Notice of Pre-AIA or AIA Status
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Detailed Action
2. This Final Office Action is responsive to Applicants’ Reply dated 4/23/26. Claims 1-12 remain pending, of which claims 1 and 7 are independent.
In this present action, the grounds for rejection have been reformulated as necessitated in accordance with further search and consideration in view of Applicants’ amendments.
Claim Rejections - 35 USC § 103
3. 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.
4. 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.
5. 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.
6. Claims 1-12 are rejected under 35 U.S.C. 103 as being unpatentable over Non-Patent Literature “eToxPred: a machine learning-based approach to estimate the toxicity of drug candidates” (“Pu”) in view of Non-Patent Literature “Detecting Potential Adverse Drug Reactions Using a Deep Neural Network Model” (“Wang”) and Non-Patent Literature “Missing Data and Multiple Imputation” (“CPH”).
Regarding claim 1, PU teaches A processor-implemented adverse drug reaction detection (Abstract: seeking to improve efficiency of drug development in launching new pharmaceuticals, specifically through computational modelling and machine learning algorithms to evaluate drug candidates in terms of their toxicity, where the employment of modelling and ML algorithms as mentioned is equivalent to a “processor-implemented” approach as recited, and where evaluation/prediction of toxicity as mentioned is equivalent to “adverse drug reaction” as recited (see, e.g., page 2, 1st column, bottom paragraph: “... toxicity prediction is an increasingly important component of modern CADD [Examiner: computer-aided drug discovery], especially considering that the collections of virtual molecules for VS may comprise tens of millions of untested compounds. Methods to predict toxicity aim at identifying undesirable or adverse effects of certain chemicals on humans, animals, plants, or the environment.”)) method, the method comprising:
receiving raw data including information on adverse events of a plurality of patients with regard to a target drug and classifying the received raw data into first data corresponding to adverse drug reactions of the target drug, second data corresponding to no adverse reactions of the target drug and drugs similar to the target drug ... and learning a machine learning model by calculating and obtaining a correlation between the target drug and adverse events related to the target drug included in a gold standard dataset that includes data corresponding to the first data and the second data among the received raw data (see the discussion under the heading “Model training, cross-validation, and evaluation”, specifically on page 6, 1st column, 1st full paragraph: “The Tox-score prediction is conducted with a binary, ET-based classifier. The training and cross-validation are carried out for the FDA-approved dataset used as positive (non-toxic) instances and the TOXNET dataset used as negative (toxic) instances. Subsequently, the toxicity predictor is trained on the entire FDA-approved/TOXNET dataset and then independently tested against the KEGG-Drug (positive, non-toxic) and T3DB (negative, toxic) sets.”,
where the positive and negative datasets mentioned here are akin to the classified “second data” and classified “first data” as recited, and the concrete “instances” mentioned that populate the aforementioned datasets are akin to the recited “raw data” indicative of adverse events (or not) relating to actual patients with respect to actual drugs, and where the Examiner understands the reference to teach the use of the positive and negative classifications to train the framework’s ML algorithm to generate the “Tox-score prediction”,
where the reference’s acquisition of information that is already classified into first and second data classes, as noted above, to be functionally equivalent in establishing ground truth / gold set information for training a machine learning to the claim’s receiving raw data and classifying it into the first and second data classes); and
determining a probability of adverse reactions for a prediction dataset ... based on the machine learning model (staying with page 6, 1st column, 1st full paragraph: “In addition, the capability of the classifier to predict specific toxicities is assessed against CP, CD, ED, and AO datasets.”, but even apart from that, the Examiner understands the trained framework to be useful in the evaluation of toxicity prediction (and hence adverse drug reaction) as pertaining to new drugs, e.g. as discussed above in relation to the claim’s preamble (citing to the Abstract), and where the Tox-score prediction is explicitly a probability as discussed in the section ‘Tox-score prediction with eToxPred’ (page 7, 2nd column): “It employs an ET classifier to compute the Tox-score ranging from 0 (a low probability to be toxic) to 1 (a high probability to be toxic).”),
wherein the received raw data is classified based on a database that is independent of the machine learning model, prior to the learning of the machine learning model (reiterating the citation provided above, from page 6, 1st column, 1st full paragraph: “The Tox-score prediction is conducted with a binary, ET-based classifier. The training and cross-validation are carried out for the FDA-approved dataset used as positive (non-toxic) instances and the TOXNET dataset used as negative (toxic) instances.”).
