Recent advancements in nanoparticle detection technology present exciting new possibilities across numerous fields like biology, medicine, and materials science. A study published recently introduces "Deep Nanometry" (DNM), which combines advanced unsupervised deep learning techniques with innovative optofluidic technologies to dramatically improve the scalability and sensitivity of nanoparticle analysis.
Nanoparticle characterization is critically important, yet traditional techniques often face significant trade-offs. Sensitivity, which refers to the minimum detectable levels of target particles, often compromises throughput, which is the number of particles detectable per second. With DNM, researchers have achieved the remarkable ability to detect polystyrene beads as small as 30 nm at throughput levels exceeding 100,000 events per second. More intriguingly, this method identifies rare extracellular vesicles (EVs) directly from non-purified serum samples—the EVs are markers linked to various diseases, including cancer, and were detected at concentrations as low as 0.002% of total particles.
The innovative method uses hydrodynamic focusing to streamline the flow of nanoparticles through the detection system, allowing for unparalleled sensitivity. Its ability to classify nanoparticles based on both size and surface markers opens new doors for accurate profiling of EVs, which are nano-sized lipid cargoes secreted by living cells, serving as vehicles for intercellular communication.
Much of the breakthrough relies on the application of deep learning: DNM employs unsupervised deep learning models to reduce background noise during detection. The researchers note, “This unsupervised approach requires only an empty-water time series in addition to the particle time series to be denoised itself for training a deep learning-based model.” Therefore, DNM can be trained on relatively straightforward setups, eliminating laborious sample purification steps often necessary with prior detection techniques.
One of the side effects of using traditional multi-parametric analysis techniques is the challenge of obtaining pure samples without loss of data integrity. Deep Nanometry addresses this directly by enabling direct detection of rare diagnostic EVs from non-purified serum. The detection of EVs is pivotal, as these particles possess distinctive combinations of specific surface proteins pivotal for early cancer detection. Overall, DNM has detected CD9 and CD147 double-positive EVs present only 0.93% of patients and 0.17% of healthy controls, with significant differentiation statistically confirmed at p < 0.05.
The methodological design of DNM is straightforward but effective. It incorporates simplified optics paired with advanced computational models. Importantly, it aims to identify very weak scattering signals from nanoparticles even when significantly overlaid with background noise—something past techniques struggled to address. “The sensitive and scalable DNM directly detects rare target extracellular vesicles (EVs) in non-purified serum,” the authors state. This provides hope for not only simplifying testing procedures but for more accurate diagnostics through seamless integration with existing clinical workflows.
Looking forward, the implementation of DNM techniques has the potential to revolutionize not only cancer diagnostics but also assist with monitoring other diseases where identification at the nanoscale is required. The researchers conclude, “We foresee the unsupervised 1D denoising technique we have developed can impact a wide range of technological fields, where detecting weak 1D signals is pivotal.” With these developments, the future of high-throughput analysis appears exceedingly promising.