DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
The terminal disclaimer filed on Oct 20, 2025 disclaiming the terminal portion of any patent granted on this application which would extend beyond the expiration date of US Patented No 12,299,884 B2 has been reviewed and is accepted. The terminal disclaimer has been recorded.
Applicant’s representative has filed a Terminal disclaimer to overcome the double
patenting rejection, therefore, the double patenting rejection is now withdrawn
Response to Arguments
Applicant's arguments filed on 10/20/2025 have been fully considered but they are not persuasive.
On pages 4-5 Applicant asserts “
PNG
media_image1.png
170
760
media_image1.png
Greyscale
PNG
media_image2.png
124
740
media_image2.png
Greyscale
”.
Response: Examiner respectfully disagree with applicant’s argument. Drake expressly discloses machine learning models configured to receive molecular data and output activities or classifications of multiple analytes. For example: Drake on page [0332] disclose “The machine learning model may receive as inputs a set of features corresponding to properties of each of a plurality of classes of molecules of a biological sample….A set of features corresponding to properties of each of the plurality of classes of molecules may be identified and to be input to a machine learning model. A feature vector of feature values from each of the plurality of sets of measured values may be generated”, Drake further on para [0454] disclose “Each set of measured values may be from one assay applied to a class of molecules ….the different assays can be lcWGS, WGBS, cf-miRNA sequencing, and protein concentration measurements”, and Drake additionally on para [0025] disclose “inputting the feature vector into the machine learning model to obtain an output classification of whether the biological sample has the specified property”. These passages confirm that Drake provides the overall ML-based framework and the target output activities of multiple molecular analytes-implemented on a computer system.
Although the examples cite biological sample features (e.g., blood, plasma, or urine, para [0322-0333]), nothing in Drake limits the feature source. A person of ordinary skill in the art would recognize that generating embeddings or feature vectors from image-based data is an art-recognized, predictable alternative to assay-based features for inferring analyte activity, as widely practiced in deep-learning-based biomarker prediction.
Drake provides the multi-analyte problem setting and the ML pipeline: the model “receives…. a plurality of classes of molecules,” forms a “feature vector…. from of the plurality of sets of measured values,” and “inputs the feature vector…to obtain an output classification.”. These quotes establish Drake’s multi-analyte outputs and computer-implemented ML system.
Feala teaches the claimed “first module-embedding-second model outputs” architecture. Feala explicitly describes a molecular neural-network design where a first component generates an embedding and subsequent components decode that embedding into multiple predicted activities or properties. (See Fig. 4: “Embedding Layer……input (None, 1024) output: (None, 512)”), and “DECODER…. REGRESSION LAYER…PREDICTED ACTIVITY”. (Fig. 1 and Fig. 10).
A person of ordinary skill in the art would have found it obvious to employ Feala’s embedding-multi-output architecture in Drake’s multi-analyte prediction system. Substituting a known image encoder (e.g., a convolutional or transformer-based module) for Feala’s embedding layer is a predictable design choice routinely used to handle image inputs. Doing so allows a single learned embedding to support multiple analyte predictions, improving generalization and computational efficiency-benefits expressly described by Feala’s multiple-head structure.
Further, it was well known in the medical ML art, it is routine to use medical images to predict physiologic/biochemical markers. (e.g., retinal photographs predicting HbA1c or lipid levels). A skilled person in the art therefore would have been motivated to feed image-derived embeddings into Drake’s multi-analyte ML model to infer analyte activities non-invasively.
The 103 rejections properly combine Drake’s disclosure of multi-analyte ML prediction systems with Feala’s explicit embedding-decoder/multi-head framework. Employing a standard image encoder as the first module to generate an embedding from a medical image would have been a routine, predictable substitution yielding the claimed system. Accordingly, the argument has been fully considered but are not persuasive, and the rejection of claim 1 is maintained.
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 text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claims 1-5, 7-12, and 14-20 are rejected under 35 U.S.C. 103 as being unpatentable over Drake et al. (US 20210210205 A1) in view of Feala et al. (US 20220122692 A1).
