Large-scale foundation model on single-cell transcriptomics.

Large pretrained models have become foundation models leading to breakthroughs in natural language processing and related fields. Developing foundation models for deciphering the 'languages' of cells and facilitating biomedical research is promising yet challenging. Here we developed a large pretrained model scFoundation, also named 'xTrimoscFoundationα', with 100 million parameters covering about 20,000 genes, pretrained on over 50 million human single-cell transcriptomic profiles. scFoundation is a large-scale model in terms of the size of trainable parameters, dimensionality of genes and volume of training data. Its asymmetric transformer-like architecture and pretraining task design empower effectively capturing complex context relations among genes in a variety of cell types and states. Experiments showed its merit as a foundation model that achieved state-of-the-art performances in a diverse array of single-cell analysis tasks such as gene expression enhancement, tissue drug response prediction, single-cell drug response classification, single-cell perturbation prediction, cell type annotation and gene module inference.

Iron acquisition and mineral transformation by cyanobacteria living in extreme environments.

Iron is an essential micronutrient for most living organisms, including cyanobacteria. These microorganisms have been found in Earth's driest polar and non-polar deserts, including the Atacama Desert, Chile. Iron-containing minerals were identified in colonized rock substrates from the Atacama Desert, however, the interactions between microorganisms and iron minerals remain unclear. In the current study, we determined that colonized gypsum rocks collected from the Atacama Desert contained both magnetite and hematite phases. A cyanobacteria isolate was cultured on substrates consisting of gypsum with embedded magnetite nanoparticles. Transmission electron microscopy imaging revealed a significant reduction in the size of magnetite nanoparticles due to their dissolution, which occurred around the microbial biofilms. Concurrently, hematite was detected, likely from the oxidation of the magnetite nanoparticles. Higher cell counts and production of siderophores were observed in cultures with magnetite nanoparticles suggesting that cyanobacteria were actively acquiring iron from the magnetite nanoparticles. Magnetite dissolution and iron acquisition by the cyanobacteria was further confirmed using large bulk magnetite crystals, uncovering a survival strategy of cyanobacteria in these extreme environments.

Adsorption of Biomineralization Protein Mms6 on Magnetite (Fe3O4) Nanoparticles.

Biomineralization is an elaborate process that controls the deposition of inorganic materials in living organisms with the aid of associated proteins. Magnetotactic bacteria mineralize magnetite (Fe3O4) nanoparticles with finely tuned morphologies in their cells. Mms6, a magnetosome membrane specific (Mms) protein isolated from the surfaces of bacterial magnetite nanoparticles, plays an important role in regulating the magnetite crystal morphology. Although the binding ability of Mms6 to magnetite nanoparticles has been speculated, the interactions between Mms6 and magnetite crystals have not been elucidated thus far. Here, we show a direct adsorption ability of Mms6 on magnetite nanoparticles in vitro. An adsorption isotherm indicates that Mms6 has a high adsorption affinity (Kd = 9.52 µM) to magnetite nanoparticles. In addition, Mms6 also demonstrated adsorption on other inorganic nanoparticles such as titanium oxide, zinc oxide, and hydroxyapatite. Therefore, Mms6 can potentially be utilized for the bioconjugation of functional proteins to inorganic material surfaces to modulate inorganic nanoparticles for biomedical and medicinal applications.

Nanoarchitected Tough Biological Composites from Assembled Chitinous Scaffolds.

Over hundreds of millions of years, organisms have derived specific sets of traits in response to common selection pressures that serve as guideposts for optimal biological designs. A prime example is the evolution of toughened structures in disparate lineages within plants, invertebrates, and vertebrates. Extremely tough structures can function much like armor, battering rams, or reinforcements that enhance the ability of organisms to win competitions, find mates, acquire food, escape predation, and withstand high winds or turbulent flow. From an engineering perspective, biological solutions are intriguing because they must work in a multifunctional context. An organism rarely can be optimally designed for only one function or one environmental condition. Some of these natural systems have developed well-orchestrated strategies, exemplified in the biological tissues of numerous animal and plant species, to synthesize and construct materials from a limited selection of available starting materials. The resulting structures display multiscale architectures with incredible fidelity and often exhibit properties that are similar, and frequently superior, to mechanical properties exhibited by many engineered materials. These biological systems have accomplished this feat through the demonstrated ability to tune size, morphology, crystallinity, phase, and orientation of minerals under benign processing conditions (i.e., near-neutral pH, room temperature, etc.) by establishing controlled synthesis and hierarchical 3D assembly of nano- to microscaled building blocks. These systems utilize organic-inorganic interactions and carefully controlled microenvironments that enable kinetic control during the synthesis of inorganic structures. This controlled synthesis and assembly requires orchestration of mineral transport and nucleation. The underlying organic framework, often consisting of polysaccharides and polypeptides, in these composites is critical in the spatial and temporal regulation of these processes. In fact, the organic framework is used not only to provide transport networks for mineral precursors to nucleation sites but also to precisely guide the formation and phase development of minerals and significantly improve the mechanical performance of otherwise brittle materials.Over the past 15 years, we have focused on a few of these extreme performing organisms, (Wang , Adv. Funct. Mater. 2013, 23, 2908; Weaver , Science 2012, 336, 1275; Huang , Nat. Mater. 2020, 19, 1236; Rivera , Nature 2020, 586, 543) investigating not only their ultrastructural features and mechanical properties but in some cases, how these assembled structures are mineralized. In specific instances, comparative analyses of multiscale structures have pinpointed which design principles have arisen convergently; when more than one evolutionary path arrives at the same solution, we have a good indication that it is the best solution. This is required for survival under extreme conditions. Indeed, we have found that there are specific architectural features that provide an advantage toward survival by enabling the ability to feed effectively or to survive against predatory attacks. In this Account, we describe 3 specific design features, nanorods, helicoids, and nanoparticles, as well as the interfaces in fiber-reinforced biological composites. We not only highlight their roles in the specific organisms but also describe how controlled syntheses and hierarchical assembly using organic (i.e., often chitinous) scaffolds lead to these integrated macroscale structures. Beyond this, we provide insight into multifunctionality: how nature leverages these existing structures to potentially add an additional dimension toward their utility and describe their translation to biomimetic materials used for engineering applications.

