IFAC Conference on Foundations of Systems Biology in Engineering | August, 2022
Developing drug candidates for complex diseases may involve selection from a library of molecules. An investigational product must demonstrate clinical benefit with an acceptable safety profile to be authorized for commercial use. Therapeutic candidates with a mixture of active components must demonstrate each component’s contribution to the mix’s overall efficacy profile.
Combinations with multiple components may prohibit the conduct of complete factorial clinical studies, requiring alternative approaches, including non-clinical models. Published evidence demonstrates the importance of considering efficacy contributions during candidate design and evaluation for complex diseases.
Meta-ranker is one of the few algorithms that facilitates a leaner experimental design and leverages advanced computational methods to reduce the number of experiments while appropriately providing selection power or allowing for interrogation of individual contribution effects per the US FDA’s fixed combination drug product rule (21CFR300.50). Meta-ranking might complement other modeling tools and provide opportunities for regulatory engagement on fixed-combination products. In non-alcoholic steatohepatitis (NASH), a rank-based approach was used previously to compare molecular candidates for multiple cell types and phenotypes, namely primary human hepatocytes, macrophages, and stellate cells.
Meta-ranker calculates investigational product candidates’ ranks (meta-ranks) in an experiment considering the measurements across multiple phenotypes of interest. The meta-ranker algorithm can be used for combination design with model-based approaches to reduce the need for extensive clinical studies to fulfill the fixed combination rule. It considers the direction and importance of each desired phenotype while comparing the beneficial effects of investigational product candidates. Here, we describe this algorithm for investigational product candidate ranking in multiple contexts where input data is a treatment screen for various phenotypes. We introduced phenotype influence factors to assign weights based on disease etiology. As input data, we used the results of two non-clinical screens in primary human hepatocytes where multiple endogenous metabolic modulators (EMMs), comprising several molecular families, including amino acids, bile acids, other intermediary substrates, and hormones, were tested as single agents or combinations at multiple concentrations. The first data set provides an example for using a meta-ranker for single EMMs, and the second data set offers a model for deconvolution of the effects of a combination and its subcomponents.