Regarding claim 1, Drake et al. teaches a system for predicting activities of one or more molecular analytes of a subject (see Abstract; Systems and methods that analyze blood-based cancer diagnostic tests using multiple classes of molecules are described. The system uses machine learning (ML) to analyze multiple analytes, for example cell-free DNA, cell-free microRNA, and circulating proteins, from a biological sample”, para [0606]; “E. Example 5: A Gene Expression Prediction Model that Uses cfDNA Fragment Coverage and Length to Predict which Genes are Highly or Lowly Expressed in cfDNA-Producing Cells” gene expression implies activity of a molecular analyte of a patient), the system comprising : one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors (see para [0050]; “a system comprising one or more computer processors and computer memory coupled thereto. The computer memory comprises machine-executable code that, upon execution by the one or more computer processors, implements any of the methods above or elsewhere herein”), the one or more programs including instructions for: receiving a medical image of the subject (see Abstract; “the system receives a biological sample, and separates a plurality of molecule classes from the sample”, see also para [0322]; “the system receives the biological sample including a plurality of classes of molecules. Example biological samples are described herein, e.g., blood, plasma, or urine. Separate samples can also be received”, Note: the system is directed to ML that integrate various sample-derived signals to predict molecular analytes; in this art, “medical images” (e.g., histology, retinal, or other clinical imaging) are well known ML inputs to infer molecular/biochemical activity from phenotype. Drake’s spec teaches flexible inputs for multi-analyte prediction implemented on standard processors); and inputting the embedding into a second module of the machine learning model to predict the activities of the one or more molecular analytes for the subject, where in the second module of the machine learning model is configured to provide as output activities of a plurality of molecular analytes (see para [0332]; “The machine learning model may receive as inputs a set of features corresponding to properties of each of a plurality of classes of molecules of a biological sample. A plurality of classes of molecules in the biological sample may be assayed to be obtained a plurality of sets of measured values representative of the plurality of classes of molecules…….set of features corresponding to properties of each of the plurality of classes of molecules may be identified and to be input to a machine learning model. A feature vector of feature values from each of the plurality of sets of measured values may be generated” Feature vector implies embeddings”, see also para [0606]; “E. Example 5: A Gene Expression Prediction Model that Uses cfDNA Fragment Coverage and Length to Predict which Genes are Highly or Lowly Expressed in cfDNA-Producing Cells” Note: gene expression implies activity of a molecular analyte of a patient). However, Drake et al. does disclose inputting the medical image from the subject into a first module of a machine learning model to obtain an embedding and does not explicitly disclose a second module of the machine learning model.
In the same field of endeavor Feala et al. teaches inputting the medical image from the subject into a first module of a machine learning model to obtain an embedding, a second module of the machine learning model (see Fig. 4, and Fig. 10, see also para [0007]; “a method of modeling a desired protein property comprising: (a) providing a first pretrained system comprising a neural net embedder and, optionally, a neural net predictor, the neural net predictor of the pretrained system being different from the desired protein property; (b) transferring at least a part of the neural net embedder of the pretrained system to a second system comprising a neural net embedder and a neural net predictor, the neural net predictor of the second system providing the desired protein property; and (c) analyzing, by the second system, the primary amino acid sequence of a protein analyte, thereby generating a prediction of the desired protein property for the protein analyte” see also para [0137]; “The second system 1011 then applies the transfer learning method to predict activity by replacing the decoder layer 1016 with a linear regression layer 1026, and further training the resulting model to predict scalar enzymatic activity values 1022 as a supervised task. The labeled sequences are split randomly into training and test sets”). Accordingly, it would have been obvious to one of ordinary skill in the before the effective filing date of the claimed invention to modify the general system uses machine learning (ML) to analyze multiple analytes of Drake et al. in view of the use a method for a machine learning used to generate models that identify associations between amino acid sequences and protein functions based on amino acid sequence information of Feala et al. in order to provide an accurate and reproducible method for predicting protein function based on an amino acid sequence (see para [0007]).
Regarding claim 2, the rejection of claim 1 is incorporated herein.
The combination of Drake et al and Feala et al. further teach wherein the predicted activity of the molecular analyte comprises amplification signature data and/or a chromosome accessibility score comprising an ATAC-seq peak value (see Drake et al para [0184]; “Biological information may also include information regarding transcription start sites, transcription factor binding sites, assay for transposase-accessible chromatin using sequencing (ATAC-seq) data, histone marker data, DNAse hypersensitivity sites (DHSs), or combinations thereof”).
Regarding claim 3, the rejection of claim 2 is incorporated herein.
The combination of Drake et al and Feala et al. further teach wherein the amplification signature data is generated based on a plurality of differentially expressed genes with respect to an amplification and a plurality of weights (see Drake et al para [0255]; “The known nucleic acid sequence for the genes is sufficient to enable one to routinely select primers to amplify any portion of the gene” see also para [0138]; “any type of nucleic acid amplification reaction can be used to amplify a target nucleic acid molecule or fragment thereof and generate an amplified product)”).
