 2-dimensional (4 by 15) array, with each element corresponding to the curvature distribution in the bin.” (pg. 16625, col. 2, para. 4, lines 13-14). This teaches a Direct Fingerprint Similarity Score (DFSS), which is a feature similarity score between two patch distributions.
Gainza et al. does not disclose determining, based on the respective feature similarity scores, an averaged feature similarity score.
However, Yin et al. discloses “For each patch
p
i
and
p
j
, we also search neighboring patches within 2.5 Å and compute all pairwise differences of the fingerprints. The best 5 pairwise similarity scores are selected as the difference between the patch
p
i
and
p
j
.” (pg. 16625, col. 2, para. 6). This teaches an Averaged Fingerprint Similarity Score (AFSS) protocol, which aggregates the best 5 similarity scores across neighboring patches into one averaged feature similarity score.
Gainza et al. does not disclose generating the respective real space ID for the patch based on the averaged feature similarity score.
However, Yin et al. discloses “The averaging procedure also provides a tentative alignment pose for the matching patches, which allows explicit alignment and comparison of the patches without undue additional computational cost.” (pg. 16622, col. 2, para. 2, lines 14-17). This teaches a real space alignment pose or identifier based on the averaged fingerprint similarity score (AFSS).
It would have been prima facie obvious to one of ordinary skill in the art to modify the molecular surface fingerprinting method disclosed by Gainza et al. to incorporate feature similarity scores disclosed by Yin et al. One would be motivated to incorporate feature similarity scores because all pairwise DFSS calculations and AFSS calculations take 0.1 s as disclosed by Yin et al. (pg. 16625, col. 1, para. 3). This indicates fast calculations of feature similarity scores, leading to efficient molecular surface fingerprinting. There is a likelihood of success, since both methods explore protein surface fingerprinting and are well known in the field of computational chemistry.
Yin et al. does not disclose obtaining one or more candidate items with known surfaces.
However, Gainza et al. discloses “scanning a large database of descriptors of potential binders” (Page 190, col. 2, lines 18-20, Fig. 5d). This suggests selecting from a large database, which contains one or more candidate items or descriptors. These descriptors are of potential binders, which implies known protein molecular surfaces.
Yin et al. does not disclose using the latent space IDs and the real space IDs to identify at least one candidate item that includes a surface resembling a surface region of the target molecule, wherein the surface region comprises multiple patches in the plurality of surface patches of the target molecule.
However, Gainza et al. discloses “Each point within a patch is assigned an array of geometric and chemical input features. The input features (chemistry and geometry) are not learned, they are precomputed properties from the molecular surface. MaSIF then learns to embed the surface patch’s input features into a numerical vector descriptor” (Page 184, col. 2, lines 13-14, Fig. 1 b, d). Also, further discloses “Specifically, the MaSIF-search workflow entails two stages: (1) scanning a large database of descriptors of potential binders and selecting the top decoys by descriptor similarity and (2) three-dimensional alignment of the complexes exploiting fingerprint descriptors of multiple points within the patch, coupled to a reranking of the predictions with a separate neural network.” (Page 190, col. 2, lines 17-23, Fig. 5d). The descriptors, which correspond to latent space IDs, are generated using the map of the geometric and chemical features, which are the polar geodesic coordinates that correspond to real space IDs. Both IDs are used to identify the matching patches in the database using the subsequent MaSIF-search workflow.
Yin et al. does not disclose using the at least one candidate item to determine an identification or a classification of the target molecule.
However, Gainza et al. discloses “scanning a large database of descriptors of potential binders and selecting the top decoys by descriptor similarity” (Page 190, col. 2, lines 18-20, Fig. 5d). This suggests that candidate items, descriptors referred to as decoys, are selected by descriptor similarity. This implies that a potential binder or target molecule can be identified or classified based on this descriptor similarity.
Yin et al. does not disclose providing the identification or the classification of the target molecule for presentation to a user.
However, Gainza et al. discloses Figure 4, which depicts how the identification or classification of the target molecule is presented to the user, in the form of three-dimensional models.
Claim 11 recites a non-transitory, computer-readable medium storing one or more instructions. Claim 17 recites a system comprising of one or more processors and a computer-readable storage device coupled to the one or more processors.
