Hybrid modeling techniques to support drug-development and post-market analysis of anti-infectives

Nowadays, decision-making processes are highly assisted by computational predictive platforms in any discipline. In drug development and global health fields, massive use of in silico tools deals with the anticipation of plausible scenarios, which overall inform on best therapeutic options. The thesis exemplifies this model-based contribution by addressing two practical open questions within the anti-infective pipeline. The first case-study regards the support of a novel granuloma-centric approach to anti-tuberculosis (TB) regimen design through the quantification of site-of-action pharmacokinetic attainments of any anti-TB compound to come. In a step-wise manner, an in vivo minimal physiologically-based pharmacokinetic (mPBPK) platform for anti-TB drug disposition was derived from a full-body version and enriched with an innovative intra-granuloma module described via ODE-based passive diffusion mechanisms. The simultaneous support of severalanti-TB agents and successful validation step provided strong indications of the design quality and universality. Afterwards, under a data augmentation paradigm, mPBPK model-based simulations of granuloma-to-plasma partition coefficients in rabbits were used to supplement non-compartmental approaches and aid the training of a minimal signature machine learning (ML) predictive tool. Sitting at the early stage of the drug development, the ML model was developed to require easy-accessible and animal-testing free descriptors of a molecule to forecast its granuloma PK scores and consequent profile. Although the reduced size of the training set, results from the cross validation endorsed the use of the ML model to close the gap from bench to bed. The second part of the thesis covers model-based postmarket therapeutic evaluations with an application coming from the anti-flu vaccine area. A natural history decision tree was applied upon simulations of a multi-group and age-structured SEIR (Susceptible-Exposed-Infected-Recovered) epidemiological model to reconstruct economical tolls of a novel vaccination strategy. In particular, the cost-benefit scenario of the introduction of the adjuvanted QIV (aQIV) vaccine in the elderly Spanish population (65+ years) was analyzed, leveraging specific real data from demographic, viral, and event cost scales. To complement the population-level section, modeling efforts on the COVID-19 pandemic are presented as ancillary activity to endorse the parallel role of non-pharmaceutical interventions in limiting an infection spread.