Search NASA⌕ Search

SEARCH · Search NASA

Results for “goodness of fit”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

The outburst of the changing-look AGN IRAS 23226-3843 in 2019

IRAS 23226-3843 has previously been classified as a changing-look active galactic nucleus (AGN) based on observations taken in the 1990s in comparison to X-ray data (Swift, XMM-Newton, and NuSTAR) and optical spectra taken after a very strong X-ray decline in 2017. In 2019, Swift observations revealed a strong rebrightening in X-ray and UV fluxes. We aimed to study this outburst in greater detail. We took follow-up Swift, XMM-Newton, and NuSTAR observations of IRAS 23226-3843 together with optical spectra (SALT and SAAO 1.9 m telescope) from 2019 until 2021. IRAS 23226-3843 showed a strong X-ray and optical outburst in 2019. It varied in the X-ray continuum by a factor of 5 and in the optical continuum by a factor of 1.6 within two months. This corresponds to a factor of 3 after correction for the host galaxy contribution. The Balmer and Fe II emission-line intensities showed comparable variability amplitudes during the outburst in 2019. The Hα emission-line profiles of IRAS 23226-3843 changed from a blue-peaked profile in the years 1997 and 1999 to a broad double-peaked profile in 2017 and 2019. However, there were no major profile variations in the extremely broad double-peaked profiles despite the strong intensity variations in 2019. One year after the outburst, IRAS 23226-3843 changed its optical spectral type and became a Seyfert type 2 object in 2020. Blue outflow components are present in the optical Balmer lines and in the Fe band in the X-rays. A deep broadband XMM-Newton/NuSTAR spectrum was taken during IRAS 23226-3843’s maximum state in 2019. This spectrum is qualitatively very similar to a spectrum taken in 2017, but by a factor of 10 higher. The soft X-ray band appears featureless. The soft excess is well modeled with a Comptonization model. A broadband fit with a power-law continuum, Comptonized soft excess, and Galactic absorption gives a good fit to the combined EPIC-pn and NuSTAR spectrum. In addition, we see a complex and broadened Fe K emission-line profile in the X-rays. The changing-look character in IRAS 23226-3843 is most probably caused by changes in the accretion rate – based on the short-term variations on timescales of weeks to months.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Testing α -attractor quintessential inflation against CMB and low-redshift data

Due to universality and attractor properties, α-attractor quintessential inflation establishes direct relations between inflationary observables such as the scalar tilt n s and the tensor-to-scalar ratio r, and late-time dark energy equation of state parameters w 0 and w a . In this work, we examine three different physically motivated regimes, considering complete freedom in the parameter α, models inspired by supergravity where α takes on values up to α = 7/3, and Starobinsky inflation (α = 1). We investigate the consistency and constraints imposed by Cosmic Microwave Background measurements from the Planck satellite, B-mode polarization data from the BICEP/Keck collaboration, and low-redshift observations. Additionally, we consider small-scale CMB measurements released by the Atacama Cosmology Telescope, which give results approaching the Harrison– Zel’dovich spectrum (n s ≈ 1). Here α-attractors lead to an improved fit over $\Lambda$CDM. For the large-scale CMB measurements, α ≳ 2 models can provide equally good fits as $\Lambda$CDM

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Diffuse X-ray scattering from polished silicon: application of the distorted wave Born approximation

Measured diffuse X-ray scattering data for a `smooth' as well as for a `rough' silicon sample were fit to theoretical expressions within the distorted wave Born approximation (DWBA). Data for the power spectral density (PSD) for both samples were also obtained by means of atomic force microscopy and optical interferometry. The Fourier transforms of trial correlation functions were fit to the PSD data and then applied to the DWBA formalism. The net correlation functions needed to fit the PSD data for each sample comprised the sum of two terms with different cutoff lengths and different self-affine fractal exponents. At zero distance these correlation functions added up to yield net values of σ 2 = (2) 2 and (71) 2 Å 2 for the smooth and rough samples, respectively. X-ray scattering data were obtained at beamline 1-BM of the Advanced Photon Source. Data and fits at values of q z = 0.05 and 0.10 Å −1 for the smooth sample are reported. Good fits for the smooth sample were obtained at both q z values simultaneously, that is, identical fitting parameters were applied at both values of q z . The smooth sample also exhibited weak Yoneda wings and a clear distinction between the strong specular scattering and the weak diffuse scattering. Data for the rough sample were qualitatively different and exhibited very weak scattering at the specular condition in contrast to extremely large Yoneda wings. Fits for the rough sample are reported for q z = 0.04, 0.05, and 0.06 Å −1 . Although the large Yoneda wings could be fit quite well in both position and amplitude, scattering near the specular condition could not be equally well fit by applying the same fitting parameters at all values of q z . Albeit imperfect, best-fitting results at the specular condition were obtained by invoking only diffuse scattering, that is, without including a separate theoretical expression for specular scattering.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2020

Observation of T = 3/2 isobaric analog states in 9 Be using p + 8 Li resonance scattering

