Predictive assignments that relate to genetic information and leverage machine learning models
Inventors
Glode, Christopher M. • Trunck, Ryan P. • Powers, Rani K. • Lescallett, Jennifer L.
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-10217048-B2
Publication Date
2019-02-26
Expiration Date
Abstract
Systems and methods are provided for performing predictive assignments pertaining to genetic information. One embodiment is a system that includes a genetic prediction server. The genetic prediction server includes an interface that acquires records that each indicate one or more genetic variants determined to exist within an individual, and a controller. The controller selects one or more machine learning models that utilize the genetic variants as input, and loads the machine learning models. For each individual in the records: the controller predictively assigns at least one characteristic to that individual by operating the machine learning models based on at least one genetic variant indicated in the records for that individual. The controller also generates a report indicating at least one predictively assigned characteristic for at least one individual, and transmits a command via the interface for presenting the report at a display.
Core Innovation
A genetic prediction system acquires records that indicate one or more genetic variants determined to exist within an individual. The system includes a genetic prediction server with an interface that acquires the records and a controller that selects one or more machine learning models that utilize genetic variants as input. For each individual in the records, the controller predictively assigns at least one characteristic by operating the one or more machine learning models, using at least one genetic variant indicated in the records for that individual as input to the one or more machine learning models.
The controller analyzes input indicating accuracy of a predictively assigned characteristic, and determines a score for a machine learning model based on the analyzed input and a cost function. The machine learning models comprise multi-layer neural networks in which nodes in different layers are coupled via weighted connections. Nodes in the input layer receive input indicating whether a different genetic variant exists within the individual, and nodes in an output layer provide output predicting whether a different characteristic is predicted for the individual.
The controller revises the weighted connections based on the score, thereby updating the neural network used to generate predictive characteristic assignments. In some implementations, the predicted characteristics are distinct from a phenotype defined by genetic variants already indicated in the records. In some implementations, genetic variants are represented by relative locations, and outputs are generated based on those relative locations, including locations assigned to categorized aspects and/or genomic location groupings. In some implementations, confidence values are determined from the machine learning model outputs and a characteristic is predictively assigned based on a confidence threshold.
Claims Coverage
The independent claims are clm-00001, clm-00008, and clm-00015. Across these independent claims, there are three core inventive groupings: acquiring genetic-variant records for individuals and selecting machine learning models to assign characteristics; using multi-layer neural networks with weighted connections where input encodes whether genetic variants exist and output predicts characteristics; scoring model accuracy using a cost function and revising weighted connections based on the score.
Acquiring genetic-variant records and predictively assigning characteristics using machine learning models
A genetic prediction system and method acquire records indicating one or more genetic variants determined to exist within an individual, select one or more machine learning models that utilize genetic variants as input, and for each individual predictively assign at least one characteristic by operating the one or more machine learning models using at least one genetic variant indicated in the records for that individual as input.
Multi-layer neural networks using genetic-variant existence input and characteristic prediction output
Each of the one or more machine learning models comprises a multi-layer neural network with multiple nodes per layer, where nodes in different layers are coupled via weighted connections; nodes in an input layer receive input indicating whether a different genetic variant exists within the individual, and nodes in an output layer provide output predicting whether a different characteristic is predicted for the individual.
Model scoring with a cost function and revising weighted connections based on the score
The system and method analyze input indicating accuracy of a predictively assigned characteristic, determine a score for a machine learning model based on the analyzed input and a cost function, and revise the weighted connections based on the score.
All independent claims cover acquiring records of genetic variants for individuals, using one or more multi-layer neural-network machine learning models with weighted connections to predict characteristics from genetic-variant existence inputs, and scoring accuracy with a cost function to revise the weighted connections. Dependent claim refinements further constrain predicted characteristics relative to a phenotype, encode genetic variants as relative locations, and apply confidence-threshold logic for characteristic assignment.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?