# US 12,763,212 B2 · Methods for predicting relative benefit of therapy options and devices thereof

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,763,212 B2
- Document ID: US12763212B2
- Status: Granted
- Kind: B2
- Application number: 18/731,179
- Filed: 2024-05-31
- Published: 2024-12-05
- Granted: 2026-10-13
- Earliest claimed: 2023-05-31
- Assignees: Perthera, Inc.
- Applicants: Perthera, Inc.
- Inventors: Emanuel Frank Petricoin, III; Edik Matthew Blais; Hieu Trung Nguyen
- Web page: https://patentgrep.com/patent/US12763212B2

## CPC

- G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA

## Abstract

Methods, non-transitory computer-readable media, and devices are disclosed that identify and validate a signature for predicting relative benefit of two therapies for frontline and therapy sequencing for patients. Described and illustrated by way of the example herein are machine learning approaches for gaining data-driven insights from the mutational landscape in metastatic pancreatic adenocarcinoma (mPDAC) and validating the signature in predicting relative benefit from FOLFIRINOX (FFX) and Gemcitabine/nab-Paclitaxel (GA) therapies. This technology inputs a patient's genomics findings and clinical data and generates predictions of relative effectiveness for the two distinct FFA and GA chemotherapy options. The predictions for an individual patient provide personalized guidance on treatment sequencing to improve patient health outcomes.

## Claims

### Claim 1

A method for improved machine learning modeling for predicting relative benefit of therapy options for a condition, the method comprising:
generating training cohorts and validation cohorts for each of first and second therapy options for a plurality of patients based on training and validation data for the patients obtained via one or more communication networks from one or more data server devices and comprising at least first molecular and clinical data;
generating an ensemble machine learning model, comprising:
sampling training and cross-validation subsets from each of the training cohorts and validation subsets from each of the validation cohorts comprising excluding from the training cohorts and the validation cohorts a set of other patients that received treatment including one or more agents aligned to both the first and second therapy options;
training machine learning models using the training and cross-validation subsets, wherein the machine learning models comprise random survival forest (RSF) models trained to predict treatment response data for each of the first and second therapy options;
validating the machine learning models using treatment-specific response data in the validation subsets to determine that the machine learning models exceed an accuracy threshold, wherein the treatment-specific response data comprises first progression-free survival data; and
combining first and second subsets of the machine learning models into the ensemble machine learning model, wherein at least a subset of models in each of the first and second subsets of the machine learning models vary by tree depth and parameter set as a result of the training;
applying the machine learning model to patient data for a patient to generate first and second treatment response data, each corresponding to one of a plurality of categories for each of the first and second therapy options, wherein the patient data is obtained via the communication networks from a client device;
generating, and outputting via the communication networks for display via a display device of the client device, a graphical user interface comprising a representation of the first and second treatment response data with respect to the categories for the first and second therapy options, respectively, to thereby inform treatment of the patient for the condition; and
treating the patient in accordance with one of the first or second therapy options as a frontline therapy based on a relative value of the first and second treatment response data, wherein the first and second therapy options comprise Gemcitabine/nab-Paclitaxel (GA) and FOLFIRINOX (FFX) chemotherapeutic agents and the treatment comprises administering chemotherapy to the patient via the GA or FFX chemotherapeutic agents to treat pancreatic adenocarcinoma (PDAC).

### Claim 2 (depends on 1)

The method of claim 1, wherein the first molecular and clinical data further comprises, for each of the patients, clusters of genomic features, second progression-free survival data, and an indication of a frontline treatment corresponding to one of the first or second therapy options.

### Claim 3 (depends on 1)

The method of claim 1, wherein the first therapy option or second therapy option comprises gemcitabine and nab-paclitaxel, 5-Fluorouracil+nal-Irinotecan+Oxaliplatin, 5-Fluorouracil+Irinotecan+Oxaliplatin, 5-Fluorouracil+Oxaliplatin, 5-Fluorouracil+Irinotecan, or 5-Fluorouracil+nal-Irinotecan, and the condition comprises pancreatic adenocarcinoma.

### Claim 4 (depends on 1)

The method of claim 1, wherein the patient data comprises second molecular and clinical data for the patient and the first and second treatment response data comprises first and second personalized therapy response predictions for the patient with respect to the first and second therapy options, respectively.

### Claim 5 (depends on 1)

The method of claim 1, wherein each of the patients is associated with one of two therapy-specific reference populations and the categories correspond to tertiles representing underperformance, expected performance, and overperformance with respect to portions of the therapy-specific reference population corresponding to one of the first or second therapy options, respectively.