A novel technique leveraging data analysis, known as fingerprint mass spectrometry, offers a way to measure the mass of separate proteins using nanoscale devices.
Scientists at Caltech have created a machine learning-based approach that enables precise measurement of the mass of individual particles and molecules with intricate nanoscale devices. This innovative method broadens the range of devices that can be used for mass measurement, aiding in protein identification, and may lead to insights into the complete proteome—the total collection of proteins in any organism.
Proteins serve as the driving force behind living organisms. Insights into which proteins are produced, their locations, and quantities can reveal significant information about an organism’s health, offer clues during disease diagnosis, and suggest possible treatment strategies. However, researchers still lack comprehensive methods to characterize entire proteomes.
“We’re discussing mass spectrometry at the level of individual molecules; this allows us to observe full proteins in real-time without the need to fragment them,” explains Michael Roukes, the Frank J. Roshek Professor of Physics, Applied Physics, and Bioengineering, and one of the authors of a paper published in Nature Communications detailing this new technique. “With a high-throughput single-molecule technique, we can measure millions of proteins in a reasonable timeframe, enabling a deeper understanding of an organism’s complete proteome, including that of humans.”
Mass spectrometry is a widely used analytical technique that scientists employ for various types of molecular investigation. The process starts with a sample, which is ionized (charged by removing electrons) and propelled along a designated path. By applying a magnetic or electric field to nudge the ions and observing their deflection, researchers can determine their mass and charge. This data can then be used to decipher the chemical composition of the sample.
This technique is utilized for various applications, including forensic trace analysis, disease biomarker detection, and pesticide residue examination. However, the initial ionization process may not be suitable for all samples, especially biological ones, which can be altered during ionization.
When it comes to tiny samples, like individual proteins, complexities arise. Recent advances in nanoelectromechanical systems (NEMS) over the past two decades have allowed for a type of mass spectrometry that bypasses the need for sample ionization. This development has made real-time mass measurements of small molecules routine. Thus, scientists can now interpret chemical species in a sample without relying on educated guesses. However, this method has ruled out certain sophisticated NEMS devices from being utilized in mass spectrometry.
In traditional NEMS mass spectrometry, a silicon element resembling a minuscule beam anchored at both ends resonates when struck, similar to a guitar string, oscillating with specific modes at distinct frequencies.
When a sample is added to this beam, the frequencies of the beam’s vibrational modes are altered. “By tracking these frequency shifts, we can deduce the sample’s mass,” says John Sader, a research professor at Caltech in aerospace and applied physics and the lead author of the new study. “However, to determine this, it’s essential to know the shape of each mode, which is critical for all current measurements.”
With the latest NEMS devices, accurately determining a precise mode shape can prove challenging due to variations or imperfections at the nanoscale that can alter the mode shapes slightly. Additionally, the advanced NEMS devices, which researchers have crafted to probe the foundational physics of the quantum world, exhibit highly intricate three-dimensional modes that closely intertwine in frequency. “You can’t merely compute the mode shapes and their frequencies theoretically and expect them to remain constant during measurements,” explains Sader.
Moreover, the exact point where a sample is placed within the device affects the frequency readings of the beam. For instance, if the sample is added near one of the beam’s ends, the frequency shift will be minimal compared to placing it closer to the center, where the vibrational amplitude is usually more significant. However, with devices measuring about one micron by one micron, visualizing the precise sample placement becomes impractical.
Fingerprints Reveal Location and Determine Mass
Sader, Roukes, and their team have introduced a new method termed “fingerprint nanoelectromechanical mass spectrometry” that addresses these challenges.
By following this technique, the researchers randomly position a single particle on the NEMS device in a state of ultrahigh vacuum and ultralow temperature. They monitor real-time changes in the frequencies of several device modes corresponding to that placement. This enables them to create a high-dimensional vector that reflects these frequency shifts, with each dimension representing a mode. By repeating this process for various particles in numerous random locations, they established a library of vectors to train the machine learning algorithm.
Interestingly, each vector acts like a fingerprint, possessing a unique shape or direction that alters distinctly based on where a particle is positioned.
“When I place a particle of unknown mass anywhere on the NEMS device—regardless of its location—and measure the vibrational mode frequencies, I obtain a vector that points in a specific direction,” Sader clarifies. “By comparing it to the vectors in the database and identifying the one most parallel to it, I can determine the unknown particle’s mass. It’s merely the ratio of the magnitudes of the two vectors.”
Roukes and Sader believe this fingerprint method can be applied across all devices. In this study, the Caltech team theoretically examined phononic crystal NEMS devices created in the lab of their colleague, Stanford physicist Amir Safavi-Naeni. These innovative NEMS devices are effective at trapping vibrations, allowing them to resonate at certain frequencies for extended periods, enabling researchers to capture high-quality measurements. This fingerprint approach facilitates mass spectrometry measurements using these cutting-edge devices. For validation, the team measured the mass of individual GroEL protein particles, known for assisting in correct protein folding in cells.
Roukes points out that conventional mass spectrometry methods pose challenges for large protein complexes and membrane proteins like GroEL for several reasons. First, these techniques yield an overall mass and charge that do not uniquely identify a specific species. Given the size of these complexes, numerous candidates could fit. “You need to differentiate these somehow,” Roukes notes. “Currently, the leading method is to fragment the molecule into pieces between 3 and 20 amino acids long.” This process is followed by employing pattern recognition to trace back to the original molecule from the fragments. “However, this approach no longer retains a unique identifier of the original structure since it’s destroyed during fragmentation.”
Roukes emphasizes that the new fingerprint method “is progressing towards an alternative known as native single-molecule mass spectrometry, where large proteins and complexes can be analyzed one by one in their natural form without any fragmentation.”
