Machine Learning-Assisted Navigation of Iron Oxide Nanoparticle Synthesis: A Meta Analysis

Abstract

The past two decades have seen a significant increase in research on the use of iron oxide nanoparticles (IONPs) for a wide range of biomedical applications. IONPs are safe, biocompatible and have increased surface areas that enhance their magnetic properties. The relationship between their size and response to the applied magnetic field underpins the importance of optimizing synthesis conditions to achieve the desired biomedical performance. Unfortunately, aggregation and difficulties in controlling their size distribution hamper the development of IONPs-containing diagnostics and therapeutics. Clearly, a better understanding of the extrinsic parameters affecting the size and magnetic properties of IONPs is needed. To address this paucity of information, I compiled a large dataset from the literature, and used machine learning to explore the relative contributions of synthesis conditions to the magnetic properties of IONPs. I determined the contribution of each experimental parameter to magnetic properties using two machine learning algorithms, regression trees and an artificial neural network. I demonstrate that computer-assisted approaches hold considerable promise for finding bespoke synthesis conditions to generate materials appropriate for specific biomedical applications.