Publications

The hierarchical prior used in Latent Gaussian models (LGMs) induces a posterior geometry prone to frustrate inference algorithms. Marginalizing out the latent Gaussian variable using an integrated Laplace approximation removes the offending geometry, allowing us to do efficient inference on the hyperparameters. To use gradient-based inference we need to compute the approximate marginal likelihood and its gradient. The adjoint-differentiated Laplace approximation differentiates the marginal likelihood and scales well with the dimension of the hyperparameters. While this method can be applied to LGMs with any prior covariance, it only works for likelihoods with a diagonal Hessian. Furthermore, the algorithm requires methods which compute the first three derivatives of the likelihood with current implementations relying on analytical derivatives. I propose a generalization which is applicable to a broader class of likelihoods and does not require analytical derivatives of the likelihood. Numerical experiments suggest the added flexibility comes at no computational cost: on a standard LGM, the new method is in fact slightly faster than the existing adjoint-differentiated Laplace approximation. I also apply the general method to an LGM with an unconventional likelihood. This example highlights the algorithm's potential, as well as persistent challenges.

Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by heterogeneous cognitive, behavioral and communication impairments. Disruption of the gut–brain axis (GBA) has been implicated in ASD although with limited reproducibility across studies. In this study, we developed a Bayesian differential ranking algorithm to identify ASD-associated molecular and taxa profiles across 10 cross-sectional microbiome datasets and 15 other datasets, including dietary patterns, metabolomics, cytokine profiles and human brain gene expression profiles. We found a functional architecture along the GBA that correlates with heterogeneity of ASD phenotypes, and it is characterized by ASD-associated amino acid, carbohydrate and lipid profiles predominantly encoded by microbial species in the genera Prevotella, Bifidobacterium, Desulfovibrio and Bacteroides and correlates with brain gene expression changes, restrictive dietary patterns and pro-inflammatory cytokine profiles. The functional architecture revealed in age-matched and sex-matched cohorts is not present in sibling-matched cohorts. We also show a strong association between temporal changes in microbiome composition and ASD phenotypes. In summary, we propose a framework to leverage multi-omic datasets from well-defined cohorts and investigate how the GBA influences ASD.

One promising way to extract information about stellar astrophysics from a gravitational-wave catalog is to compare the catalog to the outputs of stellar population synthesis modeling with varying physical assumptions. The parameter space of physical assumptions in population synthesis is high-dimensional and the choice of parameters that best represents the evolution of a binary system may depend in an as-yet-to-be-determined way on the system's properties. Here we propose a pipeline to simultaneously infer zero-age main-sequence properties and population synthesis parameter settings controlling modeled binary evolution from individual gravitational-wave observations of merging compact binaries. Our pipeline can efficiently explore the high-dimensional space of population synthesis settings and progenitor system properties for each system in a catalog of gravitational-wave observations. We apply our pipeline to observations in the third LIGO–Virgo Gravitational-Wave Transient Catalog. We showcase the effectiveness of this pipeline with a detailed study of the progenitor properties and population synthesis settings that produce mergers like the observed GW150914. Our pipeline permits a measurement of the variation of population synthesis parameter settings with binary properties, if any; we illustrate the possibility of such capability by presenting inferences for the recent GWTC-3 transient catalog that suggest that the stable mass transfer efficiency parameter may vary with primary black hole mass.

Study question:
Is there any relationship between the relative telomere length (RTL) within cumulus cells (CCs) and the outcome of assisted reproductive treatment using the corresponding oocyte?

Summary answer:
Lower RTLs in CCs were significantly associated with embryos chosen for transfer or cryopreservation. In contrast, embryos considered non-viable (discarded) tended to have higher RTLs.

We present a suite of high-resolution simulations of an isolated dwarf galaxy using four different hydrodynamical codes: {\sc Gizmo}, {\sc Arepo}, {\sc Gadget}, and {\sc Ramses}. All codes adopt the same physical model which includes radiative cooling, photoelectric heating, star formation, and supernova (SN) feedback. Individual SN explosions are directly resolved without resorting to sub-grid models, eliminating one of the major uncertainties in cosmological simulations. We find reasonable agreement on the time-averaged star formation rates as well as the joint density-temperature distributions between all codes. However, the Lagrangian codes show significantly burstier star formation, larger supernova-driven bubbles, and stronger galactic outflows compared to the Eulerian code. This is caused by the behavior in the dense, collapsing gas clouds when the Jeans length becomes unresolved: gas in Lagrangian codes collapses to much higher densities than in Eulerian codes, as the latter is stabilized by the minimal cell size. Therefore, more of the gas cloud is converted to stars and SNe are much more clustered in the Lagrangian models, amplifying their dynamical impact. The differences between Lagrangian and Eulerian codes can be reduced by adopting a higher star formation efficiency in Eulerian codes, which significantly enhances SN clustering in the latter. Adopting a zero SN delay time reduces burstiness in all codes, resulting in vanishing outflows as SN clustering is suppressed.

