Search for: All records

Conditional generation models for longitudinal sequences can produce new or modified trajectories given a conditioning input. However, they often lack control over when the condition should take effect (timing) and which variables it should influence (scope). Most methods either operate only on univariate sequences or assume that the condition alters all variables and time steps. In scientific and clinical settings, interventions instead begin at a specific moment, such as the time of drug administration or surgery, and influence only a subset of measurements while the rest of the trajectory remains unchanged. CLEF learns temporal concepts that encode how and when a condition alters future sequence evolution. These concepts allow CLEF to apply targeted edits to the affected time steps and variables while preserving the rest of the sequence. We evaluate CLEF on 8 datasets spanning cellular reprogramming, patient health, and sales, comparing against 9 state-of-the-art baselines. CLEF improves immediate sequence editing accuracy by 16.28% (MAE) on average against their non-CLEF counterparts. Unlike prior models, CLEF enables one-step conditional generation at arbitrary future times, outperforming their non-CLEF counterparts in delayed sequence editing by 26.73% (MAE) on average. We test CLEF under counterfactual inference assumptions and show up to 62.84% (MAE) improvement on zero-shot conditional generation of counterfactual trajectories. In a case study of patients with type 1 diabetes mellitus, CLEF identifies clinical interventions that generate realistic counterfactual trajectories shifted toward healthier outcomes. 

Abstract Understanding how mixed-species forests uptake subsurface water sources is critical to projecting future forest water use and stress. Variation in root water uptake (RWU) depths and volumes is common among trees but it is unclear how it is affected by species identity, local water availability or neighboring tree species compositions. We evaluated the hypothesis that RWU depths and the age of water (i.e., time since water entered soils as precipitation) taken up by red maples (Acer rubrum) varied significantly between two forested plots, both containing red maples, similar soils, topography and hydrologic conditions, but having different neighboring tree species. We measured soil moisture contents as well as stable isotopes (δ2H, δ18O) in plant xylem water and soil moisture across two years. These data were used to calibrate process-based stand-level ecohydrological models for each plot to estimate species-level RWU depths. Model calibration suggested significant differences in red maple tree RWU depths, transpiration rates and the ages of water taken up by maples across the two stands. Maple trees growing with ash and white spruce relied on significantly deeper and older water from the soil profile than maple trees growing with birch and oak. The drought risk profile experienced by maple trees differed between the plots as demonstrated by strong correlations between precipitation and model simulated transpiration on a weekly time scale for maples taking up shallow soil moisture and a monthly time scale for maples reliant on deeper soil moisture. These findings carry significant implications for our understanding of water competition in mixed-species forests and for the representation of forest rooting strategies in hydrologic and earth systems models. 

ABSTRACT Sampling of stable isotopes in plant xylem water (δ²H, δ¹⁸O) has become a ubiquitous technique to study spatiotemporal variations in the water taken up by plant roots; however, open questions remain concerning the most appropriate time of day to sample trees to obtain representative xylem water isotopic values (δ_XYLEM). We sampled the δ_XYLEMof oak and maple trees prior to solar midday (i.e., in a recommended sampling window) and then again after solar midday (i.e., outside of the recommended window) across 4 months. The paired root mean squared difference between AM and PM δ¹⁸O ranged from 1.00‰ to 1.16‰ for maples and 0.23‰ to 2.55‰ for oaks across all sampling dates. Xylem water seasonal origin index (SOI) values derived from AM and PM δ_XYLEMsamples were significantly different, though both SOI estimates supported the conclusion that maple and oak δ_XYLEMreflected summer precipitation on all sampling dates. We conclude that sampling time of day is a significant consideration in the design of δ_XYLEMsampling campaigns; however, our findings also support flexibility in the collection time of δ_XYLEMin field sites where sampling during the optimal time of day is challenging. 

Inhomogeneous patterns of chromatin-chromatin contacts within 10–100-kb-sized regions of the genome are a generic feature of chromatin spatial organization. These features, termed topologically associating domains (TADs), have led to the loop extrusion factor (LEF) model. Currently, our ability to model TADs relies on the observation that in vertebrates TAD boundaries are correlated with DNA sequences that bind CTCF, which therefore is inferred to block loop extrusion. However, although TADs feature prominently in their Hi-C maps, non-vertebrate eukaryotes either do not express CTCF or show few TAD boundaries that correlate with CTCF sites. In all of these organisms, the counterparts of CTCF remain unknown, frustrating comparisons between Hi-C data and simulations. ResultsTo extend the LEF model across the tree of life, here, we propose theconserved-current loop extrusion (CCLE) modelthat interprets loop-extruding cohesin as a nearly conserved probability current. From cohesin ChIP-seq data alone, we derive a position-dependent loop extrusion rate, allowing for a modified paradigm for loop extrusion, that goes beyond solely localized barriers to also include loop extrusion rates that vary continuously. We show that CCLE accurately predicts the TAD-scale Hi-C maps of interphaseSchizosaccharomyces pombe, as well as those of meiotic and mitoticSaccharomyces cerevisiae, demonstrating its utility in organisms lacking CTCF. ConclusionsThe success of CCLE in yeasts suggests that loop extrusion by cohesin is indeed the primary mechanism underlying TADs in these systems. CCLE allows us to obtain loop extrusion parameters such as the LEF density and processivity, which compare well to independent estimates. 

Abstract Excitons, bound electron–hole pairs, in two-dimensional hybrid organic inorganic perovskites (2D HOIPs) are capable of forming hybrid light-matter states known as exciton-polaritons (E–Ps) when the excitonic medium is confined in an optical cavity. In the case of 2D HOIPs, they can self-hybridize into E–Ps at specific thicknesses of the HOIP crystals that form a resonant optical cavity with the excitons. However, the fundamental properties of these self-hybridized E–Ps in 2D HOIPs, including their role in ultrafast energy and/or charge transfer at interfaces, remain unclear. Here, we demonstrate that >0.5 µm thick 2D HOIP crystals on Au substrates are capable of supporting multiple-orders of self-hybridized E–P modes. These E–Ps have high Q factors (>100) and modulate the optical dispersion for the crystal to enhance sub-gap absorption and emission. Through varying excitation energy and ultrafast measurements, we also confirm energy transfer from higher energy E–Ps to lower energy E–Ps. Finally, we also demonstrate that E–Ps are capable of charge transport and transfer at interfaces. Our findings provide new insights into charge and energy transfer in E–Ps opening new opportunities towards their manipulation for polaritonic devices. 

Numerical simulations have revolutionized material design. However, although simulations excel at mapping an input material to its output property, their direct application to inverse design has traditionally been limited by their high computing cost and lack of differentiability. Here, taking the example of the inverse design of a porous matrix featuring targeted sorption isotherm, we introduce a computational inverse design framework that addresses these challenges, by programming differentiable simulation on TensorFlow platform that leverages automated end-to-end differentiation. Thanks to its differentiability, the simulation is used to directly train a deep generative model, which outputs an optimal porous matrix based on an arbitrary input sorption isotherm curve. Importantly, this inverse design pipeline leverages the power of tensor processing units (TPU)—an emerging family of dedicated chips, which, although they are specialized in deep learning, are flexible enough for intensive scientific simulations. This approach holds promise to accelerate inverse materials design.