MIT team uses machine learning to find nanoparticles with high drug-loading capacities

Massachusetts Institute of Technology (MIT) researchers have used machine learning to identify pairs of small-molecule drugs and inactive ingredients that will self-assemble into nanoparticles with high therapeutic payloads. 

Nanoparticle formulations can overcome the pharmacokinetic limitations of small-molecule drugs. However, the process of developing nanoparticles varies a lot from formulation to formulation. Some combinations of drug and excipient readily self-assemble into nanoparticles that carry large payloads of the therapeutic. Other combinations are hard to make and carry very little of the drug.

Efforts to develop nanoparticle formulations have been held back by the inability to predict which mix of ingredients will have high loading capacities. To address that limitation, MIT researchers used machine learning and high-throughput experimentation to identify effective nanoformulations.

The project, details of which were published in Nature Nanotechnology, evaluated 788 therapeutic small molecules and 2,686 approved excipients. Those building blocks resulted in 2.1 million pairings. From those pairings, the researchers identified 100 self-assembling drug nanoparticles and picked two formulations for further characterization.  

One of the formulations consisted of sorafenib, the active ingredient in Nexavar, and the licorice root extract glycyrrhizin. The other formulation paired the antifungal medicine terbinafine to taurocholic acid. In in vitro and in vivo tests, the sorafenib-glycyrrhizin formulation bettered the unpaired active ingredient, suggesting it could improve on the standard of care in hepatocellular carcinoma.

The researchers see another application for the technology. As well as optimizing formulations for all patients, the platform could propose personalized sets of excipients for people who are allergic to certain ingredients.   

“We have an opportunity to think about matching the delivery system to the patient. We can account for things like drug absorption, genetics, even allergies to reduce side effects upon delivery,” Daniel Reker, Ph.D., one of the authors of the paper, told MIT News.