Broadly claiming an automated means to replace a manual function to accomplish the same result does not distinguish over the prior art. See Leapfrog Enters., Inc. v. Fisher-Price, Inc., 485 F .3d 1157, 1161, 82 USPQ2d 1687, 1691 (Fed. Cir. 2007) (“Accommodating a prior art mechanical device that accomplishes [a desired] goal to modern electronics would have been reasonably obvious to one of ordinary skill in designing children’s learning devices. Applying modern electronics to older mechanical devices has been commonplace in recent years.”); In re Venner, 262 F .2d 91, 95, 120 USPQ 193, 194 (CCPA 1958); see also MPEP § 2144.04. Furthermore, implementing a known function on a computer has been deemed obvious to one of ordinary skill in the art if the automation of the known function on a general purpose computer is nothing more than the predictable use of prior art elements according to their established functions. KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 417, 82 USPQ2d 1385, 1396 (2007); see also MPEP § 2143, Exemplary Rationales D and F. Likewise, it has been found to be obvious to adapt an existing process to incorporate Internet and Web browser technologies for communicating and displaying information because these technologies had become commonplace for those functions. Muniauction, Inc. v. Thomson Corp., 532 F.3d 1318, 1326-27, 87 USPQ2d 1350, 1357 (Fed. Cir. 2008).
With respect to claim 8:
Yin et al. does not disclose wherein the target molecule is a protein molecule, and a candidate item is a portion of a known protein molecule.
However, Gainza et al. discloses “To benchmark MaSIF-search we simulated a scenario where the binding site of a target protein is known, and one attempts to recapitulate the true binder of a protein among many other binders. Specifically, we benchmarked MaSIF-search in 100 bound protein complexes randomly selected from our testing set (disjoint from the training set). For each complex, we first selected the center of the interface in the target protein (see Methods), and then attempted to recover the bound complex within the 100 binder proteins comprising the test set (Fig. 5d).” (Page 190, col. 2, lines 29-37, Fig. 5). This suggests that the target protein is the target molecule to be identified among 100 bound protein complexes. To do this, the candidate item is implied to be a binder, where the portion of this protein molecule is docked to the target as seen in Fig. 5d.
With respect to claim 9:
Yin et al. does not disclose wherein the target molecule is an antigen, and a candidate item is an epitope.
However, Gainza et al. discloses “We used MaSIF-site to predict three such designed interfaces that have been experimentally validated: an influenza inhibitor (Fig. 4a), a homo-oligomeric cage protein (Fig. 4b), and an epitope-scaffold used as an immunogen (Fig. 4c). The designs were based on wild-type scaffold proteins with no binding activity, and in each case, we compared their interface score with that of the noninteracting wild type.” (Page 189, col 2, lines 9-15, Fig. 4). Candidate items such as homo-oligomeric cage proteins can serve as epitopes as they contain regions that can be recognized by the immune system. Therefore, the wild-type scaffold proteins and their respective designs serve as antigens and thus target molecules because Fig. 4 depicts the prediction of protein-protein interaction sites from candidate items on a set of target molecules.
Yin et al. does not disclose using the identification or the classification of the antigen to design or identify an antibody for the antigen based on the epitope.
However, Gainza et al. discloses “Overall, MaSIF-site may help to identify the sites of interactions with other proteins for PPI validation, paratope/epitope prediction or small molecule binding sites, for cases where evolutionary or experimental information may not be available.” (Page 189, col. 2, lines 17-20). This implies that protein-protein interaction site validation and epitope prediction can lead to the identification of the antigen when identifying sites of interactions with other proteins. This can assist in designing or identifying an antibody for the antigen, which implies a case where evolutionary or experimental information may not be available.
With respect to claim 10:
Yin et al. does not disclose wherein the neural network comprises multiple layers and each of the latent space IDs is generated using the same layers.
However, Gainza et al. discloses “MaSIF applies a geometric deep neural network to these input features using the polar coordinates to spatially localize features. The neural network consists of one or more layers applied sequentially; a key component of the architecture is the geodesic convolution, generalizing the classical convolution to surfaces and implemented as an operation on local patches.” (Page 185, col. 2, lines 7-12, Fig. 1d). Also, further discloses “Fingerprint descriptors are computed for each patch using application-specific neural network architectures, which contain reusable building blocks (geodesic convolutional layers).” (Fig. 1d). This suggests that the neural network contains several layers and the fingerprint descriptors (latent space IDs) are generated using reusable convolutional layers.
