Integration of machine learning-based histopathology and hepatic transcriptomic data identifies genes associated with portal inflammation and ductular proliferation as predictors of disease progression in advanced fibrosis due to nash