Differentia Bio's AI models matched later clinical outcomes
Differentia Bio says its disease-modeling platform identified therapeutic mechanisms between 2023 and 2024 that were later supported by independent clinical results. One approach was backed by an FDA-approved treatment, while the other received support from early clinical data.
- Differentia identified B-cell and IL-4/Th2 therapeutic approaches in IgG4-RD between 2023 and 2024.
- The FDA approved inebilizumab in April 2025 as the first treatment for IgG4-RD.
- The pivotal Phase 3 trial reduced the risk of flares by 87 percent compared with placebo.
- An early study of dupilumab found clinical improvement in four out of five patients.
- The DHARMA platform builds digital disease models from patient data and biological knowledge.
Two predicted therapeutic approaches received clinical support
San Francisco-based Differentia Bio said on November 3, 2026, that its AI-based model for IgG4-related disease, or IgG4-RD, identified B-cell-targeting interventions and the IL-4/Th2 pathway as promising therapeutic strategies. The company presented its research at international conferences from 2023 onward, including ACoP, PAGE and NBC events, before publicly outlining its platform for autoimmune diseases in August 2024.
Subsequent developments supported both approaches. In April 2025, the FDA approved inebilizumab, a CD19-targeting therapy, as the first treatment for IgG4-RD. In its pivotal Phase 3 trial, the treatment reduced the risk of disease flares by 87 percent compared with placebo.
An early independent study reported findings on dupilumab in January 2026. The treatment inhibits IL-4Rα signaling. Four out of five patients showed clinical improvement and reduced IgG4 levels, while levels normalized in three patients.
The model narrowed its focus to biological mechanisms
Differentia said its approach differs from retrospective AI assessments that use outcomes already known. The company documented its therapeutic hypotheses before the subsequent clinical developments. At the same time, it stressed that broader, prospective validation is still needed. The current results show that mechanism-based AI predictions can be compared with human clinical evidence generated later.
According to the company, the platform begins working before biological data are collected. It combines artificial intelligence, biological knowledge graphs and systems biology to determine which biological questions, pathways and multi-omic measurements are needed to understand a disease.
The IgG4-RD model integrated single-cell RNA sequencing, flow cytometry and multiplex cytokine profiling from the same patients. Using tens of thousands of cells and millions of molecular measurements, the model filtered out the elements relevant to treatment decisions from more than 30 biological components and over 100 mechanistic interactions.
The B-cell and IL-4/Th2 pathways proved promising, while mechanisms studied in some unsuccessful clinical programs, including SLAMF7, did not emerge as significant disease drivers and were therefore not included in the original model.
The DHARMA platform builds virtual disease environments
Differentia's DHARMA™ AI Disease Digital Twin Engine platform combines generative AI, biological knowledge graphs, public scientific and clinical evidence, and detailed patient data from academic and clinical collaborations. The company says this enables the creation of digital twins operating across multiple scales that model both biological interactions and differences between patients.
The models are calibrated using pharmaceutical clients' existing preclinical and clinical data, after which they generate predictions. When new, client-specific clinical data become available, the models can be recalibrated, and updated predictions can support further development decisions.
The company also develops bespoke disease environments where research and development teams can run virtual experiments, compare therapeutic strategies and explore potential paths for clinical development. The solution is intended for oncology, autoimmune and metabolic diseases, as well as different treatment modalities.
The aim is to improve drug development decisions
According to Differentia Bio, this type of modeling can help developers evaluate disease mechanisms, clinical trial strategies, patient populations, dosing approaches, indications and treatment combinations before clinical trials begin. The approach aligns with the regulatory expansion of model-informed drug development, which the announcement said was linked in 2026 to the finalization of ICH M15 and its introduction in several major regulatory markets.
The company has established business relationships with several multinational pharmaceutical companies ranked among the world's ten largest drugmakers, as well as biotechnology developers, clinical research organizations and investment institutions. Differentia aims to make predictive modeling an ongoing capability in pharmaceutical research and development. Based on the results presented, AI's role could extend beyond molecule design to predicting how treatments are likely to perform in human disease.
