PROJECT SUMMARY AND ABSTRACT
Cancer nanomedicine is a rapidly growing area of medical research which may overcome the intrinsic limits of
convention cancer therapies for more effective and safer cancer treatment. Yet, the challenge remains in the
development of cancer nanomedicine owing to the inability for nanoparticles (NPs) to efficiently deliver
therapeutic agents into solid tumors. It has been well documented that delivering NPs into biological fluids results
in the formation of NP-protein corona which endows NPs a biological identify by unpredictably altering the
uptake, biodistribution and toxicity of NPs, thus hindering the therapeutic potential of nanomedicines. Therefore,
the objective of this proposal is to use artificial intelligence (AI) algorithms to take into account NPs-interactions
protein corona fingerprints into the prediction model to improve the delivery efficiency of NPs to tumor. Our
hypothesis is that the proposed AI-based computational model will be capable of precisely predicting the
pharmacokinetic profiles of NPs and NPs libraries with desired optimal targeting efficiency by considering the
distinct protein corona fingerprints. Our methodology proposed herein can find a general use for reducing the
number of nanomedicines that need to be tested at the early-stage preclinical trials and could identify a targeted
library of novel NPs with desired delivery efficiency to tumor. Two Specific Aims of this proposal were formulated
to test the hypothesis. Aim 1: To establish a generative adversarial network (GAN) model for the predictions of a
protein corona fingerprint corresponding to physicochemical properties of distinct NPs. Aim 2: A neural ordinary
differential equation (NODE)-driven pharmacokinetic model will be developed to predict the temporal
biodistribution and tumor delivery efficiency of the protein corona-NPs complex. This design is necessary
because the protein corona patterns, which are dependent on the physiochemical properties of NPs, can
significantly change the in vivo fate of NPs. Experimental data for the AI model training are from our recently
published Nano-Tumor Database, in which tumor delivery efficiency and time-course pharmacokinetic profiles
in different tissues were evaluated. The models will be made available as a user-friendly interface, through AI-
guided design, to assist the discovery of the novel nanomedicine libraries with desired tumor delivery efficiency.
The proposed research is significant as this study addresses possible solutions to the unmet need for understanding
the interaction between protein corona and NPs that impacts their targeting to tumor. This project has broad
impacts because, upon successful completion, the AI-based system can provide proof-of-art technology for the
design of smart nanomedicines to select an optimized nanomedicine library displaying maximum efficacy at
tumor sites. Additionally, since our computational approach can be extended to different NPs and species, this
work points towards a novel way of designing and optimizing nanomedicines for a wide variety of applications.
