Pharma R&D Today

Ideas and Insight supporting all stages of Drug Discovery & Development

Select category
Search this blog

Connecting the data sources to understand the safety of bioactivities

Posted on June 16th, 2017 by in Pharma R&D

Digital_monitor-1307227__340

Earlier this week, I was in Berlin, Germany speaking for the Elsevier Professional Services team at the Drug Discovery Summit (Monday, June 12, 15:30) at the Hotel Palace Berlin on informatics analyses to provide early warning for clinical issues from bioassay results.A study published in Nature Reviews Drug Discovery suggests that 82% of preclinical closures, and 62% in Phase I are due to safety concerns.

DDD summit 1

This illustrates the need for better predictive tools to help reduce failure by a) early closure of projects, b) alteration of trial inclusion criteria, and c) modification of patient safety plans to reduce patient risk.

The Elsevier services team have been studying ways to link clinical effects mined from regulatory filings to the reported bioassays of drugs to find if any bioassays appear to be statistically related to the clinical effects. This effort relies on high-quality data normalization from text mining of regulatory filings, journals, and patents to create standard names for adverse events, drugs, and protein targets.

The event–target associations are measured with 2×2 contingency matrices, using the same statistical analyses used to determine the efficacy of medical diagnostic tests. The statistical strength of the relationship was measured with the chi-squared statistic, and the risk implied by target binding measured with the likelihood ratio.

The 2×2 table below is an example for the relationship between HRH1, histamine receptor binding, and QT prolongation.

DDD summit 2

The chi-squared metric for this relationship is 165, indicating a very high confidence level for the association. The likelihood ratio suggests that if a drug is observed to inhibit HRH1 the likelihood of observing QT prolongation is increased by a factor of 9.

We have created predictive models for targets such as HRH1 to improve the model for the set of drugs where the HRH1 assay was not performed or reported. The relationships were also supported by pathway analyses to identify any known biological pathways connecting the target and event to distinguish ‘real’ relationships from spurious ones.

This analysis was carried out over all adverse event – drug – target combinations — nearly 10 million combinations — to identify patterns and relationships to help evaluate drug safety. The results were analyzed in tables with the most highly predictive targets, as well as with network diagrams to understand how drugs and targets interact.

Liver toxicity is a common cause of drug failure. The image at the head of this article shows a network of statistically significant relationships around drugs and targets implicated to be related to cholestasis and cholestatic jaundice. The drugs in the diagram are those that have been observed to have those adverse events at therapeutic doses in FDA filings.


 

All opinions shared in this post are the author’s own.

R&D Solutions for Pharma & Life Sciences

We're happy to discuss your needs and show you how Elsevier's Solution can help.

Contact Sales
Matthew Clark


Life Sciences R&D Solution Consultant

Connect on LinkedIn
https://www.elsevier.com/rd-solutions