With respect to claim 20:
Yin et al. does not disclose identifying multiple clusters of vertices on the target molecule with vertices mapped to vertices of one or more candidate items; and performing the operations for each of the clusters to identify one or more surfaces on the one or more candidate items as resembling respective surface regions of the target molecule.
However, Gainza et al. discloses “RANSAC selects three random points from the binder patch and uses the computed descriptors to find the closest points in the target patch by descriptor distance. Using these three newly found correspondences, RANSAC attempts to align the source patch to the target patch. RANSAC iterates 2,000 times and selects the transformation with the highest number of points within 1.0 Å between binder and target.” (Methods, Section “Structural alignment and rescoring”, lines 6-11). This describes the method of mapping, where the attempt to align the source patch (candidate item) to the target patch (target molecule) is performed multiple times to identify the transformation with the highest number of points, which indicates the candidate item with a surface resembling a surface region of the target molecule. The closest points in the target patch implies the multiple clusters of vertices on the target molecule to be identified with the vertices mapped to vertices of the candidate item, which is further described in the three random points from the binder patch used along with the descriptors.
Claims 2-7, 12-16, and 18-19 are rejected under 35 U.S.C. 103 as being unpatentable over Gainza et al. (Nature Methods, 2019, 17(2), 184-192) and Yin et al. (Proceedings of the National Academy of Sciences, 2009, 106(39), 16622-16626), as applied to claims 1, 8-11, 17, and 20 above, in view of Jain, A.N. (Journal of Computer-Aided Molecular Design, 1996, 10, 427-440).
Gainza et al. and Yin et al. are applied to claims 1, 8-11, 17, and 20 above.
With respect to claims 2, 12, and 18:
Yin et al. does not disclose wherein identifying a first candidate item that includes a first surface resembling the surface region of the target molecule.
However, Gainza et al. discloses “scanning a large database of descriptors of potential binders and selecting the top decoys by descriptor similarity” (Page 190, col. 2, lines 18-20, Fig. 5d). This suggests that candidate items, descriptors referred to as decoys, are selected by descriptor similarity, which means they will have surfaces similar to that of the target molecule.
Yin et al. does not disclose mapping vertices of the target molecule to vertices of the first candidate item, wherein a first vertex on the target molecule is mapped to a second vertex on the first candidate item when a difference between at least one feature at the first vertex and at the second vertex is within a predetermined threshold.
However, Gainza et al. discloses “RANSAC selects three random points from the binder patch and uses the computed descriptors to find the closest points in the target patch by descriptor distance. Using these three newly found correspondences, RANSAC attempts to align the source patch to the target patch. RANSAC iterates 2,000 times and selects the transformation with the highest number of points within 1.0 Å between binder and target.” (Methods, Section “Structural alignment and rescoring”, lines 6-11). This suggests that the correspondences are the vertices being mapped between the target molecule and the binder (first candidate item), where the distance between their points are within 1.0 Å. This implies a difference between at least one feature because the points are fingerprint descriptors that have feature information associated to them.
Yin et al. does not disclose identifying a cluster of vertices on the target molecule that are each within a predetermined threshold distance from at least one of the mapped vertices on the target molecule.
However, Gainza et al. discloses “RANSAC selects three random points from the binder patch and uses the computed descriptors to find the closest points in the target patch by descriptor distance.” (Methods, Section “Structural alignment and rescoring”, lines 6-8). This suggests that the closest points in the target molecule is the cluster of vertices that are within a predetermined threshold distance, which is implied in the points being the closest to each other.
Yin et al. does not disclose identifying, as the surface region, surface patches associated with the vertices of the cluster on the target molecule.
However, Gainza et al. discloses “MaSIF-search will produce similar descriptors for pairs of interacting patches (low Euclidean distances between fingerprint descriptors), and dissimilar descriptors for noninteracting patches (larger Euclidean distance between fingerprint descriptors). Thus, identifying potential binding partners is reduced to a comparison of numerical vectors.” (Page 190, col. 1, lines 12-18, Fig. 5a). This implies that surface patches can be identified through the comparison of their associated fingerprint descriptors.
Yin et al. does not disclose providing the first candidate item as an item that includes the first surface resembling the surface region of the target molecule.
However, Gainza et al. discloses “The top decoy patches with the shortest fingerprint descriptor distance to the target patch are selected as a shortlist of potential binding partners.” (Methods, Section “Structural alignment and rescoring”, lines 2-4). The shortlist of potential binding partners contains a first candidate item.
