Related Articles
Advancing extrapolative predictions of material properties through learning to learn using extrapolative episodic training
Recent advancements in machine learning have demonstrated its potential to significantly accelerate the discovery of new materials. Central to this progress is the development of rapidly computable property predictors, which allow identifying novel materials with the desired properties from vast material spaces. However, the limited availability of data resources poses a significant challenge in data-driven material research, particularly hindering the exploration of innovative materials beyond the boundaries of existing data. Although machine-learning predictors are inherently interpolative, establishing a general methodology to create an extrapolative predictor remains a fundamental challenge. In this study, we leveraged the attention-based architecture of neural networks and a meta-learning algorithm to enhance extrapolative generalization capabilities. Meta-learners trained repeatedly on arbitrarily generated extrapolative tasks show outstanding generalization for unexplored material spaces. Through the tasks of predicting the physical properties of polymeric materials and hybrid organic–inorganic perovskites, we highlight the potential of such extrapolatively trained models, particularly their ability to rapidly adapt to unseen material domains in transfer-learning scenarios.
Predation-resistant Pseudomonas bacteria engage in symbiont-like behavior with the social amoeba Dictyostelium discoideum
The soil amoeba Dictyostelium discoideum acts as both a predator and potential host for diverse bacteria. We tested fifteen Pseudomonas strains that were isolated from transiently infected wild D. discoideum for ability to escape predation and infect D. discoideum fruiting bodies. Three predation-resistant strains frequently caused extracellular infections of fruiting bodies but were not found within spores. Furthermore, infection by one of these species induces secondary infections and suppresses predation of otherwise edible bacteria. Another strain can persist inside of amoebae after being phagocytosed but is rarely taken up. We sequenced isolate genomes and discovered that predation-resistant isolates are not monophyletic. Many Pseudomonas isolates encode secretion systems and toxins known to improve resistance to phagocytosis in other species, as well as diverse secondary metabolite biosynthetic gene clusters that may contribute to predation resistance. However, the distribution of these genes alone cannot explain why some strains are edible and others are not. Each lineage may employ a unique mechanism for resistance.
Coevolution between marine Aeromonas and phages reveals temporal trade-off patterns of phage resistance and host population fitness
Coevolution of bacteria and phages is an important host and parasite dynamic in marine ecosystems, contributing to the understanding of bacterial community diversity. On the time scale, questions remain concerning what is the difference between phage resistance patterns in marine bacteria and how advantageous mutations gradually accumulate during coevolution. In this study, marine Aeromonas was co-cultured with its phage for 180 days and their genetic and phenotypic dynamics were measured every 30 days. We identified 11 phage resistance genes and classified them into three categories: lipopolysaccharide (LPS), outer membrane protein (OMP), and two-component system (TCS). LPS shortening and OMP mutations are two distinct modes of complete phage resistance, while TCS mutants mediate incomplete resistance by repressing the transcription of phage genes. The co-mutation of LPS and OMP was a major mode for bacterial resistance at a low cost. The mutations led to significant reductions in the growth and virulence of bacterial populations during the first 60 days of coevolution, with subsequent leveling off. Our findings reveal the marine bacterial community dynamics and evolutionary trade-offs of phage resistance during coevolution, thus granting further understanding of the interaction of marine microbes.
Abundant Sulfitobacter marine bacteria protect Emiliania huxleyi algae from pathogenic bacteria
Emiliania huxleyi is a unicellular micro-alga that forms massive oceanic blooms and plays key roles in global biogeochemical cycles. Mounting studies demonstrate various stimulatory and inhibitory influences that bacteria have on the E. huxleyi physiology. To investigate these algal-bacterial interactions, laboratory co-cultures have been established by us and by others. Owing to these co-cultures, various mechanisms of algal-bacterial interactions have been revealed, many involving bacterial pathogenicity towards algae. However, co-cultures represent a significantly simplified system, lacking the complexity of bacterial communities. In order to investigate bacterial pathogenicity within an ecologically relevant context, it becomes imperative to enhance the microbial complexity of co-culture setups. Phaeobacter inhibens bacteria are known pathogens that cause the death of E. huxleyi algae in laboratory co-culture systems. The bacteria depend on algal exudates for growth, but when algae senesce, bacteria switch to a pathogenic state and induce algal death. Here we investigate whether P. inhibens bacteria can induce algal death in the presence of a complex bacterial community. We show that an E. huxleyi-associated bacterial community protects the alga from the pathogen, although the pathogen occurs within the community. To study how the bacterial community regulates pathogenicity, we reduced the complex bacterial community to a five-member synthetic community (syncom). The syncom is comprised of a single algal host and five isolated bacterial species, which represent major bacterial groups that are naturally associated with E. huxleyi. We discovered that a single bacterial species in the reduced community, Sulfitobacter pontiacus, protects the alga from the pathogen. We further found that algal protection from P. inhibens pathogenicity is a shared trait among several Sulfitobacter species. Algal protection by bacteria might be a common phenomenon with ecological significance, which is overlooked in reduced co-culture systems.
Enantioselective C–H annulations enabled by either nickel- or cobalt-electrocatalysed C–H activation for catalyst-controlled chemodivergence
Enantioselective electrocatalysis shows unique potential for the sustainable assembly of enantiomerically enriched molecules. This approach allows electro-oxidative C–H activation to be performed paired to the hydrogen evolution reaction. Recent progress has featured scarce transition metals with limited availability. Here we reveal that the earth-abundant 3d transition metals nickel and cobalt exhibit distinctive performance for enantioselective electrocatalysis with chemodivergent reactivity patterns. Enantioselective desymmetrizations of strained bicyclic alkenes were achieved through C–H annulations. A data-driven optimization of chiral N,O-bidentate salicyloxazoline-type ligands was crucial for enhancing enantioselectivity in nickel electrocatalysis. Notably, in the transition state of the enantio-determining step, secondary weak attractive π–π and CH–π interactions were identified, reflecting the informed adaptations in the ligand design. Detailed mechanistic investigations by experimental and computational studies revealed for the nickel electrocatalysis a C–N bond-forming reductive elimination from nickel(III) and for the cobalt electrocatalysis a C–C bond-forming nucleophilic addition from cobalt(III) as the product-determining steps.
Responses