Modern maxims for an AI oracle
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.
Modelling and design of transcriptional enhancers
Transcriptional enhancers are the genomic elements that contain critical information for the regulation of gene expression. This information is encoded through precisely arranged transcription factor-binding sites. Genomic sequence-to-function models, computational models that take DNA sequences as input and predict gene regulatory features, have become essential for unravelling the complex combinatorial rules that govern cell-type-specific activities of enhancers. These models function as biological ‘oracles’, capable of accurately predicting the activity of novel DNA sequences. By leveraging these oracles, DNA sequences can be optimized towards designed synthetic enhancers with tailored cell-type-specific or cell-state-specific activities. In parallel, generative artificial intelligence is rapidly advancing in genomics and enhancer design. Synthetic enhancers hold great promise for a wide range of biomedical applications, from facilitating fundamental research to enabling gene therapies.
Genetic analysis of a Yayoi individual from the Doigahama site provides insights into the origins of immigrants to the Japanese Archipelago
Mainland Japanese have been recognized as having dual ancestry, originating from indigenous Jomon people and immigrants from continental East Eurasia. Although migration from the continent to the Japanese Archipelago continued from the Yayoi to the Kofun period, our understanding of these immigrants, particularly their origins, remains insufficient due to the lack of high-quality genome samples from the Yayoi period, complicating predictions about the admixture process. To address this, we sequenced the whole nuclear genome of a Yayoi individual from the Doigahama site in Yamaguchi prefecture, Japan. A comprehensive population genetic analysis of the Doigahama Yayoi individual, along with ancient and modern populations in East Asia and Northeastern Eurasia, revealed that the Doigahama Yayoi individual, similar to Kofun individuals and modern Mainland Japanese, had three distinct genetic ancestries: Jomon-related, East Asian-related, and Northeastern Siberian-related. Among non-Japanese populations, the Korean population, possessing both East Asian-related and Northeastern Siberian-related ancestries, exhibited the highest degree of genetic similarity to the Doigahama Yayoi individual. The analysis of admixture modeling for Yayoi individuals, Kofun individuals, and modern Japanese respectively supported a two-way admixture model assuming Jomon-related and Korean-related ancestries. These results suggest that between the Yayoi and Kofun periods, the majority of immigrants to the Japanese Archipelago originated primarily from the Korean Peninsula.
Five millennia of mitonuclear discordance in Atlantic bluefin tuna identified using ancient DNA
Mitonuclear discordance between species is readily documented in marine fishes. Such discordance may either be the result of past natural phenomena or the result of recent introgression from previously seperated species after shifts in their spatial distributions. Using ancient DNA spanning five millennia, we here investigate the long-term presence of Pacific bluefin tuna (Thunnus orientalis) and albacore (Thunnus alalunga) -like mitochondrial (MT) genomes in Atlantic bluefin tuna (Thunnus thynnus), a species with extensive exploitation history and observed shifts in abundance and age structure. Comparing ancient (n = 130) and modern (n = 78) Atlantic bluefin MT genomes from most of its range, we detect no significant spatial or temporal population structure, which implies ongoing gene flow between populations and large effective population sizes over millennia. Moreover, we identify discordant MT haplotypes in ancient specimens up to 5000 years old and find that the frequency of these haplotypes has remained similar through time. We therefore conclude that MT discordance in the Atlantic bluefin tuna is not driven by recent introgression. Our observations provide oldest example of directly observed MT discordance in the marine environment, highlighting the utility of ancient DNA to obtain insights in the long-term persistence of such phenomena.
Parallel scaling of elite wealth in ancient Roman and modern cities with implications for understanding urban inequality
Rapid urbanization and rising inequality are pressing global concerns, yet inequality is an ancient trait of city life that may be intrinsically connected to urbanism itself. Here we investigate how elite wealth scales with urban population size across culture and time by analyzing ancient Roman and modern cities. Using Bayesian models to address archeological uncertainties, we uncovered a consistent correlation between population size and physical expressions of elite wealth in urban spaces. These patterns suggest the presence of an ancient, enduring mechanism underlying urban inequality. Supported by an agent-based network simulation and informed by the settlement scaling theory, we propose that the observed patterns arise from common preferential attachment in social networks—a simple, yet powerful, driver of unequal access to interaction potential. Our findings open up new directions in urban scaling research and underscore the importance of understanding long-term urban dynamics to chart a course toward a fairer urban future.
Responses