Gainza et al. and Yin et al. do not disclose aligning the cluster on the target molecule with multiple vertices of the first candidate item by using gradient descent, the multiple vertices being within the first surface on the first candidate item.
However, Jain discloses “The three critical requirements on a scoring function for a molecular docking system are accuracy, speed, and tolerance to inaccurate poses of putative ligands in protein binding sites. The function F defined here satisfies these requirements. The expected mean error of predicted affinity, estimated by cross-validation across a diverse set of binding sites and ligands, is 1.0 log unit. Using a simple optimization to speed up the identification of protein atoms near the ligand, the time to compute the affinity of benzamidine to trypsin is 0.03 s, which is fast enough for use in a docking search engine. F is a continuous and differentiable function whose maxima correspond closely to crystallographically determined structures. So, imprecise putative ligand poses can be efficiently optimized by gradient descent.” (Conclusions, lines 1-15).
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention if some motivation in the prior art would have led that person to substitute in the prior art teachings for the instant claim limitations. Gainza et al. teaches alignment of the patches of the target molecule to that of the candidate item using the RANSAC algorithm in Open 3D (Methods, Section “Structural alignment and rescoring”, Supplementary Fig. 6). However, Jain teaches that alignment, or docking, can also be done using gradient descent. This technique is more advantageous because it is a simple optimization method that significantly speeds up the identification of protein atoms near the ligand, and ultimately the affinity between two molecules. One of ordinary skill in the art would recognize that substituting the RANSAC algorithm for gradient descent would have a predictable result as both function to optimize the alignment between target and candidate molecules.
With respect to claims 3, 13, and 19:
Yin et al. and Jain do not disclose determining a spatial similarity score for the cluster based on a 3D distance between the vertices of the cluster and vertices on the first surface of the first candidate item.
However, Gainza et al. discloses “To discriminate true alignments we trained a separate neural network to score binder patches after the alignment step (Supplementary Fig. 6). Once a patch alignment has been made, the nearest neighbor on the binder in 3D space to each point in the target is searched, establishing correspondences (Supplementary Fig. 6b). Then, the input to the neural network is the 3D Euclidean distance, the MaSIF-search fingerprint distance and the product of the normal between correspondences. The output is a predicted score on the alignments.” (Methods, Section “Neural network for scoring aligned patches”, lines 1-8). Gainza et al. further discloses “For each point in an aligned patch we found its nearest neighbor (in 3D space, after alignment) on the target patch; for each pair of (binder, target) points we measured MaSIF-search fingerprint descriptor distance; the Euclidean distance in 3D space and dot products between their normals.” (Methods, Section “Neural network for scoring aligned patches”, lines 14-19). This suggests that a neural network is used to score vertices of the target patch (cluster) based on its 3D Euclidean distance to the vertices of the binder patch (first candidate item), where the predicted score is a spatial similarity score.
Yin et al. and Jain do not disclose wherein the first candidate item is provided in response to determining that the spatial similarity score is greater than a threshold score.
However, Gainza et al. discloses “A second-stage alignment and scoring method generates the complexes based on the identified fingerprints. The top decoy patches with the shortest fingerprint descriptor distance to the target patch are selected as a shortlist of potential binding partners.” (Methods, Section “Structural alignment and rescoring”, lines 1-4). The top decoy patches with the shortest fingerprint descriptor distance indicates that there is a threshold score the spatial similarity score must be above, and the shortlist of potential binding partners contains a first candidate item.
With respect to claim 4:
Yin et al. and Jain do not disclose identifying multiple clusters of vertices on the target molecule with vertices mapped to vertices of one or more candidate items; and for each of the clusters, identifying one or more surfaces on the one or more candidate items as resembling respective surface regions of the target molecule.
However, Gainza et al. discloses “RANSAC selects three random points from the binder patch and uses the computed descriptors to find the closest points in the target patch by descriptor distance. Using these three newly found correspondences, RANSAC attempts to align the source patch to the target patch. RANSAC iterates 2,000 times and selects the transformation with the highest number of points within 1.0 Å between binder and target.” (Methods, Section “Structural alignment and rescoring”, lines 6-11). This describes the method of mapping, where the attempt to align the source patch (candidate item) to the target patch (target molecule) is performed multiple times to identify the transformation with the highest number of points, which indicates the candidate item with a surface resembling a surface region of the target molecule. The closest points in the target patch implies the multiple clusters of vertices on the target molecule to be identified with the vertices mapped to vertices of the candidate item, which is further described in the three random points from the binder patch used along with the descriptors.
