INSPIRE: An Integrated Agent Based System for Hypothesis Generation within Cancer Datasets

Cancer research has become an extremely data rich environment, with huge batteries of tests performed to quantify and categorise tumours. Multiple analyses of biochemical, molecular and immunohistological markers on tissue samples generate large and complex data sets to compare with clinical and pathological parameters. The manual data analysis procedures by scientists have become impractical and automation is becoming the only method for complete and comprehensive analysis of the search space. To this end, a multi-agent system has been developed to automate the data analysis process. Project agents are tasked with overseeing the analysis. Initially, they create a list of hypotheses based on all parameters associated with the project. Agents then commence data collection and aggregation by directly interfacing with the various data sources. Conventional statistical tests can then be performed under agent control to determine the significance of these hypotheses. The final stage is to present collated results online. INSPIRE uses automated data retrieval and analysis on large and diverse cancer datasets to allow for the quick identification of significant results amongst the noise of the large dataset. This results in a more streamlined research process, which makes large cohort, multivariate projects easier to manage in a secure, user-friendly, web-based data management system.