Genotoxic mode of action predictions from a multiplexed flow cytometric assay and a machine learning approach

S.M. Bryce, D.T. Bernacki, J.C. Bemis, S.D. Dertinger

Litron Laboratories

Environmental and Molecular Mutagenesis
Volume 57, Issue 3, April 2016, Pages 171–189

Creating a high throughput assay that provides both mode of action (MoA) information and high specificity would solve many of the problems currently plaguing in vitro genotoxicity tests. Described below is a multiplexed, one step, add-and-read assay developed by Litron Laboratories that addresses these challenges. This “MultiFlow” method uses flow cytometry and machine learning to classify compounds as clastogens, aneugens or non-genotoxicants.

MultiFlow was designed to provide information on several biomarkers that are highly relevant to both DNA damage and repair pathways:

  • γH2AX is a marker of DNA double strand breaks
  • Phospho-histone H3 is a marker of mitotic cells
  • p53 is an indicator of general genotoxic stress

For this study we examined the effects of 67 reference compounds in human lymphoblastoid TK6 cells. Cell cultures were grown, treated, processed and analyzed in 96 well plates. The data from the various endpoints were analyzed by logistic regression to provide clastogenic, aneugenic or non-genotoxicant probabilities. The MultiFlow approach achieved 94% concordance based on a priori classification of MoA chemicals.

Overall, these studies demonstrate the benefits of using a high content flow cytometric assay combined with machine learning to classify compounds according to their genotoxic MoA. This information is highly valuable for improved decision-making and human health risk assessment.