With respect to claims 5 and 15:
Yin et al. and Jain do not disclose further comprising filtering out, from the multiple clusters, clusters that have less than a predetermined number of vertices.
However, Gainza et al. discloses “RANSAC iterates 2,000 times and selects the transformation with the highest number of points within 1.0 Å between binder and target.” (Methods, Section “Structural alignment and rescoring”, lines 10-11). This suggests that clusters that have the highest number of points within a specified distance between the candidate item and target molecule are selected, which further implies excluding clusters with the lowest number of vertices.
With respect to claims 6 and 16:
Yin et al. and Jain do not disclose ranking each cluster in the multiple clusters based on one or more of (i) number of mapped vertices in the cluster, (ii) a ratio of the number of mapped vertices in the cluster to a total number of vertices in the cluster, (iii) the number of vertices on the at least one candidate item mapped to one or more vertices of the cluster, and (iv) a ratio of the number of vertices on the at least one candidate item mapped to one or more vertices of the cluster, to a total number of vertices on the candidate item.
However, Gainza et al. discloses a “three-dimensional alignment of the complexes exploiting fingerprint descriptors of multiple points within the patch, coupled to a reranking of the predictions with a separate neural network (see Methods and Supplementary Fig. 6).” (Page 190, col. 2, lines 20-23, Fig. 5d). This indicates that while identifying multiple clusters of vertices through the alignment of complexes, a ranking process follows.
Yin et al. and Jain do not disclose filtering out, from the multiple clusters, clusters that are ranked lower than a specific threshold rank.
However, Gainza et al. discloses “Each aligned patch was limited to 200 points, if the size of the aligned patch was greater than 200 points it was randomly sampled and if it was lower than 200 points it was zero-padded. Thus, the input to the network is a matrix of size 200,3 (200 point pairs with three features per pair).” (Methods, Section “Neural network for scoring aligned patches”, lines 20-23). This suggests that clusters lower than 200 points are excluded or filtered out, making 200 points the threshold rank in this case.
With respect to claim 7:
Yin et al. and Jain do not disclose wherein the at least one feature at a vertex includes one or more of a shape index, a distance-dependent curvature, a hydropath, a continuum electrostatics, and a number of free electrons/protons at that vertex.
However, Gainza et al. discloses “For each vertex within the patch, we compute two geometric features (shape index and distance-dependent curvature) and three chemical features (hydropathy index, continuum electrostatics and the location of free electrons and proton donors).” (Page 185, col. 1, lines 7-11, Fig. 1b). This suggests all five of the features that can be associated to a vertex.
With respect to claim 14:
Yin et al. and Jain do not disclose identifying multiple clusters of vertices on the target molecule with vertices mapped to vertices of one or more candidate items; and performing the operations for each of the clusters to identify one or more surfaces on the one or more candidate items as resembling respective surface regions of the target molecule.
However, Gainza et al. discloses “RANSAC selects three random points from the binder patch and uses the computed descriptors to find the closest points in the target patch by descriptor distance. Using these three newly found correspondences, RANSAC attempts to align the source patch to the target patch. RANSAC iterates 2,000 times and selects the transformation with the highest number of points within 1.0 Å between binder and target.” (Methods, Section “Structural alignment and rescoring”, lines 6-11). This describes the method of mapping, where the attempt to align the source patch (candidate item) to the target patch (target molecule) is performed multiple times to identify the transformation with the highest number of points, which indicates the candidate item with a surface resembling a surface region of the target molecule. The closest points in the target patch implies the multiple clusters of vertices on the target molecule to be identified with the vertices mapped to vertices of the candidate item, which is further described in the three random points from the binder patch used along with the descriptors.
Response to Arguments
Applicant’s arguments, see pg. 5-6, filed 4/16/2026, with respect to the rejection(s) of claim(s) 1-20 under 35 U.S.C. § 103 have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Gainza et al. (Nature Methods, 2019, 17(2), 184-192).
Conclusion
No claims are allowed.
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.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Jammy Luo whose telephone number is (571)272-2358. The examiner can normally be reached Monday - Friday, 9:00 AM - 5:00 PM 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, Larry D Riggs can be reached at (571)270-3062. 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.N.L./Examiner, Art Unit 1686
/LARRY D RIGGS II/Supervisory Patent Examiner, Art Unit 1686
Prosecution Insights