Spatially distributed complex organic matter detected in an ancient river valley in Jezero crater, Mars - Science | AAAS

We’re sharing this article early to provide faster access to peer-reviewed, accepted research. It is citable and carries a permanent DOI. This version is subject to further edits and will be replaced automatically by the final Version of Record. All legal disclaimers apply. AbstractMany biologically important molecules are structural isomers that share the same molecular mass yet differ in shape, bonding and function. This creates an ambiguity for mass spectrometry imaging (MSI), where chemically distinct species can be indistinguishable on the mass-to-charge (m/z) scale and therefore cannot be reliably mapped in heterogeneous materials. Here we demonstrate that Orbitrap secondary ion mass spectrometry (OrbiSIMS) can provide an orthogonal contrast that enables both identification and spatial separation of isomeric molecules directly on surfaces. By exploiting structure-dependent secondary-ion dissociation and gas-phase collisional effects during ion transmission, we discriminate isomeric trisaccharides raffinose, maltotriose and melezitose and spatially resolve raffinose and maltotriose in mixtures.

image: Overview of memory-efficient training techniques for transformers The integration of deep learning—particularly Transformer-based models—has profoundly reshaped scientific research paradigms. In domains such as drug discovery, weather forecasting, and protein structure prediction, these models enable the extraction of complex patterns from large-scale data. However, as model resolution and parameter counts continue to grow, GPU memory consumption increases rapidly, creating a “memory wall” that confines large-scale training to costly, high-end hardware and limits broader adoption in scientific research.

image: Overview of memory-efficient training techniques for transformers The integration of deep learning—particularly Transformer-based models—has profoundly reshaped scientific research paradigms. In domains such as drug discovery, weather forecasting, and protein structure prediction, these models enable the extraction of complex patterns from large-scale data. However, as model resolution and parameter counts continue to grow, GPU memory consumption increases rapidly, creating a “memory wall” that confines large-scale training to costly, high-end hardware and limits broader adoption in scientific research.