Hierarchical cluster analysis and nonlinear mixed-effects modelling for candidate biomarker detection in preclinical models of cancer.

Background: Preclinical models of cancer can be of translational benefit when assessing how different biomarkers are regulated in response to particular treatments. Detection of molecular biomarkers in preclinical models of cancer is difficult due inter-animal variability in responses, combined with limited accessibility of longitudinal data.

Methods: Nonlinear mixed-effects modelling (NLME) was used to analyse tumour growth data based on expected tumour growth rates observed 7 days after initial doses (DD7) of Radiotherapy (RT) and Combination of RT with DNA Damage Response Inhibitors (DDRi). Cox regression was performed to confirm an association between DD7 and survival. Hierarchical Cluster Analysis (HCA) was then used to identify candidate biomarkers impacting responses to RT and RT/DDRi and these were validated using NLME.

Results: Cox regression confirmed significant associations between DD7 and survival. HCA of RT treated samples, combined with NLME confirmed significant associations between DD7 and Cluster specific CD8 Ki67 MFI, as well as DD7 and cluster specific Natural Killer cell density in RT treated mice.

Conclusion: Application of NLME, as well as HCA of candidate biomarkers may provide additional avenues to assess the effect of RT in MC38 syngeneic tumour models. Additional studies would need to be conducted to confirm association between DD7 and biomarkers in RT/DDRi treated mice.