Etiology-independent fibrosis severity scoring by quantitative digital pathology image analysis

compounds.We have previously reported the performance of novel Digital Pathology Quantitative AI to generate automatic, continuous and direct fibrosis endpoints to quantify fibrosis severity and compound treatment response.Here, we expand this validation effort to increasing concentartions of Selonsertib, Firsocostat and resmetiron (MGL-3196) and their combinations.Method: Human in vitro 3D InSight TM liver microtissues containing primary hepatocytes, Kupffer cells, endothelial cells and hepatic stellate cells in 96-well plates were used to model NASH progression using a defined cocktail of free fatty acids, LPS and high levels of sugars.For compound efficacy testing, the 3D NASH tissues were simultaneously treated with selonsertib (2 mM and 10 mM), firsocostat (0.5 mM and 10 mM), MGL-3196 (0.005 mM and 0.05 mM) andcombination of selonsertib (10 mM) with firsocostat (0.5 mM) for a total of 9 groups (n = 18 to 21 in each group).Liver microtissues slices were stained with Sirus Red and digitaly imaged at 40X.FibroNest TM , a cloud-based quantitative image analysis and quantitative AI platform, was used to quantify the fibrosis phenotype along 32 traits for collagen content, morphometry, and architecture.Each trait is described by up to seven quantitative fibrosis traits (qFTs, 315 in total).Principal qFTs are automatically detected and combined into a normalized Phenotypic Composite Fibrosis Score (Ph-FCS).Each qFT is described individulally for relative severity (green to red) in Phenotypic Heatcharts.Results: The Ph-FCS offers a significant detection threshold and dynamic range to evaluate the antifibrotic response the seven treatment arms (box plot chart and p-value table below).Firsocostat (10 mM) and MGL-3196 (0.05 mM) antifibrotic effects are significant and comparable.The combination of selonsertib (10 mM) with the low dose of firsocostat (0.5 mM) does not demonstrate any synergestic effect.The dose response effects are poorly detected except for the MGL-3196 arms ( p = 0.075) which demonstrate that the result is driven by the compounds, not the Ph-CFS score and method.Each principal qFT is described individually for relative severity (green to red) in phenotypic heatcharts which can be used to quantified differences in the fibrosis phenotype in each group, and quantify specific effects of each drug (and dose) on the collagen distribution, collagen fibers morphometry and fibrosis architecture. Conclusion:The combination of FibroNest TM imaging analysis for automated quantification of histological fibrosis severity phenotype within vitro 3D InSight TM human NASHmodel provide powerful platform for anti-fibrotic drug-candidates response evaluation.