Method and system for assessing drug efficacy using multiple graph kernel fusion
Inventors
Frieder, Ophir • YAO, Hao-Ren • CHANG, Der-Chen
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12211623-B2
Publication Date
2025-01-28
Expiration Date
Abstract
Embodiments of the present systems and methods may provide techniques to predict the success or failure of a drug used for disease treatment. For example, a method of determining drug efficacy may include, for a plurality of patients, generating a directed acyclic graph from health related information of each patient comprising nodes representing a medical event of the patient, at least one first edge connecting the first node to an additional node, each additional edge connecting nodes representing two consecutive medical events, the edge having a weight based on a time difference between the two consecutive medical events, capturing a plurality of features from each directed acyclic graph, generating a binary graph classification model on captured features of each directed acyclic graph, determining a probability that a drug or treatment will be effective using the binary graph classification model, and determining a drug to be prescribed to a patient based on the determined probability.
Core Innovation
The invention provides a method of computing a probable drug efficacy implemented in a computer system. For each of a plurality of patients, the method generates and stores a graph comprising nodes that represent medical events and edges connecting two consecutive medical events, where each edge weight is based on a time difference between the two consecutive medical events. The approach captures features from each patient graph by generating a topological ordering based on an order of occurrence of a label associated with each node and generating a topological sequence comprising levels indicating an order of occurrence of a same node label.
From each graph, the method generates a temporal signature comprising a series of total passage times between nodes in a union of a plurality of topological sequences. Using the temporal signatures, it generates a temporal proximity kernel between pairs of temporal signatures. It further generates a shortest path kernel by calculating an edge walk similarity on at least one shortest path graph for pairs of graphs and generates a node kernel by comparing node labels of pairs of graphs.
The temporal proximity kernel, the shortest path kernel, and the node kernel are fused, and a classifier model is trained using the captured features of each graph. The trained classifier model is applied to determine a probability that a drug or treatment will be effective for a particular patient. The determined probability is then used to determine a drug or treatment to be prescribed to the particular patient.
Claims Coverage
The provided partial content includes three independent claims: a method, a system, and a computer program product. Across these independent claims, the central inventive features include patient-specific time-weighted medical-event graphs, feature extraction using topological/temporal ordering and multiple graph kernels, kernel fusion, classifier training, probabilistic efficacy determination, and prescribing based on the determined probability.
Patient medical-event graph with time-difference weighted edges
For each of a plurality of patients, generating and storing in the memory a graph comprising nodes representing medical events and edges connecting two consecutive medical events, where each edge has a weight based on a time difference between the two consecutive medical events.
Temporal and topological feature capture from graph structure
Capturing a plurality of features from each graph by generating a topological ordering based on an order of occurrence of a label associated with each node, generating a topological sequence comprising levels indicating an order of occurrence of a same node label, and generating a temporal signature comprising a series of total passage times between nodes in a union of a plurality of topological sequences.
Kernel construction from temporal signatures, shortest paths, and node labels
Generating a temporal proximity kernel between a plurality of pairs of temporal signatures, generating a shortest path kernel by calculating an edge walk similarity on at least one shortest path graph for a plurality of pairs of graphs, and generating a node kernel by comparing node labels of a plurality of pairs of graphs.
Fusion of temporal proximity, shortest path, and node kernels
Fusing the temporal proximity kernel, the shortest path kernel, and the node kernel.
Classifier training and probabilistic drug efficacy determination
Training a classifier model using a plurality of the captured features of each graph and applying the classifier model to determine a probability that a drug or treatment will be effective for a particular patient.
Prescribing based on determined drug efficacy probability
Determining, using the processor, a drug or treatment to be prescribed to the particular patient based on the determined probability.
All three independent claims cover the same core pipeline: build a patient medical-event graph with time-difference weighted edges, compute temporal and topological features including temporal signatures, construct temporal proximity/shortest path/node kernels, fuse kernels, train a classifier, output a probability of drug effectiveness, and determine prescribing based on that probability.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?