Pu does not appear to explicitly teach the further limitations for:
classifying the received raw data into ... third data comprising data that is neither the first data nor the second data
determining a probability of adverse reactions for a prediction data set based on the machine learning model, as discussed above, but where the prediction data set is including the third data
At best, Pu teaches validating the trained model against other data (i.e., not the training data as discussed above), which could be equivalent to the recited “third data” (since it would not be included as either the first or second data as recited/mapped). See, again, page 6, 1st column, 1st full paragraph (“In addition, the capability of the classifier to predict specific toxicities is assessed against CP, CD, ED, and AO datasets.”), or Pu’s applicability of its trained framework to any new drug being considered for market (which Pu contemplates as a use case, per its Abstract) and which would then not be even available for use as training data but rather would be a useful inference case application for the model/framework as developed. The Examiner reasons that either of these scenarios could possibly read on Applicants’ recitation of third data.
However, to the extent that these aspects of Pu as addressed above are not sufficient to fully read on the limitations addressed immediately just above, the Examiner further relies upon WANG and CPH to teach what Pu otherwise lacks, see e.g., the following:
Wang also teaches a ML-driven adverse drug reaction framework that is trained on known ADR information and then used to provide useful predictions for new drugs. See “Abstract” and “Introduction” sections on pages 1-2. Moreover, Wang contemplates an ingestion and modelling aspect that differentiates between known and unknown data, as processed. See, e.g., “Adverse Drug Reaction Detection Deep Neural Network Model Description”, on page 4, citing to Figure 2, where the framework clearly has the capability to encode/embed for both certainty scenarios in a way that would read on Applicants’ recitation of a classified third data that was not classified as either the first/negative or second/positive data.
CPH teaches a modeling consideration where information is either categorizable as known/available or missing, e.g. similar to Wang’s aspect discussed just above. See CPH’s Overview. CPH further teaches imputation as a viable way to fill in missing data with plausible value. See, e.g., item 4 on page 4. The Examiner reasons that the prediction aspects of both Pu and Wang are a manner of imputing a plausible value, based on a trained framework, to complete the information that is missing, e.g. a new drug without known actual data/history per both Pu and Wang.
Pu and Wang both relate to adverse drug reaction, or similar, predictions for new drug products. Hence, they are similarly directed and therefore analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Wang’s flexibility to actively embed/encode for missing/new data, e.g. for a new drug without known history, into a framework such as Pu’s, with a reasonable expectation of success, such that the prediction motivation expressed by Pu and Wang could be realized in a manner of filling that particular embedding/encoding, as CPH suggests in more literal fashion.
Regarding claim 2, Pu in view of Wang and CPH teach the method of claim 1, as discussed above, and further wherein the first data, the second data and the third data are classified based on the database including information about the adverse reactions of the target drug and the drugs similar to the target drug based on a predetermined standard (Pu’s page 9, 1st column, before the first full paragraph: “According to Fig. 5B, a Tox-score of 0.58 the most effectively discriminates between toxic and non-toxic molecules, yielding an MCC (Eq. 4) of 0.52. Employing this threshold gives a high TPR of 0.71 at a low FPR of 0.19.”). The motivation for combining the references is as discussed above in relation to claim 1.
