BNBenjamin E. Nelson

Benjamin E. Nelson

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Benjamin E. Nelson has a total of 120 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Top Cited Papers

    TranscriptFormer: A generative cell atlas across 1.5 billion years of evolution

    by James D. Pearce, Sara E. Simmonds, Gita Mahmoudabadi, Lakshmi Krishnan, Giovanni Palla, Ana-Maria Istrate, Alexander Tarashansky, Benjamin Nelson, Omar Valenzuela, Donghui Li, et al.

    Single-cell transcriptomics is revolutionizing our understanding of cellular diversity, yet comparing transcriptional programs across the tree of life remains challenging. We developed TranscriptFormer, a family of generative foundation models trained on up to 112 million cells spanning 1.53 billion years of evolution across 12 species. We demonstrate state-of-the-art performance on cell type classification, even for species separated over 685 million years of evolution, and zero-shot disease state identification in human cells. Developmental trajectories, phylogenetic relationships and cellular hierarchies emerge naturally in TranscriptFormer’s representations without any explicit training on these annotations. This work establishes a powerful framework for quantitative single-cell analysis and comparative cellular biology, thus demonstrating that universal principles of cellular organization can be learned and predicted across the tree of life.


  • Open Access

    A Cross-Species Generative Cell Atlas Across 1.5 Billion Years of Evolution: The TranscriptFormer Single-cell Model

    by James D Pearce, Sara E Simmonds, Gita Mahmoudabadi, Lakshmi Krishnan, Giovanni Palla, Ana-Maria Istrate, Alexander Tarashansky, Benjamin Nelson, Omar Valenzuela, Donghui Li, et al.

    Single-cell transcriptomics has revolutionized our understanding of cellular diversity, yet our understanding of the transcriptional programs across the tree of life remains limited. Here we present TranscriptFormer, a family of generative foundation models trained on up to 112 million cells spanning 1.53 billion years of evolution across 12 species. By jointly modeling gene identities and expression levels using a novel generative architecture, TranscriptFormer encodes multi-scale biological structure, functioning as a queryable virtual cell atlas. We demonstrate state-of-the-art performance on both in-distribution and out-of-distribution cell type classification, with robust performance even for species separated by over 685 million years of evolution. TranscriptFormer can also perform zero-shot disease state identification in human cells and accurately transfers cell state annotations across species boundaries. As a generative model, TranscriptFormer can be prompted to predict cell type-specific transcription factors and gene-gene interactions that align with independent experimental observations. Developmental trajectories, phylogenetic relationships and cellular hierarchies emerge naturally in TranscriptFormer’s representations without any explicit training on these annotations. This work establishes a powerful framework for quantitative single-cell analysis, and comparative cellular biology, thus demonstrating that universal principles of cellular organization can be learned and predicted across the tree of life.


  • Top Cited Papers

    A data portal for providing standardized annotations for cryo-electron tomography

    by Utz Ermel, Anchi Cheng, Jun Xi Ni, Jessica Gadling, Manasa Venkatakrishnan, Kira Evans, Jeremy Asuncion, Andrew Sweet, Janeece Pourroy, Zun Shi Wang, et al.


  • Open Access

    Sensitivity Analyses of Exoplanet Occurrence Rates from Kepler and Gaia

    by Megan Shabram, Natalie Batalha, Susan E. Thompson, Danley C. Hsu, Eric B. Ford, Jessie L. Christiansen, Daniel Huber, Travis Berger, Joseph Catanzarite, Benjamin E. Nelson, et al.

    We infer the number of planets-per-star as a function of orbital period and planet size using $Kepler$ archival data products with updated stellar properties from the $Gaia$ Data Release 2. Using hierarchical Bayesian modeling and Hamiltonian Monte Carlo, we incorporate planet radius uncertainties into an inhomogeneous Poisson point process model. We demonstrate that this model captures the general features of the outcome of the planet formation and evolution around GK stars, and provides an infrastructure to use the $Kepler$ results to constrain analytic planet distribution models. We report an increased mean and variance in the marginal posterior distributions for the number of planets per $GK$ star when including planet radius measurement uncertainties. We estimate the number of planets-per-$GK$ star between 0.75 and 2.5 $R_{\oplus}$ and 50 to 300 day orbital periods to have a $68\%$ credible interval of $0.49$ to $0.77$ and a posterior mean of $0.63$. This posterior has a smaller mean and a larger variance than the occurrence rate calculated in this work and in Burke et al. (2015) for the same parameter space using the $Q1-Q16$ (previous $Kepler$ planet candidate and stellar catalog). We attribute the smaller mean to many of the instrumental false positives at longer orbital periods being removed from the $DR25$ catalog. We find that the accuracy and precision of our hierarchical Bayesian model posterior distributions are less sensitive to the total number of planets in the sample, and more so on the characteristics of the catalog completeness and reliability and the span of the planet parameter space.