Regarding claim 4, the rejection of claim 3 is incorporated herein.
The combination of Drake et al and Feala et al. further teach wherein the amplification signature data comprises a gene- specific copy number amplification (CNA) (see Drake et al page [0365; “TABLE-US-00001 TABLE 1 Gene Seq Name CNV p-value Feature p-value”).
Regarding claim 5, the rejection of claim 1 is incorporated herein.
The combination of Drake et al and Feala et al. further teach the one or more programs further including instructions for: using the third machine learning model to determine a measure of significance or prognostic value of the molecular analyte to dynamically select a subset of molecular analytes for subsequent use (see Drake et al para [0343]; “Machine Learning 45:5 2001) are used to make a classification (e.g., diagnostic or prognostic call) from the measured values (e.g., intensity data) for the assay data (e.g., transcript set) or their products” see also para [0492]; “The regions can be dynamically assigned by discovery. It is possible to take a number of samples from different classes and discover which regions are the most differentially methylated between the different classifications. One then selects a subset to be differentially methylated and uses these for classification”).
Regarding claim 7, the rejection of claim 5 is incorporated herein.
The combination of Drake et al and Feala et al. further teach wherein the second module of the machine learning model and/or the third module of the machine learning model are trained using transfer learning (see Feala et al. para [0058]; “transfer learning is used to enhance predictive accuracy. Transfer learning is a machine learning technique where a model developed for one task can be reused as the starting point for a model on a second task”). Accordingly, it would have been obvious to one of ordinary skill in the before the effective filing date of the claimed invention to modify the general system uses machine learning (ML) to analyze multiple analytes of Drake et al. in view of the use a method for a machine learning used to generate models that identify associations between amino acid sequences and protein functions based on amino acid sequence information of Feala et al. in order to boost predictive accuracy on a task where there is limited data by having the model learn a on a related task where data is abundant (see para [0058]).
Regarding claim 8, the rejection of claim 1 is incorporated herein.
The combination of Drake et al and Feala et al. further teach wherein the first module of the machine learning model has been trained using a plurality of medical images of a first cohort (see Drake et al para [0359]; “The machine learning predictor may be trained using datasets e.g., datasets generated by performing multi-analyte assays of biological samples of individuals) from one or more sets of cohorts of patients having cancer as inputs and known diagnosis (e.g., staging and/or tumor fraction) outcomes of the subjects as outputs to the machine learning predictor”).
Regarding claim 9, the rejection of claim 1 is incorporated herein.
The combination of Drake et al and Feala et al. further teach wherein the first module of the machine learning model comprises an embedding module (see Feala et al. para [00117]; “the first model comprises an embedder and the second model comprises a predictor… the first machine learning software module trains the first model on a first training data set comprising at least 10,000 protein properties”).
Regarding claim 10, the rejection of claim 8 is incorporated herein.
The combination of Drake et al and Feala et al. further teach wherein the second module of the machine learning model has been trained using one or more molecular analyte data sets obtained from a second cohort (see Drake et al para [0359]; “The machine learning predictor may be trained using datasets e.g., datasets generated by performing multi-analyte assays of biological samples of individuals) from one or more sets of cohorts of patients having cancer as inputs and known diagnosis (e.g., staging and/or tumor fraction) outcomes of the subjects as outputs to the machine learning predictor”).
Regarding claim 11, the rejection of claim 10 is incorporated herein.
The combination of Drake et al and Feala et al. further teach wherein the one or more molecular analyte data sets comprises: gene expression data; copy number amplification (CNA) data; chromatin accessibility data; DNA methylation data; histone modification; RNA data; protein data; spatial biology data; whole-genome sequencing (WGS) data (see Drake et al. para [0110]; “FIGS. 36A-36G show cfDNA derived V-plots around TSS regions used to predict gene expression”); see also para [0576]; “Paired-end whole-genome sequencing (WGS) was performed on plasma DNA samples obtained from 937 control subjects and 524 patients diagnosed with CRC”) somatic mutation data; germline mutation data; or any combination thereof (see Drake et al. para [0120]; “A somatic mutation can refer to nucleic acids inducing acquired or abnormal variations (e.g., cancers, obesity, symptoms, diseases, disorders, etc.). Germline variants are inherited, and thus correspond to an individual's genetic differences that he or she is born relative to a canonical human genome. Somatic variants are variants that occur in the zygote or later on at any point in cell division, development, and aging. In some examples, an analysis can distinguish between germline variants, e.g., private variants, and somatic mutations”).
