Supervised learning predicts the in vivo fate of engineered nanomaterials

ABSTRACT: Despite having exquisite control over nanoparticle design, controlling nanoparticle fate in vivo remains a major barrier for clinical translation. This is because we do not understand how nanoparticles interact with the surrounding environment in vivo and how this lack of control contributes towards organ accumulation. The suggested link between nanoparticle interactions and organ accumulation are the proteins that adsorb to the nanoparticle surface following administration. How this network of proteins changes during nanoparticle transport, and its influence over the fate of where nanoparticles distribute inside of the body is fundamentally not understood. Here we developed a simple workflow to show that the evolution of proteins on the surface of nanoparticles predicts the biological fate of nanoparticles in vivo. This workflow involves extracting nanoparticles at multiple time points from circulation, isolating the proteins off the surface, and training a neural network to predict nanoparticle biological fate using the proteins as inputs and clearance and organ accumulation as outputs. In a double-blind study, we validated the model by predicting nanoparticle clearance and spleen and liver accumulation with 76-97% accuracy. This work demonstrates that a link between surface adsorbed proteins and the biological fate of nanomaterials exists, and that it can be predicted using the workflow we designed. As we acquire more training data, the strength of these relationships will become more powerful. With more training data we will use more sophisticated neural networks to identify proteins and pathways to target, or create more effective nanomaterial designs to improve clinical translation.

INSTRUMENT(S): LTQ Orbitrap Elite

ORGANISM(S): Mus Musculus (mouse)

TISSUE(S): Blood Serum

SUBMITTER: Jonathan Krieger  

LAB HEAD: Warren Chan

PROVIDER: PXD011354 | Pride | 2024-09-14

REPOSITORIES: Pride