A natural impact-resistant bicontinuous composite nanoparticle coating.

Nature utilizes the available resources to construct lightweight, strong and tough materials under constrained environmental conditions. The impact surface of the fast-striking dactyl club from the mantis shrimp is an example of one such composite material; the shrimp has evolved the capability to localize damage and avoid catastrophic failure from high-speed collisions during its feeding activities. Here we report that the dactyl club of mantis shrimps contains an impact-resistant coating composed of densely packed (about 88 per cent by volume) ~65-nm bicontinuous nanoparticles of hydroxyapatite integrated within an organic matrix. These mesocrystalline hydroxyapatite nanoparticles are assembled from small, highly aligned nanocrystals. Under impacts of high strain rates (around 104 s-1), particles rotate and translate, whereas the nanocrystalline networks fracture at low-angle grain boundaries, form dislocations and undergo amorphization. The interpenetrating organic network provides additional toughening, as well as substantial damping, with a loss coefficient of around 0.02. An unusual combination of stiffness and damping is therefore achieved, outperforming many engineered materials.

Mechanism of water extraction from gypsum rock by desert colonizing microorganisms.

Microorganisms, in the most hyperarid deserts around the world, inhabit the inside of rocks as a survival strategy. Water is essential for life, and the ability of a rock substrate to retain water is essential for its habitability. Here we report the mechanism by which gypsum rocks from the Atacama Desert, Chile, provide water for its colonizing microorganisms. We show that the microorganisms can extract water of crystallization (i.e., structurally ordered) from the rock, inducing a phase transformation from gypsum (CaSO4·2H2O) to anhydrite (CaSO4). To investigate and validate the water extraction and phase transformation mechanisms found in the natural geological environment, we cultivated a cyanobacterium isolate on gypsum rock samples under controlled conditions. We found that the cyanobacteria attached onto high surface energy crystal planes ({011}) of gypsum samples generate a thin biofilm that induced mineral dissolution accompanied by water extraction. This process led to a phase transformation to an anhydrous calcium sulfate, anhydrite, which was formed via reprecipitation and subsequent attachment and alignment of nanocrystals. Results in this work not only shed light on how microorganisms can obtain water under severe xeric conditions but also provide insights into potential life in even more extreme environments, such as Mars, as well as offering strategies for advanced water storage methods.

Large-Scale Low-Rank Matrix Learning with Nonconvex Regularizers.

Low-rank modeling has many important applications in computer vision and machine learning. While the matrix rank is often approximated by the convex nuclear norm, the use of nonconvex low-rank regularizers has demonstrated better empirical performance. However, the resulting optimization problem is much more challenging. Recent state-of-the-art requires an expensive full SVD in each iteration. In this paper, we show that for many commonly-used nonconvex low-rank regularizers, the singular values obtained from the proximal operator can be automatically threshold. This allows the proximal operator to be efficiently approximated by the power method. We then develop a fast proximal algorithm and its accelerated variant with inexact proximal step. It can be guaranteed that the squared distance between consecutive iterates converges at a rate of $O(1/T)$O(1/T), where $T$T is the number of iterations. Furthermore, we show the proposed algorithm can be parallelized, and the resultant algorithm achieves nearly linear speedup w.r.t. the number of threads. Extensive experiments are performed on matrix completion and robust principal component analysis. Significant speedup over the state-of-the-art is observed.

A Ranking Approach on Large-Scale Graph With Multidimensional Heterogeneous Information.

Graph-based ranking has been extensively studied and frequently applied in many applications, such as webpage ranking. It aims at mining potentially valuable information from the raw graph-structured data. Recently, with the proliferation of rich heterogeneous information (e.g., node/edge features and prior knowledge) available in many real-world graphs, how to effectively and efficiently leverage all information to improve the ranking performance becomes a new challenging problem. Previous methods only utilize part of such information and attempt to rank graph nodes according to link-based methods, of which the ranking performances are severely affected by several well-known issues, e.g., over-fitting or high computational complexity, especially when the scale of graph is very large. In this paper, we address the large-scale graph-based ranking problem and focus on how to effectively exploit rich heterogeneous information of the graph to improve the ranking performance. Specifically, we propose an innovative and effective semi-supervised PageRank (SSP) approach to parameterize the derived information within a unified semi-supervised learning framework (SSLF-GR), then simultaneously optimize the parameters and the ranking scores of graph nodes. Experiments on the real-world large-scale graphs demonstrate that our method significantly outperforms the algorithms that consider such graph information only partially.