TL;DR: MIT engineers reported an autonomous, pilot-scale platform that produces loaded lipid nanoparticles at a requested size with no human intervention.

On September 25, 2026, a team led by Peter Sagmeister, Aniket Udepurkar and Cedric Devos published results in ACS Nano describing a closed-loop system that links a size-control production method to inline dynamic light scattering for real-time feedback, according to ACS Nano. The platform adds automated design of experiments, dynamic parameter sweeps and Bayesian optimization to map process-property relationships for loaded lipid nanoparticles while reducing experimental burden, according to ACS Nano. The work was reported the same day by MIT News as a way to make RNA delivery particles faster and with tighter control of size and shape for vaccines and therapeutics.

What was published on September 25

The paper is titled Autonomous Lipid Nanoparticle Engineering, according to ACS Nano. The listed authors are Peter Sagmeister, Aniket Pradip Udepurkar, Cedric Devos, Joy I. Ren, Konstantinos Zinelis, Sofiya Chubich, Krystian Ganko, Andy Y. Liu, Dylan Nguyen, Richard D. Braatz and Allan S. Myerson, according to ACS Nano.

The publisher is the American Chemical Society, according to ACS Nano. The manuscript record lists receipt on August 12, 2026, revision receipt on September 15, 2026, acceptance on September 16, 2026 and online publication on September 25, 2026, according to ACS Nano.

Three contributors shared equal credit. P.S., A.U. and C.D. contributed equally to the work, according to ACS Nano. The senior corresponding author is Allan S. Myerson, with contact listed as [email protected], according to ACS Nano.

The funding record names the National Cancer Institute award P30-CA14051 and the Food and Drug Administration award 75F40122C00200, according to ACS Nano. A competing-interest statement says authors P.S., A.U., C.D., R.D.B. and A.S.M. are listed as inventors on a provisional patent application filed by MIT that covers the size- and morphology-control method and the autonomous platform for automated process development described in the manuscript, according to ACS Nano.

MIT News identified the three lead authors as MIT postdocs Aniket Udepurkar, Peter Sagmeister and Cedric Devos, with Allan Myerson as senior author, according to MIT News. The MIT News account was written by Anne Trafton and dated September 25, 2026, according to MIT News.

Why lipid nanoparticles matter for nucleic acid medicines

Lipid nanoparticles are the leading vehicles for encapsulating and delivering nucleic acid therapeutics, according to ACS Nano. RNA vaccines and other nucleic acid therapeutics are typically packaged within fatty molecules known as lipid nanoparticles, according to MIT News.

Vaccines based on mRNA work by delivering instructions to cells to produce a harmless version of a viral protein, which prompts the immune system to generate a response, according to MIT News. If mRNA is injected on its own, it will be quickly broken down in the body, according to MIT News.

Delivery packaging addresses that limitation. “These are really a revolutionary type of therapeutics, but they need some kind of delivery vehicle to bring them to the right cells in the body,” says Cedric Devos, an MIT postdoc and one of the lead authors, according to MIT News.

For the mRNA Covid-19 vaccines and other mRNA therapeutics, scientists have used lipid nanoparticles for that packaging, according to MIT News. LNPs usually consist of four components — an ionizable lipid, a phospholipid, cholesterol and a lipid attached to a molecule of polyethylene glycol, which helps to stabilize the LNP, according to MIT News.

The new platform is framed as process development rather than a new therapeutic itself. The authors state that LNP process development remains labor-intensive and empirical, constrained by coupled quality attributes and limited mechanistic insight, according to ACS Nano. Current methods for designing new lipid nanoparticles often require time-consuming trial-and-error experiments, with limited investigation of desired particle size and shape, according to MIT News.

The core result: target size on demand

The central claim is direct. The team presents an autonomous, pilot-scale platform for accelerating LNP process development by identifying critical process parameters that produce LNPs with target size attributes, according to ACS Nano.

In practical terms, the researchers can specify a particle size and the system will generate that size, according to MIT News. The system also measures the resulting particles to make sure they are the right size and, if not, adjusts the delay time and other factors to steer them to the right size, according to MIT News.

That closed loop is enabled by inline measurement. The platform combines a size-control production method with inline dynamic light scattering for real-time feedback, enabling closed-loop experimentation and accelerated optimization, according to ACS Nano. The researchers incorporated a commercially available dynamic light scattering device that can measure the sizes of the particles as they are formed, according to MIT News.

The authors report that the platform rapidly maps complex process-property relationships for loaded LNPs while minimizing experimental burden, according to ACS Nano. They state that the resulting data-rich outputs were used to develop a predictive model that quantitatively links process parameters to particle quality attributes, according to ACS Nano.

Operable in fully autonomous mode, the platform provides a scalable and flexible framework for rational LNP manufacturing and accelerates the broader development of nucleic acid therapeutics, according to ACS Nano. The new process, which can be run automatically without any human intervention, could greatly speed up the development of new RNA and DNA therapeutics, the researchers say, according to MIT News.

How conventional mixing limited control

Standard LNP formation mixes two fluid streams at high speed, according to MIT News. One stream consists of lipid molecules suspended in ethanol, and the other contains mRNA dissolved in an acidic buffer, according to MIT News.

Those solutions are not mixed at equal flow rates in the conventional approach. Instead, there is about three times more of the mRNA solution than the lipid solution, according to MIT News. This unequal ratio helps promote the formation of lipid nanoparticles containing mRNA, but it does not offer precise control over the size or shapes of the particles, according to MIT News.

