Systems and methods for genomic data extraction and medical event predictions

A trained machine learning model, specifically a large language model with a self-supervised transformer for time series, addresses the inefficiencies in processing genetic reports and irregular data, achieving improved predictive performance for medical events by 10 percentage points, particularly in chronic myelomonocytic leukemia.

WO2026117740A1PCT designated stage Publication Date: 2026-06-04MAYO FOUNDATION FOR MEDICAL EDUCATION & RESEARCH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
MAYO FOUNDATION FOR MEDICAL EDUCATION & RESEARCH
Filing Date
2025-11-26
Publication Date
2026-06-04

AI Technical Summary

Technical Problem

Current techniques for generating predictions associated with medical events, such as chronic myelomonocytic leukemia, face inefficiencies due to the complexity of genetic reports, non-standardized formats, sparse and irregularly sampled multivariate data, and the inability to handle such data effectively, leading to inaccurate results.

Method used

The use of a trained machine learning model, particularly a large language model, to process genetic reports, medical test data, and demographic data, utilizing a self-supervised transformer for time series (STraTS) to handle sparse and irregular data, improving predictive performance by approximately 10 percentage points.

Benefits of technology

The techniques provide more accurate predictions, enhancing the efficiency of genetic mutation data extraction and improving predictive performance by approximately 10 percentage points compared to conventional methods, with a concordance index of about 0.7 for survival prediction of chronic myelomonocytic leukemia.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure US2025057363_04062026_PF_FP_ABST
    Figure US2025057363_04062026_PF_FP_ABST
Patent Text Reader

Abstract

Systems and methods for generating a prediction associated with a medical event include receiving a genetic report associated with a patient; generating, by processing the genetic report using a first machine learning model trained to identify one or more genetic mutations from the genetic report, genetic mutation data; receiving medical test data associated with the patient including irregularly sampled multivariate time-series data; receiving demographic data associated with the patient; training a second machine learning model on historical patient data to predict an occurrence of a medical event; generating, by processing the genetic mutation data, the medical test data, and the demographic data using the second machine learning model trained to predict the occurrence of the medical event, a medical event prediction associated with the patient; and providing the medical event prediction to a user device.
Need to check novelty before this filing date? Find Prior Art