Regarding claim 12, the rejection of claim 11 is incorporated herein.
The combination of Drake et al and Feala et al. further teach wherein the one or more molecular analyte data sets comprise: a gene expression value comprising an abundance of a transcript (see Drake et al para [0249]; “When using small cfRNA (including one-RNA and miRNA) as an analyte, the measured values relate to the abundance for these cfRNAs. Their transcripts are of a certain size, and each transcript is stored, and the number of cfRNAs found for each can be counted”); a copy number amplification value; an amplification signature value; abundance of one or more histone modifications comprising a ChIP-seq value; abundance of one or more mRNA sequences; abundance of one or more proteins; the presence of one or more somatic mutations; the presence of one or more germline mutations; the presence or absence of one or more specific DNA methylation marks in one or more specific genomic regions, or any combination thereof (see Drake et al para [0414]; “The data can be curated from relevant open databases such as GTEX, TCGA, and ENCODE. This includes ChIP-seq, RNA-seq, and eQTL. A record for each subject can include fields with the measured data and labels of the subjects”).
Regarding claim 14, the rejection of claim 1 is incorporated herein.
The combination of Drake et al. and Feala et al. further teach wherein the one or more programs further include instructions for: receiving a medical image of a new subject (see Drake et al. para [0446]; “At block 530, new samples types are acquired, or potentially more samples of a same type, e.g., to increase the number of samples in a cohort”); obtaining an embedding by providing the medical image of the new subject to the first module of the machine learning model (see Drake et al. para [0332]; “A set of features corresponding to properties of each of the plurality of classes of molecules may be identified and to be input to a machine learning model. A feature vector of feature values from each of the plurality of sets of measured values may be generated”); mapping the embedding based on domain adaptation (see Drake et al. para [0294]; “machine learning systems can be leveraged to assess the effectiveness of a given dataset generated from a given assay or plurality of assays and run on a given analyte to add to the overall prediction accuracy of classification. In this manner, a new biological/health/diagnostics question can be tackled to design a new assay”),
Regarding claim 15, the rejection of claim 14 is incorporated herein.
The combination of Drake et al. and Feala et al. further teach wherein mapping the embedding based on domain adaptation comprises: inputting the embedding obtained based on the medical image of the new subject into a fourth module of the machine learning model (see Feala et al. para [0060]; “With transfer learning, the pretrained first system or a portion thereof can be transferred to form part of a second system. The one or more layers of the neural net embedder can be frozen when used in the second system. In some embodiments, the second system comprises the neural net embedder or a portion thereof from the first system. In some embodiments, the second system comprises a neural net embedder and a neural net predictor. The neural net predictor can include one or more output layers for generating a final output or prediction. The second system can be trained using a second training data set that is labeled according to the protein function or property of interest. As used herein, an embedder and a predictor can refer to components of a predictive model such as neural net trained using machine learning”).
Regarding claim 16, the rejection of claim 1 is incorporated herein.
The combination of Drake et al. and Feala et al. further teach the one or more programs further including instructions for: generating an annotation map of the predicted activity of the molecular analyte (see Drake et al. para [0088]; “FIG. 14 shows a heatmap of chromosomal structure scores determined from Hi-C sequencing of the same region of the genome as in FIG. 13” see also para [0211]; “the position of cfDNA sequence reads within the genome can be determined by “mapping” the sequence to a reference genome. Mapping can be performed with the aid of computer algorithms including, for example, the Needleman-Wunsch algorithm, the BLAST algorithm, the Smith-Waterman algorithm, a Burrows-wheeler alignment, a suffix tree, or a custom-developed algorithm”); and overlaying the annotation map on the medical image (see para [0564]; “The feature set was obtained by counting a number of cfDNA fragments having a mapping quality of at least 60 that overlapped with each of the annotated gene regions by at least one base, thereby producing a “gene feature” set (D=24,152, covering 1352 Mb) for each sample”).
Regarding claim 17, the rejection of claim 16 is incorporated herein.
The combination of Drake et al. and Feala et al. further teach wherein the annotation map includes a visualization distinguishing healthy tissue from diseased tissue (see Drake et al. para [0576]; “TABLE-US-00006 TABLE 6 Number of healthy and cancer samples used for CRC experiments”).
Regarding claim 18, the rejection of claim 1 is incorporated herein.