Differentiable simulation enables gradients to be back-propagated through physics simulations. In this way, one can learn the dynamics and properties of a physics system by gradient-based optimization or embed the whole differentiable simulation as a layer in a deep learning model for downstream tasks, such as planning and control. However, differentiable simulation at its current stage is not perfect and might provide wrong gradients that deteriorate its performance in learning tasks. In this paper, we study differentiable rigid-body simulation with contacts. We find that existing differentiable simulation methods provide inaccurate gradients when the contact normal direction is not fixed - a general situation when the contacts are between two moving objects. We propose to improve gradient computation by continuous collision detection and leverage the time-of-impact (TOI) to calculate the post-collision velocities. We demonstrate our proposed method, referred to as TOI-Velocity, on two optimal control problems. We show that with TOI-Velocity, we are able to learn an optimal control sequence that matches the analytical solution, while without TOI-Velocity, existing differentiable simulation methods fail to do so.

Several recent works have introduced highly compact representations of single-particle Green's functions in the imaginary time and Matsubara frequency domains, as well as efficient interpolation grids used to recover the representations. In particular, the intermediate representation with sparse sampling and the discrete Lehmann representation (DLR) make use of low rank compression techniques to obtain optimal approximations with controllable accuracy. We consider the use of the DLR in dynamical mean-field theory (DMFT) calculations, and in particular show that the standard full Matsubara frequency grid can be replaced by the compact grid of DLR Matsubara frequency nodes. We test the performance of the method for a DMFT calculation of Sr$_2$RuO$_4$ at temperature $50$K using a continuous-time quantum Monte Carlo impurity solver, and demonstrate that Matsubara frequency quantities can be represented on a grid of only 36 nodes with no reduction in accuracy, or increase in the number of self-consistent iterations, despite the presence of significant Monte Carlo noise.

The millisecond pulsar J1713+0747 underwent a sudden and significant pulse shape change between April 16 and 17, 2021 (MJDs 59320 and 59321). Subsequently, the pulse shape gradually recovered over the course of several months. We report the results of continued multi-frequency radio observations of the pulsar made using the Canadian Hydrogen Intensity Mapping Experiment (CHIME) and the 100-meter Green Bank Telescope (GBT) in a three-year period encompassing the shape change event, between February 2020 and February 2023. As of February 2023, the pulse shape had returned to a state similar to that seen before the event, but with measurable changes remaining. The amplitude of the shape change and the accompanying TOA residuals display a strong non-monotonic dependence on radio frequency, demonstrating that the event is neither a glitch (the effects of which should be independent of radio frequency, ν) nor a change in dispersion measure (DM) alone (which would produce a delay proportional to ν−2). However, it does bear some resemblance to the two previous "chromatic timing events" observed in J1713+0747 (Demorest et al. 2013; Lam et al. 2016), as well as to a similar event observed in PSR J1643-1224 in 2015 (Shannon et al. 2016).

Cryo-electron microscopy (cryo-EM) has recently become a leading method for obtaining high-resolution structures of biological macromolecules. However, cryo-EM is limited to biomolecular samples with low conformational heterogeneity, where most conformations can be well-sampled at various projection angles. While cryo-EM provides single-molecule data for heterogeneous molecules, most existing reconstruction tools cannot retrieve the ensemble distribution of possible molecular conformations from these data. To overcome these limitations, we build on a previous Bayesian approach and develop an ensemble refinement framework that estimates the ensemble density from a set of cryo-EM particle images by reweighting a prior conformational ensemble, e.g., from molecular dynamics simulations or structure prediction tools. Our work provides a general approach to recovering the equilibrium probability density of the biomolecule directly in conformational space from single-molecule data. To validate the framework, we study the extraction of state populations and free energies for a simple toy model and from synthetic cryo-EM particle images of a simulated protein that explores multiple folded and unfolded conformations.