Search for: All records

Environmentally covarying local adaptation is a form of cryptic local adaptation in which the covariance of the genetic and environmental effects on a phenotype obscures the divergence between locally adapted genotypes. Here, we systematically document the magnitude and drivers of the genetic effect (V G ) for two forms of environmentally covarying local adaptation: counter- and cogradient variation. Using a hierarchical Bayesian meta-analysis, we calculated the overall effect size of V G as 1.05 and 2.13 for populations exhibiting countergradient or cogradient variation, respectively. These results indicate that the genetic contribution to phenotypic variation represents a 1.05 to 2.13 s.d. change in trait value between the most disparate populations depending on if populations are expressing counter- or cogradient variation. We also found that while there was substantial variance among abiotic and biotic covariates, the covariates with the largest mean effects were temperature (2.41) and gamete size (2.81). Our results demonstrate the pervasiveness and large genetic effects underlying environmentally covarying local adaptation in wild populations and highlight the importance of accounting for these effects in future studies. 

Abstract Due to their non-motile nature, plants rely heavily on mutualistic interactions to obtain resources and carry out services. One key mutualism is the plant–microbial mutualism in which a plant trades away carbon to a microbial partner for nutrients like nitrogen and phosphorous. Plants show much variation in the use of this partnership from the individual level to entire lineages depending upon ecological, evolutionary and environmental context. We sought to determine how this context dependency could result in the promotion, exclusion or coexistence of the microbial mutualism by asking if and when the partnership provided a competitive advantage to the plant. To that end, we created a 2 × 2 evolutionary game in which plants could either be a mutualist and pair with a microbe or be a non-mutualist and forgo the partnership. Our model includes both frequency dependence and density dependence, which gives us the eco-evolutionary dynamics of mutualism evolution. As in all models, mutualism only evolved if it could offer a competitive advantage and its net benefit was positive. However, surprisingly the model reveals the possibility of coexistence between mutualist and non-mutualist genotypes due to competition between mutualists over the microbially obtained nutrient. Specifically, frequency dependence of host strategies can make the microbial symbiont less beneficial if the microbially derived resources are shared, a phenomenon that increasingly reduces the frequency of mutualism as the density of competitors increases. In essence, ecological competition can act as a hindrance to mutualism evolution. We go on to discuss basic experiments that can be done to test and falsify our hypotheses. 

While the advent of Graph Neural Networks (GNNs) has greatly improved node and graph representation learning in many applications, the neighborhood aggregation scheme exposes additional vulnerabilities to adversaries seeking to extract node-level information about sensitive attributes. In this paper, we study the problem of protecting sensitive attributes by information obfuscation when learning with graph structured data. We propose a framework to locally filter out pre-determined sensitive attributes via adversarial training with the total variation and the Wasserstein distance. Our method creates a strong defense against inference attacks, while only suffering small loss in task performance. Theoretically, we analyze the effectiveness of our framework against a worst-case adversary, and characterize an inherent trade-off between maximizing predictive accuracy and minimizing information leakage. Experiments across multiple datasets from recommender systems, knowledge graphs and quantum chemistry demonstrate that the proposed approach provides a robust defense across various graph structures and tasks, while producing competitive GNN encoders for downstream tasks. 

Abstract The budding field of materials informatics has coincided with a shift towards artificial intelligence to discover new solid-state compounds. The steady expansion of repositories for crystallographic and computational data has set the stage for developing data-driven models capable of predicting a bevy of physical properties. Machine learning methods, in particular, have already shown the ability to identify materials with near ideal properties for energy-related applications by screening crystal structure databases. However, examples of the data-guided discovery of entirely new, never-before-reported compounds remain limited. The critical step for determining if an unknown compound is synthetically accessible is obtaining the formation energy and constructing the associated convex hull. Fortunately, this information has become widely available through density functional theory (DFT) data repositories to the point that they can be used to develop machine learning models. In this Review, we discuss the specific design choices for developing a machine learning model capable of predicting formation energy, including the thermodynamic quantities governing material stability. We investigate several models presented in the literature that cover various possible architectures and feature sets and find that they have succeeded in uncovering new DFT-stable compounds and directing materials synthesis. To expand access to machine learning models for synthetic solid-state chemists, we additionally presentMatLearn. This web-based application is intended to guide the exploration of a composition diagram towards regions likely to contain thermodynamically accessible inorganic compounds. Finally, we discuss the future of machine-learned formation energy and highlight the opportunities for improved predictive power toward the synthetic realization of new energy-related materials. 

Prader-Willi syndrome (PWS) is caused by the loss of function of the paternally inherited 15q11-q13 locus. This region is governed by genomic imprinting, a phenomenon in which genes are expressed exclusively from one parental allele. The genomic imprinting of the 15q11-q13 locus is established in the germline and is largely controlled by a bipartite imprinting centre. One part, termed the Prader-Willi syndrome imprinting center (PWS-IC), comprises a CpG island that is unmethylated on the paternal allele and methylated on the maternal allele. The second part, termed the Angelman syndrome imprinting centre, is required to silence the PWS_IC in the maternal germline. The loss of the paternal contribution of the imprinted 15q11-q13 locus most frequently occurs owing to a large deletion of the entire imprinted region but can also occur through maternal uniparental disomy or an imprinting defect. While PWS is considered a contiguous gene syndrome based on large-deletion and uniparental disomy patients, the lack of expression of only non-coding RNA transcripts from the SNURF-SNRPN/SNHG14 may be the primary cause of PWS. Patients with small atypical deletions of the paternal SNORD116 cluster alone appear to have most of the PWS related clinical phenotypes. The loss of the maternal contribution of the 15q11-q13 locus causes a separate and distinct condition called Angelman syndrome. Importantly, while much has been learned about the regulation and expression of genes and transcripts deriving from the 15q11-q13 locus, there remains much to be learned about how these genes and transcripts contribute at the molecular level to the clinical traits and developmental aspects of PWS that have been observed.