Protein Motion at Scale
Multiplexed HDX-MS shows that protein domains with similar structures can occupy markedly different energetic and conformational landscapes.
A large-scale study of protein dynamics has shown that protein domains with nearly identical three-dimensional structures can behave very differently over time, with entire structural regions varying in stability and unfolding behavior.
Called multiplexed hydrogen–deuterium exchange mass spectrometry (mHDX-MS), the approach tracks how quickly backbone hydrogens exchange with deuterium in solution. Regions exposed through structural fluctuations exchange more readily than protected regions, allowing the researchers to infer local stability and the energetic barriers separating different conformations.
Rather than analyzing proteins individually, the team measured thousands of sequences in parallel under identical conditions. The resulting dataset covered 5,778 protein domains between 28 and 64 amino acids long across ten structural families, providing a large-scale comparison of how amino-acid sequence influences conformational behavior.
“We developed a multiplexed experimental strategy that can map the energy landscape [the different energy states and conformations] of hundreds of these molecules in a single experiment,” said first author Allan Ferrari in a recent press release.
The measurements showed that similar static structures can conceal markedly different dynamics. In some proteins, fluctuations involved coordinated opening of entire structural elements, including alpha-helices and beta-sheets, rather than isolated local changes. Sequence variation could also alter these energetic landscapes without substantially changing the protein’s dominant three-dimensional structure.
“We’re observing a fundamental aspect of structural biology that was previously difficult to access: the various conformations that proteins can adopt beyond their most common structure,” Ferrari said.
The dataset also supported protein engineering. Using the measured dynamics, the researchers selected two mutations predicted to stabilize a naturally flexible region of one protein, with subsequent experiments confirming the expected effect.
The authors are now expanding the dataset across specific protein families and investigating how altered dynamics contribute to disease mechanisms
What Makes Dopamine Vesicles Different
Differences in trafficking and recycling proteins provide a molecular basis for the distinct behavior of dopamine and glutamate vesicles.
Synaptic vesicles that carry dopamine have a distinct protein composition from those used for glutamate, providing a molecular explanation for differences in neurotransmitter release, according to a recent study.
The researchers isolated vesicles carrying either vesicular monoamine transporter 2 (VMAT2), which packages dopamine and other monoamines, or vesicular glutamate transporter 2 (VGLUT2). Quantitative mass spectrometry then compared their protein composition, revealing differences in both abundance and isoform usage across several families involved in vesicle trafficking, recycling, and release.
“Our research sheds light on the different ways brain cells release chemical messengers, which the cells use to communicate with each other,” said senior author Katlin Silm in a Cedars-Sinai press release. “We identified key differences between brain cells that release dopamine and those that release other brain chemicals.”
The proteomic differences were subsequently validated in primary neurons and mouse brain tissue. One of the clearest differences involved SCAMP5, a membrane-trafficking protein whose abundance differed between the two vesicle populations. Removing SCAMP5 selectively impaired recycling of VGLUT2-containing vesicles while leaving VMAT2-targeted vesicles largely unaffected, linking compositional differences to distinct vesicle behavior rather than simply cataloguing different proteins.
The findings build on earlier evidence that dopamine- and glutamate-containing vesicles differ in recycling kinetics, biogenesis, and their response to firing frequency. By resolving the proteins carried by each population, the study provides a molecular basis for those functional differences.
The authors suggest that defining how this specialized vesicle machinery supports dopamine signaling could help explain the long-term vulnerability of dopamine-producing neurons. “The differences we identified help explain the unique properties of dopamine release and lay the groundwork to explore how this affects the long-term stability of dopamine-producing brain cells,” said Silm.
Adaptive Control for ToF-SIMS
PACE-SIMS adapts an eight-hour ToF-SIMS experiment as data arrive while keeping higher-level scientific and safety decisions under researcher supervision.
An AI-controlled workflow has completed an eight-hour time-of-flight secondary ion mass spectrometry (ToF-SIMS) experiment with checkpoint-based quality control, adapting measurements as data arrived while keeping higher-level scientific and safety decisions under researcher supervision.
Called PACE-SIMS (Pause–Assess–Correct–Execute), the system turns scientific questions and quality criteria specified in natural language into an experimental plan. After human approval, an AI agent operates an unmodified commercial ToF-SIMS instrument, evaluating each measurement before deciding whether to proceed, adjust parameters, repeat at a spare location, or return control to the researcher.
