System and method for biomarker-outcome prediction and medical literature exploration

Inventors

Krasnoslobodtsev, Artem • Backis, Danius Jean • Babakhani, Pouya • Jo{hacek over (c)}ys, {hacek over (Z)}ygimantas • Tal, Roy • Knuff, Charles Dazler

Assignees

RO5 Inc

Abstract

A system and method for biomarker-outcome prediction and medical literature exploration which utilizes a data platform to analyze, optimize, and explore the knowledge contained in or derived from clinical trials. The system utilizes a knowledge graph and data analysis engine capabilities of the data platform. The knowledge graph may be used to link biomarkers with molecules, proteins, and genetic data to provide insight into the relationship between biomarkers, outcomes, and adverse events. The system uses natural language processing techniques on a large corpus of medical literature to perform advanced text mining to identify biomarkers associated with adverse events and to curate a comprehensive profile of biomarker-outcome associations. These associations may then be ranked to identify the most-common biomarker-outcome association pairs. Having a comprehensive profile of ranked biomarker-outcome data allows the system to predict biomarkers associated with a given disease and serious adverse events linked to biomarker data.

Core Innovation

The invention provides a biomarker-outcome prediction and medical literature exploration system that uses a data platform with a knowledge graph and a data analysis engine including natural language processing. The system is configured for automated text mining of clinical trial publications and other medical literature to identify relationships between biomarkers and outcomes, including adverse events and outcomes. The mined relationships are represented as biomarker-adverse event associations and biomarker-outcome relationships derived from text.

A computing system receives a clinical trial publication and automatically scrapes the clinical trial publications for biomarker and outcome word pairs. The system computes co-occurrence statistics by computing a total number of times the outcome word appears after the biomarker word in a window of words for all clinical trial publications. The system then computes probabilities of occurrence for each biomarker and for each outcome by dividing the computed total counts by the total number of publications in the database, and computes the probability of each biomarker occurring with each outcome word.

The system calculates an association score for each biomarker and outcome word pair using normalized pointwise mutual information based on the computed probability of occurrence for each biomarker, the probability of occurrence for each outcome, and the probability each biomarker occurring with each outcome. The platform supports exploring and ranking biomarker associations derived from the computed association scores and provides an interactive clinical trial explorer with navigable graph and map interfaces and explanatory context between biomarkers and outcomes.

Claims Coverage

The document provides two independent claims directed to calculating an association score between a biomarker and an outcome using automated text mining of clinical trial publications and a normalized pointwise mutual information association calculation. Across both independent claims, the inventive features focus on extracting biomarker-outcome word pairs, computing windowed co-occurrence counts, deriving occurrence probabilities, and computing an association score from those probabilities via normalized pointwise mutual information.

Both independent claims cover an end-to-end association scoring approach that starts with automated scraping of clinical trial publications to extract biomarker and outcome word pairs, proceeds to windowed co-occurrence counting and probability computation using counts normalized by the total number of publications, and finishes by computing an association score via normalized pointwise mutual information for each biomarker-outcome word pair.

Stated Advantages

Documented Applications