Skip to main content
2023 BMES Annual Meeting Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back

Bioinformatics, Computational and Systems Biology

Translational systems modeling

Computational Translation of Mouse Models of Osteoarthritis Predicts Human Disease Identifies Age and Sex-Dependent Cross-species Disease Biology

Saturday, October 14, 2023
1:00 PM – 1:15 PM PDT
Location: 331 (Center, Level 3)
Introduction:: Osteoarthritis (OA) affects an estimated 250 million people worldwide and its prevalence continues to rise as obesity rates increase and the population ages [1]. Animal models of OA, which involve surgical, mechanical, or biochemical disruption to a synovial joint, are necessary to study the mechanisms underlying the development and progression of OA and to evaluate interventions [2]. Direct translation of observations from preclinical models to humans remains challenging and confounds understanding of human OA mechanisms and the development of interventions. The objectives of our study were to evaluate two models of mouse PTOA, destabilization of the medial meniscus (DMM) and noninvasive anterior cruciate ligament rupture (ACLR), against available human OA samples to determine which human features could be predicted by preclinical data.

Materials and Methods:: Methods were based on conditioning mouse-human translational modeling on human variables through interaction effects. Gene Expression Omnibus was used to identify human Control vs. arthroscopic partial meniscectomy (APM) and Control vs. Osteoarthritic (OA) expression data and mouse injury datasets including Control vs. ACLR and Sham vs. DMM. Feature selection of genes was conducted by performing generalized linear modeling on mouse data. Genes with a q-value less than 0.2 after false discover rate (FDR) were preserved. Principal component analysis (PCA) was run on both sets of mouse data to decrease dimensionality of data while identifying latent variables. Human data was projected on mouse principal components (PCs) through multiplication by PC loadings. Down sampling was performed on significant PCs before principal component regression (PC-R) was performed to include interaction effects between PCs and human covariates. This allows us to stratify biology predicted to translate to specific human populations and mouse biology that can be explained by human disease risk factors. Significant PCs of these models were plotted against each other for visual data clustering comparison.

Results, Conclusions, and Discussions:: Results and Discussion: Initially, we built smaller PC-R models using human data projection on mouse PCs to predict human disease status, age, and sex. This allowed us to down sample the selection of mouse PCs to those most relevant for human translation in the final TransComp-R models. The ACLR Control vs. OA model had significant PC1 (p = 0.021, OA), and PC4 (p = 0.041, sex). TransComp-R models combining mouse PCs with human covariates resulted in p-values for the selected PCs, Age, Sex, and PCs combined with these two clinical factors. Though a larger number of significant PCs was seen in the DMM model than the ACLR model, the lower ranks represented a smaller influence on the variance explained in the data by that latent variable. The significance of PC1 in the ACLR Control vs. OA case presents evidence for the cross-species modeling of OA by mice who undergo ACLR. Age was validated as statistically significant across models both through visual clustering of points on PCA plots and through p-values in models produced from TransComp-R analysis. Of note, human sex was identified as a significant covariate in ACLR TransComp-R modeling despite the procedure being performed on only male mice. These differences suggest that predictions including mouse PC1 interactions with human sex are stronger than those based on human data alone.

Conclusion: Preclinical studies of OA must take care when selecting an animal model to evaluate disease progression or assess treatments. Using TransComp-R, we identify gene signatures in two established OA mouse models that predict human OA and show that different models encode distinct axes of translatable biology that depend on human age as a risk factor and sex as a biological variable. The use of ACLR to model osteoarthritis in mice does have significant latent variables that encompass more variance in biological signals than that of DMM models. These results encourage future studies to validate findings in larger data sets and further consider the role of sex-related injury response mechanisms to injury.


Acknowledgements (Optional): :

References (Optional): :
[1] Hunter. (2019). Osteoarthritis. The Lancet., 393(10182), 1745–1759. https://doi.org/10.1016/S0140-6736(19)30417-9




[2] Fang, Hang, and Frank Beier. "Mouse models of osteoarthritis: modelling risk factors and assessing outcomes." Nature Reviews Rheumatology 10.7 (2014): 413-421.