PredictCan Biotechnologies has published a new study in Frontiers in Toxicology, in collaboration with Centre Antoine Lacassagne in Nice, demonstrating the potential of its individual-centric liver platform to address one of the major challenges in drug development: Drug-Induced Liver Injury (DILI).
DILI remains an important cause of clinical development failure, regulatory restrictions, and post-marketing drug withdrawal. One of the limitations of conventional preclinical approaches is their ability to reproduce not only human liver toxicity, but also the interindividual variability that determines why the same drug may be well tolerated by most individuals while causing severe liver injury in others.
PredictCan’s approach addresses this challenge by combining human serum-derived educated liver spheroids (Liver hSES™) with AI-driven analysis to evaluate drug response across individualized human biological backgrounds.
Capturing DILI Beyond Acute Toxicity
An important finding of the study was the ability of the platform to distinguish between acute and chronic toxicity profiles.
With ximelagatran, no toxicity was detected under acute exposure conditions, whereas prolonged exposure revealed a clear DILI response. This pattern reflects the clinical profile of the compound, for which hepatotoxicity emerges following extended treatment rather than short-term exposure.
Beyond detecting toxicity, mechanistic investigations identified immune-mediated pathways associated with the observed response, illustrating the ability of the platform to move from toxicity detection toward understanding the biological processes involved.
From Population Risk to Individual Susceptibility
The study also investigated whether PredictCan’s approach could capture toxicity at the individual patient level.
Using serum from three patients with ER+/HER2− breast cancer treated with ribociclib, individualized liver models were generated and evaluated in vitro. The platform predicted grade 3 DILI in the patient who subsequently experienced liver toxicity, while no toxicity was observed in models generated from the two patients without clinical DILI.
The concordance between the individualized in vitro responses and the corresponding clinical outcomes illustrates the potential of PredictCan’s technology to identify individual susceptibility to drug-induced liver injury, a dimension that population-average preclinical models are not designed to capture.
Combining Human Biological Diversity with AI-Driven Risk Assessment
The study further demonstrates the value of combining individualized biological models with computational analysis.
PredictCan’s proprietary AI-driven approach integrates toxicity information across a cohort to characterize compounds according to both DILI severity and incidence. This enables drug candidates to be evaluated not only according to whether liver toxicity occurs, but also according to how severe the response is and how frequently susceptible profiles emerge within the tested population.
Together, these capabilities provide a framework for moving from conventional toxicity assessment toward a more population-aware and individual-centric evaluation of drug safety.
Supporting More Informed Drug Development Decisions
PredictCan’s technology is designed to complement existing preclinical and regulatory approaches rather than replace established gold standards.
By providing information on chronic toxicity, interindividual variability, susceptible profiles, and potential mechanisms earlier in development, the platform can contribute additional evidence for weight-of-evidence assessments, compound prioritization, mechanistic investigation, and decisions on subsequent testing.
The study therefore represents an important scientific validation of PredictCan’s vision: bringing human biological diversity into preclinical drug safety assessment to better understand not only whether a compound may cause toxicity, but also who may be at risk, under which exposure conditions, and through which biological mechanisms.
