Rational design of active pharmaceutical ingredient-based co-assembled nanoparticles for cancer immunotherapy.

Combination strategy is crucial for enhancing cancer therapeutic efficacy, but co-delivery of multiple active pharmaceutical ingredients (APIs) remains challenging. Although nanomedicines address spatiotemporal co-delivery of APIs, it typically requires extensive excipients and formulation screening, which hampers drug discovery efficiency. Herein, we have developed a data-driven design workflow aided by ML to accelerate the rational design of ICG-templated API co-assembled nanoparticles for immunomodulation and antitumor synergistic therapy. The machine learning models can identify the important physicochemical properties that affect co-assembly of APIs and ICG, and may provide references for the design of nanoparticles. To illustrate the practicality of this workflow, compounds with lysosomal activity and drugs with synergistic effects were respectively explored to prepare nanomedicines for immunomodulation or synergistic anti-tumor effects. In one case, the models predicted co-assembled nano-adjuvants from lysosomal toll-like receptor 7/8 agonists to enhance antitumor immunity. In another case, the models launched multi-drug co-assembled nanomedicines including AZD7762 (AZD) and camptothecin (CPT) for synergistic cancer therapy. Both of the nanomedicines showed superior immune activation and antitumor efficiency than the free APIs. This study provides a ML-aided workflow to accelerate the design of co-assembled nanomedicines and offers feasible strategies to enhance the efficacy of anti-tumor therapies.