The workflow was tested in a blinded study of tungsten oxide films containing buried oxygen-18 tracer layers. Sample composition, mounting order, and independent crater-depth measurements were withheld while the agent completed 35 positive- and negative-ion measurements over 8.1 hours. Because SIMS progressively sputters away the analyzed material, failed measurements cannot simply be repeated at the same location.
During the run, the agent made three unscripted corrections that a fixed queue would have missed. It correctly recovered the films’ oxygen-18 enrichment ranking, while sputter rates inferred from the SIMS profiles agreed with independent atomic force microscopy measurements to within 1.05 percent.
The workflow also identified a 5.3 percent systematic difference in isotope measurements between ion polarities, derived a calibration for tungsten oxide composition, and inferred how the oxygen-18 tracer had been delivered during film deposition. Those conclusions were subsequently checked against information withheld during the blinded analysis.
The work is currently available as a preprint. The authors argue that checkpoint-based autonomy may be particularly useful for destructive analytical techniques, where delayed detection of a failed or uninformative measurement can mean both lost instrument time and irretrievably consumed sample.
MALDI Without the Site Effect
DALMA separates biological signal from site-specific variation so MALDI-TOF models can generalize to unseen clinical centers.
Machine-learning models built from MALDI-TOF spectra often lose accuracy when moved between hospitals, because technical differences between sites can become entangled with the biological signal used for identification. A new framework aims to separate those two sources of variation so that models can generalize without site-specific retraining.
Domain alignment for MALDI-TOF MS, or DALMA, uses a shared encoder to capture features that remain informative across acquisition centers, while domain-specific decoders absorb laboratory-dependent variation during training. Species labels help organize the latent representation around biological identity. Once trained, the center-specific components are removed, leaving a single encoder that can process spectra from previously unseen sites.
The researchers benchmarked the approach across seven MALDI-TOF datasets from three countries, spanning six bacterial groups and differences in instrumentation and preprocessing. After representation learning, spectra became more comparable across centers while retaining the biological separation needed for microbial identification.
The key test was whether biological discrimination survived that alignment. At two clinical centers excluded from training, DALMA maintained balanced microbial-identification accuracy above 0.90 and generally outperformed conventional representation-learning, domain-adaptation, and pretrained MALDI models.
The same latent representation also supported antimicrobial-resistance prediction in Klebsiella pneumoniae. Resistance labels were incorporated during training, allowing the model to preserve features relevant to susceptibility alongside species-level information.
DALMA also provided a way to recognize spectra that fell outside the range represented during training. Excluding those atypical cases improved performance at both unseen centers, with balanced accuracy at one site rising from 0.949 to 0.997 while retaining 87.7 percent of samples.
As a preprint, the study remains to be peer reviewed, but the authors argue that separating biological signal from site-specific variation and flagging unfamiliar spectra could make cross-center MALDI-TOF models more dependable.
(Mass) Spectacular and Strange
How to Train Your Dragon (Fruit)
Dragon fruit does not earn its name quietly. A recent multi-omics study suggests its vivid peel colors emerge from several pigment pathways changing hands as the fruit ripens.
Researchers at Hainan University examined four pitaya cultivars with contrasting peel and pulp colors, collecting peel tissue at the green-fruit, color-transition, and mature stages. The team combined untargeted LC-QTOF metabolomics with Illumina RNA sequencing, pathway analysis, and qRT-PCR validation to connect pigment metabolites with gene activity.
Across 36 samples, metabolomics detected 2,682 peaks and annotated 2,670 metabolites, while transcriptome sequencing identified 5,195 previously unrecognized genes. Red-peel cultivars accumulated higher levels of betalains, particularly the betacyanins betanidin, gomphrenin-I, and lampranthin II, alongside increased expression of key betalain-biosynthesis genes including CYP76AD1, DODA, and glycosyltransferases.
In ‘Yanwoguo’, the yellow peel looked less like a shortage of red-pigment ingredients than a failure to convert them. Tyrosine, the betalain precursor, was abundant, but betalain-related genes remained weakly expressed. Instead, the cultivar accumulated more carotenoid and flavonoid metabolites, including 25 flavonoids found at especially high levels.
As the fruit matured, chlorophyll-biosynthesis genes generally declined, while degradation-related genes increased, helping remove the green background and reveal the pigments beneath. The findings recast pitaya peel color as a coordinated handover between pigment pathways rather than a simple red-or-yellow switch. They also identify candidate genes that could help breeders develop more visually distinctive cultivars.
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