Regarding claim 3, Pu in view of Wang and CPH teach the method of claim 1, as discussed above, and further wherein the machine learning model is a learning model that implements one of a gradient boosting machine and a random forest algorithm (Pu’s page 4, 1st column, 3rd full paragraph discussing the use of a random tree algorithm specifically to perform the toxicity prediction, and where the Examiner submits that the algorithm is a type of a random forest algorithm). The motivation for combining the references is as discussed above in relation to claim 1.
Regarding claim 4, Pu in view of Wang and CPH teach the method of claim 1, as discussed above, and further wherein the learning of the machine learning model comprises learning the gold standard dataset to be randomly divided into a training dataset and an evaluation dataset according to a predetermined ratio (beginning on page 5, 2nd column, under the section titled ‘Model training, cross-validation, and evaluation’, k-fold cross-validation approach is mentioned in divvying up the available data into training data and validation data in accordance with a fixed ratio, and a repetition and averaging approach thereto which the Examiner understands to involve randomization to form the different folds/subsets, and on page 6, 1st column, 1st full paragraph, it is expressed that this method is used to evaluate the toxicity predictor/classifier specifically). The motivation for combining the references is as discussed above in relation to claim 1.
Regarding claim 5, Pu in view of Wang and CPH teach the method of claim 4, as discussed above, and further wherein the learning of the machine learning model comprises: first learning the machine learning model by implementing the training dataset (as previously-cited, see the discussion under the heading “Model training, cross-validation, and evaluation”, specifically on page 6, 1st column, 1st full paragraph: “The Tox-score prediction is conducted with a binary, ET-based classifier. The training and cross-validation are carried out for the FDA-approved dataset used as positive (non-toxic) instances and the TOXNET dataset used as negative (toxic) instances. Subsequently, the toxicity predictor is trained on the entire FDA-approved/TOXNET dataset and then independently tested against the KEGG-Drug (positive, non-toxic) and T3DB (negative, toxic) sets.”); and setting a threshold to have a maximum area under the curve (AUC) of a receiver operating characteristics (ROC) curve with the evaluation dataset for the first learned machine learning model (bottom of page 7’s 2nd column, under the heading “Tox-score prediction with eToxPred”, it is discussed that the ROC curve accuracy threshold with AUC is 0.82). The motivation for combining the references is as discussed above in relation to claim 1.
Regarding claim 6, Pu in view of Wang and CPH teach the method of claim 5, as discussed above, and further wherein the determining of the possibility of adverse reactions for the prediction dataset comprises further determining whether there are adverse reactions based on the possibility of adverse reactions of the prediction dataset and the set threshold (bottom of page 7’s 2nd column, under the heading “Tox-score prediction with eToxPred”, it is discussed that the ROC curve accuracy threshold with AUC is 0.82 and that a Tox-score of 0.58 is the most effective discriminator between toxic and not). The motivation for combining the references is as discussed above in relation to claim 1.
Regarding claim 7, the claim includes the same or similar limitations as claim 1 discussed above, and is therefore rejected under the same rationale.
Regarding claim 8, the claim includes the same or similar limitations as claim 2 discussed above, and is therefore rejected under the same rationale.
Regarding claim 9, the claim includes the same or similar limitations as claim 3 discussed above, and is therefore rejected under the same rationale.
Regarding claim 10, the claim includes the same or similar limitations as claim 4 discussed above, and is therefore rejected under the same rationale.
Regarding claim 11, the claim includes the same or similar limitations as claim 5 discussed above, and is therefore rejected under the same rationale.
Regarding claim 12, the claim includes the same or similar limitations as claim 6 discussed above, and is therefore rejected under the same rationale.
Conclusion
7. Applicants’ amendment necessitated the new ground(s) of rejection presented in this Office Action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicants are 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.
8. Any inquiry concerning this communication or earlier communications from the examiner should be directed to SHOURJO DASGUPTA whose telephone number is (571)272-7207. The examiner can normally be reached M-F 8am-5pm CST.
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, Tamara Kyle can be reached at 571 272 4241. 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.
/SHOURJO DASGUPTA/Primary Examiner, Art Unit 2144
Prosecution Insights