  • Quantifying the Bayesian Evidence for a Planet in Radial Velocity Data

    by Benjamin E. Nelson, Eric B. Ford, Johannes Buchner, Ryan Cloutier, Rodrigo F. Díaz, João P. Faria, Nathan C. Hara, Vinesh M. Rajpaul, Surangkhana Rukdee

    We present results from a data challenge posed to the radial velocity (RV) community: namely, to quantify the Bayesian “evidence” for n = {0, 1, 2, 3} planets in a set of synthetically generated RV data sets containing a range of planet signals. Participating teams were provided the same likelihood function and set of priors to use in their analysis. They applied a variety of methods to estimate , the marginal likelihood for each n-planet model, including cross-validation, the Laplace approximation, importance sampling, and nested sampling. We found the dispersion in across different methods grew with increasing n-planet models: ∼3 for zero planets, ∼10 for one planet, ∼102–103 for two planets, and >104 for three planets. Most internal estimates of uncertainty in for individual methods significantly underestimated the observed dispersion across all methods. Methods that adopted a Monte Carlo approach by comparing estimates from multiple runs yielded plausible uncertainties. Finally, two classes of numerical algorithms (those based on importance and nested samplers) arrived at similar conclusions regarding the ratio of for n- and (n + 1)-planet models. One analytic method (the Laplace approximation) demonstrated comparable performance. We express both optimism and caution: we demonstrate that it is practical to perform rigorous Bayesian model comparison for models of ≤3 planets, yet robust planet discoveries require researchers to better understand the uncertainty in and its connections to model selection.


  • Forming Gliese 876 through Smooth Disk Migration

    by Adam M. Dempsey, Benjamin E. Nelson

    We run a suite of dissipative N-body simulations to determine which regions of phase space for smooth disk migration are consistent with the GJ876 system, an M-dwarf hosting three planets orbiting in a chaotic 4:2:1 Laplace resonance. We adopt adaptive mesh refinement (AMR) methods that are commonly used in hydrodynamical simulations to efficiently explore the parameter space defined by the semimajor axis and eccentricity damping timescales. We find that there is a large region of phase space that produces systems in the chaotic Laplace resonance and a smaller region consistent with the observed eccentricities and libration amplitudes for the resonant angles. Under the assumptions of Type I migration for the outer planet, we translate these damping timescales into constraints on the protoplanetary disk surface density and thickness. When we strongly (weakly) damp the eccentricities of the inner two Laplace planets, these timescales correspond to disk surface densities around ten thousand (a few hundred) grams per square centimeter and disk aspect ratios between 1% and 10%. Additionally, smooth migration produces systems with a range of chaotic timescales, from decades and centuries to upward of thousands of years. In agreement with previous studies, the less chaotic regions of phase space coincide with the system being in a low-energy double apsidal corotation resonance. Our detailed modeling of multiplanetary systems coupled with our AMR exploration method enhances our ability to map out the parameter space of planet formation models and is well suited to study other resonant chain systems such as Trappist-1, Kepler-60, and others.


  • Evidence for Two Hot-Jupiter Formation Paths

    by Benjamin E. Nelson, Eric B. Ford, Frederic A. Rasio

    Disk migration and high-eccentricity migration are two well-studied theories to explain the formation of hot Jupiters. The former predicts that these planets can migrate up until the planet–star Roche separation ( ) and the latter predicts they will tidally circularize at a minimum distance of 2 . Considering long-running radial velocity and transit surveys have identified a couple hundred hot Jupiters to date, we can revisit the classic question of hot-Jupiter formation in a data-driven manner. We approach this problem using data from several exoplanet surveys (radial velocity, Kepler, HAT, and WASP) allowing for either a single population or a mixture of populations associated with these formation channels, and applying a hierarchical Bayesian mixture model of truncated power laws of the form to constrain the population-level parameters of interest (e.g., location of inner edges, γ, mixture fractions). Within the limitations of our chosen models, we find that the current radial velocity and Kepler sample of hot Jupiters can be well explained with a single truncated power-law distribution with a lower cutoff near 2 , a result that still holds after a decade, and . However, the HAT and WASP data show evidence for multiple populations (Bayes factor ). We find that reside in a component consistent with disk migration ( ) and in one consistent with high-eccentricity migration ( ). We find no immediately strong connections with some observed host star properties and speculate on how future exoplanet surveys could improve upon hot-Jupiter population inference.


