Induced pluripotent stem cells (iPSCs) provide an inexhaustible source of cells for modeling disease and testing drugs. Here we develop a bioinformatic approach to detect differences between the genomic programs of iPSCs derived from diseased versus normal human cohorts as they emerge during in vitro directed differentiation. Using iPSCs generated from a cohort carrying mutations (PiZZ) in the gene responsible for alpha-1 antitrypsin (AAT) deficiency, we find that the global transcriptomes of PiZZ iPSCs diverge from normal controls upon differentiation to hepatic cells. Expression of 135 genes distinguishes PiZZ iPSC-hepatic cells, providing potential clues to liver disease pathogenesis. The disease-specific cells display intracellular accumulation of mutant AAT protein, resulting in increased autophagic flux. Furthermore, we detect beneficial responses to the drug carbamazepine, which further augments autophagic flux, but adverse responses to known hepatotoxic drugs. Our findings support the utility of iPSCs as tools for drug development or prediction of toxicity.