PII-009 - NOVEL IN VITRO TUBERCULOSIS ASSAYS AND THEIR UTILITY IN PREDICTING IN VIVO DRUG EFFICACY.
Thursday, March 23, 2023
5:00 PM – 6:30 PM EDT
J. Goh1, E. Nuermberger2, V. Dartois3, R. Savic1; 1University of California, San Francisco, San Francisco, CA, USA, 2Johns Hopkins University, Baltimore, MD, USA, 3Hackensack Meridian Health, Nutley, NJ, USA.
University of California, San Francisco San Francisco, California, United States
Background: Multiple in vitro assays mimicking different tuberculosis (TB) infection niches are used to test a new compound's efficacy. However, there is no consensus on which assays are most predictive of in vivo drug potency. Earlier, we showed that in vivo EC50 is portable between mouse models and patients with TB after adjusting for protein binding and can be used to predict dose-response in early clinical trials. Our goal is to extend this translational platform and develop a further link between in vitro assay results and in vivo estimates of EC50 from mouse infection models. Methods: Data from 31 unique in vitro assays were compiled from literature and collaborators for 9 first- and second-line TB drugs. Feature selection based on assay correlation and individual assay predictiveness to in vivo EC50 resulted in 4 main clusters. One cluster was not predictive. From the remaining 3, the top 2-3 positive correlations were selected, resulting in 10 features for multiple linear regression. Results: Multiple linear regression to predict in vivo EC50 had good predictions in both the training and testing sets (R2train = 0.865, RMSEtrain = 0.69, R2test = 0.837, RMSEtest = 0.63). The minimum number of features necessary to achieve a similar level of prediction proved to be a combination of 3 assays: an ex vivo rabbit caseum assay, a human macrophage model, and a high cholesterol assay (R2 = 0.926, RMSE = 0.480). Our findings reflect the heterogeneous niches of TB and the need for complementary assays to best predict in vivo efficacy. Conclusion: These recommendations on in vitro assays to conduct prior to animal studies can help to prioritize which drugs are the most likely to succeed clinically, saving resources and accelerating development.