The combination of Drake et al. and Feala et al. further teach wherein the plurality of medical images comprises: one or more histopathology images; one or more magnetic resonance imaging (MRI) images; one or more computerized tomography (CT) scans; or any combination thereof (see Drake et al. para [0352]; “This clinical test may comprise an imaging test, a blood test, a computed tomography (CT) scan, a magnetic resonance imaging (MRI) scan, an ultrasound scan, a chest X-ray, a positron emission tomography (PET) scan, a PET-CT scan, or any combination thereof”).
Regarding claim 19, the scope of claim 19 is fully encompassed by the scope of claim 1, accordingly, the rejection of claim 1 is fully applicable here
Regarding claim 20, the scope of claim 20 is fully encompassed by the scope of claim 1, accordingly, the rejection of claim 1 is fully applicable here (see also Drake et al. para [0049]; “Another aspect of the present disclosure provides a non-transitory computer-readable medium comprising machine-executable code that, upon execution by one or more computer processors, implements any of the methods above or elsewhere herein”).
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Drake et al. in view of Feala et al. as applied in claim 1 above, and further in view of Hegde et al. et al. (US 20190369098 A1).
Regarding claim 6, the rejection of claim 5 is incorporated herein.
The combination of Drake et al. and Feala et al. further teach wherein the patient is a first patient (see Drake et al. para [0086]; “FIGS. 12E-12H show PCA of cfDNA, CpG methylation, cf-miRNA and protein counts as a function of patient diagnosis”), the method further comprising: predicting, using the trained first and second module of the machine learning model, activity of at least one of the subset of molecular analytes from a medical image of a second patient (see para [0359]; “The machine learning predictor may be trained using datasets e.g., datasets generated by performing multi-analyte assays of biological samples of individuals) from one or more sets of cohorts of patients having cancer as inputs and known diagnosis (e.g., staging and/or tumor fraction) outcomes of the subjects as outputs to the machine learning predictor”). However, the combination of Drake et al. and Feala et al. as a whole does not teach identifying, based on the predicted activity, an Antibody-Drug Conjugate (ADC) therapy for the second patient.
In the same field of endeavor Hegde et al. teach identifying, based on the predicted activity, an Antibody-Drug Conjugate (ADC) therapy for the second patient (see para [0167]; “a sample from an individual is compared, e.g., to make a predictive, diagnostic, prognostic, and/or therapeutic determination”, see also para [0512]; “an immunoconjugate is an antibody-drug conjugate (ADC) in which an antibody is conjugated to one or more drugs”). Accordingly, it would have been obvious to one of ordinary skill in the before the effective filing date of the claimed invention to modify the general system uses machine learning (ML) to analyze multiple analytes of Drake et al. in view of the use a method for a machine learning used to generate models that identify associations between amino acid sequences and protein functions based on amino acid sequence information of Feala et al. and diagnostic methods, therapeutic methods, and compositions for the treatment of cancer of Hegde et al. in order to conjugated to a radioactive atom to form a radioconjugate (see para [0167]).
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Drake et al. in view of Feala et al. as applied in claim 1 above, and further in view of An et al. NPL “A deep learning model designed for Raman spectroscopy with a novel hyperparameter optimization method)”.
Regarding claim 13, the rejection of claim 1 is incorporated herein. The combination of Drake et al. and Feala et al. additionally teach a multilayer perception including multiple layers of neurons (see para 0437]; “the classifying of the biological sample is performed by a classifier trained and constructed according to one or more of…. multilayer perceptron” multilayer perceptron include multiple layers of neurons including final layer to compute an output, however, the combination of Drake et al. and Feala et al. does not specifically disclose heads.
In the same field of endeavor An et al. teaches wherein the second module of the machine learning model comprises one or more heads (see page 2, section “2.3.2. MLP head”, and Fig. 2”). Accordingly, it would have been obvious to one of ordinary skill in the before the effective filing date of the claimed invention to modify the general system uses machine learning (ML) to analyze multiple analytes of Drake et al. in view of the use a method for a machine learning used to generate models that identify associations between amino acid sequences and protein functions based on amino acid sequence information of Feala et al. and determine vibrational modes of molecules and provide a structural fingerprint by which molecules can be identified of An et al. in order to measure the pH prediction value corresponding to the input spectrum (see page 2, section 2.3.2).
Conclusion
THIS ACTION IS MADE FINAL. 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 WINTA GEBRESLASSIE whose telephone number is (571)272-3475. The examiner can normally be reached Monday-Friday9:00-5:00.
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, Andrew Bee can be reached at 571-270-5180. 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.
/WINTA GEBRESLASSIE/ Examiner, Art Unit 2677
/ANDREW W BEE/ Supervisory Patent Examiner, Art Unit 2677