  • An empirically derived three-dimensional Laplace resonance in the Gliese 876 planetary system

    by Benjamin E. Nelson, Paul M. Robertson, Matthew J. Payne, Seth M. Pritchard, Katherine M. Deck, Eric B. Ford, Jason T. Wright, Howard T. Isaacson

    We report constraints on the three-dimensional orbital architecture for all four planets known to orbit the nearby M dwarf Gliese 876 based solely on Doppler measurements and demanding long-term orbital stability. Our data set incorporates publicly available radial velocities taken with the ELODIE and CORALIE spectrographs, High Accuracy Radial velocity Planet Searcher (HARPS), and Keck HIgh Resolution Echelle Spectrometer (HIRES) as well as previously unpublished HIRES velocities. We first quantitatively assess the validity of the planets thought to orbit GJ 876 by computing the Bayes factors for a variety of different coplanar models using an importance sampling algorithm. We find that a four-planet model is preferred over a three-planet model. Next, we apply a Newtonian Markov chain Monte Carlo algorithm to perform a Bayesian analysis of the planet masses and orbits using an N-body model in three-dimensional space. Based on the radial velocities alone, we find that a 99 per cent credible interval provides upper limits on the mutual inclinations for the three resonant planets (Φcb < 6 $_{.}^{\circ}$ 20 for the ${c}$ and ${b}$ pair and Φbe < 28 $_{.}^{\circ}$ 5 for the ${b}$ and ${e}$ pair). Subsequent dynamical integrations of our posterior sample find that the GJ 876 planets must be roughly coplanar (Φcb < 2 $_{.}^{\circ}$ 60 and Φbe < 7 $_{.}^{\circ}$ 87), suggesting that the amount of planet–planet scattering in the system has been low. We investigate the distribution of the respective resonant arguments of each planet pair and find that at least one argument for each planet pair and the Laplace argument librate. The libration amplitudes in our three-dimensional orbital model support the idea of the outer three planets having undergone significant past disc migration.


  • Open Access

    An Empirically Derived Three-Dimensional Laplace Resonance in the Gliese 876 Planetary System

    by Benjamin E. Nelson, Paul Robertson, Matthew J. Payne, Seth M. Pritchard, Katherine M. Deck, Eric B. Ford, Jason T. Wright, Howard Isaacson

    We report constraints on the three-dimensional orbital architecture for all four planets known to orbit the nearby M dwarf Gliese 876 based solely on Doppler measurements and demanding long-term orbital stability. Our dataset incorporates publicly available radial velocities taken with the ELODIE and CORALIE spectrographs, HARPS, and Keck HIRES as well as previously unpublished HIRES velocities. We first quantitatively assess the validity of the planets thought to orbit GJ 876 by computing the Bayes factors for a variety of different coplanar models using an importance sampling algorithm. We find that a four-planet model is preferred over a three-planet model. Next, we apply a Newtonian MCMC algorithm to perform a Bayesian analysis of the planet masses and orbits using an n-body model in three-dimensional space. Based on the radial velocities alone, we find that a 99% credible interval provides upper limits on the mutual inclinations for the three resonant planets ($\Phi_{cb}<6.20^\circ$ for the "c" and "b" pair and $\Phi_{be}<28.5^\circ$ for the "b" and "e" pair). Subsequent dynamical integrations of our posterior sample find that the GJ 876 planets must be roughly coplanar ($\Phi_{cb}<2.60^\circ$ and $\Phi_{be}<7.87^\circ$), suggesting the amount of planet-planet scattering in the system has been low. We investigate the distribution of the respective resonant arguments of each planet pair and find that at least one argument for each planet pair and the Laplace argument librate. The libration amplitudes in our three-dimensional orbital model supports the idea of the outer-three planets having undergone significant past disk migration.