System and method for automated pharmaceutical research utilizing context workspaces
Inventors
Tal, Roy • Jocys, Zygimantas • Knuff, Charles Dazler
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11264140-B1
Publication Date
2022-03-01
Expiration Date
Abstract
A system and method for an automated pharmaceutical research utilizing contextual workspaces comprising a workspace drive engine, a data analysis engine, one or more machine and deep learning modules, a knowledge graph, and a workspace interface, which can create a virtual research workspace where data files containing biochemical data related to current research can be uploaded, which automatically processes and analyzes the uploaded data file to autonomously extract a plurality of information related to the uploaded data file, which performs various similarity searches on the uploaded data, and which formats and displays all the extracted information in the workspace, such that the workspace may provide a deeper contextualized view of the uploaded biochemical data.
Core Innovation
The invention relates to automated pharmaceutical research utilizing context workspaces. A workspace database stores information related to uploaded biochemical and biomedical data files, extracted information, and saved workspace data. A biochemical database comprises biochemical and biomedical information related to genes, disease, proteins, molecules, drugs, clinical trials, assays, and biological pathways/mechanisms.
A workspace drive engine creates and stores a workspace, receives an uploaded data file, creates and stores a copy of the uploaded data file, and parses the uploaded data file into individual biochemical or biomedical entities. The engine determines whether each individual entity exists within the biochemical database, sends undetected entities to a data analysis engine, and for detected entities extracts all information related to the entity from a knowledge graph.
The engine creates a graph-based representation of the extracted information and receives data related to a decoded vector from the data analysis engine. The data analysis engine inputs the undetected entity into a trained encoder to create an abstract vector representation in an abstract latent space of a three-dimensional convolutional neural network, computes distance between abstract vectors, compares the computed distance to a predetermined similarity threshold, and decodes the abstract vector that indicates similarity, returning data related to the decoded vector for workspace display.
Claims Coverage
The document provides two independent claims, each focusing on automated pharmaceutical research using context workspaces with knowledge-graph extraction and a 3D-CNN latent-space encoder/decoder guided by a predetermined similarity threshold.
Automated pharmaceutical research system with context workspaces and knowledge-graph extraction with 3D-CNN latent-space decoding
A system comprising a workspace database, a biochemical database, a workspace drive engine that creates and stores workspaces, ingests uploaded biochemical or biomedical entities, parses entities, determines detected versus undetected entities against the biochemical database, extracts information from a knowledge graph for detected entities, creates a graph-based representation, and formats decoded-vector data, extracted information, and the graph-based representation for display; and a data analysis engine that encodes undetected entities using a trained encoder in a 3D convolutional neural network latent space, computes abstract-vector distance, compares the distance to a predetermined similarity threshold, and decodes the abstract vector and returns workspace-drive-engine data related to the decoded vector.
Automated pharmaceutical research method with context workspaces and knowledge-graph extraction with 3D-CNN latent-space decoding
A method of creating and storing a workspace, receiving an uploaded data file and storing a copy, parsing the uploaded data file into individual biochemical or biomedical entities, determining detected versus undetected entities against a biochemical database, sending undetected entities to a data analysis engine, extracting all information related to detected entities from a knowledge graph and creating a graph-based representation, receiving decoded-vector-related data from the data analysis engine and extracting all information related to the decoded vector from the knowledge graph, formatting decoded-vector data, extracted information, and the graph-based representation for display in the workspace, inputting the undetected entity into a trained encoder to create an abstract vector representation in a 3D convolutional neural network latent space, computing distance between abstract vectors, comparing the computed distance to a predetermined similarity threshold, and decoding the abstract vector and returning workspace-drive-engine data related to the decoded vector as a response.
Across both independent claims, the inventive core is the combination of context workspaces with uploaded biochemical/biomedical entity parsing and detected/undetected determination against a biochemical database, knowledge-graph-driven extraction and graph-based representation for detected entities, and a data analysis engine using a trained 3D convolutional neural network encoder, latent-space distance thresholding, and decoder to decode information for undetected entities and return formatted results for workspace display.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?