TPA

Tumor Protocol Assist (TPA), developed by Neureveal Inc, is an AI-powered clinical decision support tool designed to assist virtual Molecular Tumor Boards (vMTB) in identifying the most appropriate treatment protocols for cancer patients. By combining advanced data retrieval, semantic indexing, and proprietary querying techniques, TPA bridges the gap between complex oncology guidelines and individualized patient data to deliver evidence-based treatment recommendations.

TPA is built for oncology teams, clinical researchers, and institutions involved in medical treatment recommendations, research and development, and clinical trials. The platform's tools and methodologies are aligned with approaches used by major healthcare institutions for similar use cases.

Guideline Retrieval and Data Processing

• Retrieves the latest oncology guidelines from authoritative sources including NCCN, ASCO, and Cancer.gov.
• Extracts and comprehends guideline content, breaking it down into independent, discrete data units.
• Stores data in individual units that are semantically identified and labelled for precise retrieval.
• Builds meaningful and weighted relationships between the various data units to reflect clinical context.
• Indexes and stores the processed data in a secondary database in numerical format for efficient querying.
• Automated data updates ensure the knowledge base remains current whenever guidelines change.

Patient Data Analysis and Protocol Recommendation

• Receives comprehensive patient data and extracts the clinically relevant information needed for protocol matching.
• Applies a proprietary querying approach to search the guideline database using patient-extracted data points.
• Returns multiple candidate responses, each evaluated independently by the model.
• Reviews all candidate responses and determines the single most appropriate treatment protocol for the patient.
• Submits the chosen treatment protocol together with all relevant references from the original guideline sources.
• Performs a faithfulness check on the returned protocol, scoring adherence to guidelines on a scale of 0 to 1 to ensure reliability and transparency.

Platform Architecture and Operations

• Microservice architecture that is fully containerized and orchestrated for scalability and resilience.
• Ongoing remote monitoring of model results with continuous reporting capabilities.
• Continuous model updates, retraining, and improvement to maintain accuracy over time.
• A dedicated clinical trial AI module for cancer research is planned as a future offering.

Related Personalized Medicine Applications

• Lung Cancer Property Graph Database: Enables personalized treatment strategies by analyzing complex interrelationships among genetic mutations, clinical histories, and treatment outcomes.
• Gene Co-expression Networks: Constructs and analyzes gene co-expression networks using RNA-seq data from The Cancer Genome Atlas (TCGA), applying graph algorithms to identify significantly altered genes and relationships across cancer types, supporting the discovery of tumor-specific expression programs and potential therapeutic targets.
• Hetionet: Models drug efficacy and interactions by connecting data on treatments, compounds, diseases, and biological pathways, supporting research into drug repurposing and the identification of new therapeutic targets in oncology.

TPA is delivered via a containerized, orchestrated microservice architecture, making it suitable for deployment in enterprise healthcare environments. The platform is positioned as an end-to-end solution for oncology treatment decision support, clinical research, and cancer-focused clinical trials.