That limitation motivates the study. “The size and shape of LNPs could not be reliably controlled by any previous production method,” says Allan Myerson, a professor of the practice in MIT’s Department of Chemical Engineering and the senior author, according to MIT News.

Tuning size and shape is presented as biologically relevant. Tuning the size and shape of LNPs can open the possibility of targeting different organs and tissues, according to MIT News. The assessment of candidates remains downstream. “This method can help you determine what are the parameters that will generate specific size and shape attributes, before you take those particles and see which one will perform best,” says Cedric Devos, an MIT postdoc and one of the lead authors, according to MIT News.

The two-step method behind autonomy

The autonomous system builds on a two-step mixing strategy the MIT team described in ACS Nano in the prior year, according to MIT News. In that earlier work, the team showed that breaking the mixing process into two steps enabled much better control of particle size, according to MIT News.

In the first step, mRNA and lipids are mixed together at equal flow rates, according to MIT News. Then, after a short delay, more of the buffer is added, which halts the growth of the particles, according to MIT News. Longer delays produce larger particles, according to MIT News.

Residence time is the control lever. Residence time is the time it takes between the first mixer and the second mixer, according to MIT News. Keeping that residence time long means particles grow more, while keeping it short keeps them small, according to MIT News. The MIT account describes this as a lever over lipid nanoparticle manufacturing that was not available before, according to MIT News.

The same two-step approach also addressed shape. By changing the concentration of the buffer added in the second step, the researchers were able to transform the particles from spheres to elongated particles that resemble avocados, according to MIT News. Both interventions — changing the delay residence time and changing the buffer composition — allow control of particle size and shape without otherwise altering the composition of the LNPs, according to MIT News.

The new ACS Nano paper automated that two-step mixing process, according to MIT News. One author summarized the sequence as a methods advance followed by automation. “The first paper really unlocked the new methodology to make lipid nanoparticles, to truly engineer them by size and shape,” Sagmeister says, according to MIT News.

What “autonomous” means in this system

Autonomy here refers to experiment design, execution, measurement and adjustment in a loop. The platform has built-in automated design of experiments, dynamic parameter sweeps and Bayesian optimization, and the platform enables rapid, data-rich exploration of complex design spaces, according to ACS Nano.

Automated design of experiments organizes test conditions without manual selection of each run. Dynamic parameter sweeps vary process inputs over time to collect dense response data. Bayesian optimization uses prior results to choose subsequent conditions toward a target size attribute. Those three capabilities are listed together as built-in functions of the platform, according to ACS Nano.

Inline dynamic light scattering supplies the feedback signal. Dynamic light scattering measures particle size as particles form, allowing the closed loop to compare the observed size with the requested size, according to MIT News and ACS Nano. When the measurement differs from the target, the system adjusts delay time and other factors to move toward the requested size, according to MIT News.

Shape remains partly outside the loop. The system can also produce particles of different shapes, but measuring those shapes has to be done outside of the automated system, according to MIT News. That distinction matters because size is directly steered by inline sensing, while shape assessment requires separate characterization.

The researchers used the data gathered from the automated experiments to train a machine-learning model that can predict the combination of factors that will generate a particular size or shape, according to MIT News. In the paper’s terms, the data-rich outputs were used to develop a predictive model that quantitatively links process parameters to particle quality attributes, according to ACS Nano. The approach allowed investigation of how changing inputs affects size and shape on a much faster timescale than currently possible, according to MIT News.

Who did the work and how it was built

The author list spans senior faculty, postdoctoral researchers and undergraduates, according to ACS Nano and MIT News. Richard D. Braatz and Allan S. Myerson are the senior co-authors listed on the ACS Nano record, according to ACS Nano. Allan Myerson is identified as a professor of the practice in MIT’s Department of Chemical Engineering and senior author of the study, according to MIT News.

Undergraduate contributors played a software role. Through MIT’s Undergraduate Research Opportunities Program, applied mathematics and computer science undergraduates Joy Ren, Sofiya Chubich and Dylan Nguyen gained hands-on experience, learning how advanced software engineering can be integrated with chemical engineering to develop an automated platform for LNP process development, according to MIT News. Those three students are also named authors on the ACS Nano paper, according to ACS Nano.

Additional co-authors named on the paper include Konstantinos Zinelis, Krystian Ganko and Andy Y. Liu, according to ACS Nano. The supporting information is described as containing hardware and control-software detail, complete experimental protocols and extended information for automated design of experiments, time-dynamic process sweeps, Bayesian optimization and machine-learning model development, with additional figures, tables and diagnostics, according to ACS Nano.

What this changes for RNA development work

The immediate use is development speed and precision, not a therapeutic result. The platform is intended to identify critical process parameters that produce target size attributes, according to ACS Nano. Developers can therefore screen size and shape variants more systematically before testing biological performance, according to MIT News.

The distinction between making particles and selecting winners is explicit in the researchers’ description. The method helps determine the parameters that generate specific size and shape attributes before those particles are evaluated for which one performs best, according to MIT News. That framing keeps manufacturing optimization separate from later functional testing.

Scale and transferability are also part of the claim. The platform is described as pilot-scale, operable in fully autonomous mode and flexible for rational LNP manufacturing, according to ACS Nano. The authors state that it provides a scalable framework and accelerates broader development of nucleic acid therapeutics, according to ACS Nano.

For U.S. science and technology readers, the funding and intellectual-property record adds context. Support included a National Cancer Institute award and a Food and Drug Administration contract, according to ACS Nano. MIT filed a provisional patent application covering the size- and morphology-control method and the autonomous platform, with five authors listed as inventors, according to ACS Nano.

Sources referenced