Background: Resonance scattering has been extensively used to study the structure of exotic, neutron-deficient nuclei. Extension of the resonance scattering technique to neutron-rich nuclei was suggested more than 20 years ago. This development is based on the isospin conservation law. In spite of broad field of the application, it has never gained a wide-spread acceptance. Purpose: To benchmark the experimental approach to study the structure of exotic neutron-rich nuclei through resonance scattering on a proton target. Method: In this work, the excitation function for p + 8 Li resonance scattering is measured using a thick target by recording coincidence between light and heavy recoils, populating T = 3/2 isobaric analog states (IAS) in 9 Be. Results: A good fit of the 8 Li(p, p) 8 Li resonance elastic scattering excitation function was obtained using previously tentatively known 5/2 – T = 3/2 state at 18.65 MeV in 9 Be and a new broad T = 3/2 s-wave state—the 5/2 + at 18.5 MeV. These results fit the expected isomirror properties for the T = 3/2 A = 9 isoquartet. Conclusions: Our analysis confirmed isospin as a good quantum number for the investigated highly excited T = 3/2 states and demonstrated that studying the structure of neutron-rich exotic nuclei through IAS is a promising approach.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2024

Limits on non-relativistic matter during Big-bang nucleosynthesis

Big-bang nucleosynthesis (BBN) probes the cosmic mass-energy density at temperatures ~10 MeV to ~100 keV. Here, we consider the effect of a cosmic matter-like species that is non-relativistic and pressureless during BBN. Such a component must decay; doing so during BBN can alter the baryon-to-photon ratio, η, and the effective number of neutrino species. We use light element abundances and the cosmic microwave background (CMB) constraints on η and N ν to place constraints on such a matter component. We find that electromagnetic decays heat the photons relative to neutrinos, and thus dilute the effective number of relativistic species to N eff < 3 for the case of three Standard Model neutrino species. Intriguingly, likelihood results based on Planck CMB data alone find N ν = 2.800 ± 0.294, and when combined with standard BBN and the observations of D and 4 He give N ν = 2.898 ± 0.141. While both results are consistent with the Standard Model, we find that a nonzero abundance of electromagnetically decaying matter gives a better fit to these results. Our best-fit results are for a matter species that decays entirely electromagnetically with a lifetime τ X = 0.89 sec and pre-decay density that is a fraction ξ = (ρ X /ρ rad |10 MeV = 0.0026 of the radiation energy density at 10 MeV; similarly good fits are found over a range where ξτ X 1/2 is constant. On the other hand, decaying matter often spoils the BBN+CMB concordance, and we present limits in the (τ X ,ξ) plane for both electromagnetic and invisible decays. For dark (invisible) decays, standard BBN (i.e. ξ = 0) supplies the best fit. We end with a brief discussion of the impact of future measurements including CMB-S4.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

The DESI one-per cent survey: exploring the halo occupation distribution of luminous red galaxies and quasi-stellar objects with AbacusSummit

ABSTRACT We present the first comprehensive halo occupation distribution (HOD) analysis of the Dark Energy Spectroscopic Instrument (DESI) One-Percent Survey luminous red galaxy (LRG) and Quasi Stellar Object (QSO) samples. We constrain the HOD of each sample and test possible HOD extensions by fitting the redshift-space galaxy 2-point correlation functions in 0.15 < r < 32 h−1 Mpc in a set of fiducial redshift bins. We use AbacusSummit cubic boxes at Planck 2018 cosmology as model templates and forward model galaxy clustering with the AbacusHOD package. We achieve good fits with a standard HOD model with velocity bias, and we find no evidence for galaxy assembly bias or satellite profile modulation at the current level of statistical uncertainty. For LRGs in 0.4 < z < 0.6, we infer a satellite fraction of $f_\mathrm{sat} = 11\pm 1~{y{\ \mathrm{per\,cent}}}$, a mean halo mass of $\log _{10}\overline{M}_h/M_\odot =13.40^{+0.02}_{-0.02}$, and a linear bias of $b_\mathrm{lin} = 1.93_{-0.04}^{+0.06}$. For LRGs in 0.6 < z < 0.8, we find $f_\mathrm{sat}=14\pm 1~{{\ \mathrm{per\,cent}}}$, $\log _{10}\overline{M}_h/M_\odot =13.24^{+0.02}_{-0.02}$, and $b_\mathrm{lin}=2.08_{-0.03}^{+0.03}$. For QSOs, we infer $f_\mathrm{sat}=3^{+8}_{-2}\mathrm{per\,cent}$, $\log _{10}\overline{M}_h/M_\odot = 12.65^{+0.09}_{-0.04}$, and $b_\mathrm{lin} = 2.63_{-0.26}^{+0.37}$ in redshift range 0.8 < z < 2.1. Using these fits, we generate a large suite of high fidelity galaxy mocks, forming the basis of systematic tests for DESI Y1 cosmological analyses. We also study the redshift-evolution of the DESI LRG sample from z = 0.4 up to z = 1.1, revealling significant and interesting trends in mean halo mass, linear bias, and satellite fraction.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2020

The Atacama Cosmology Telescope: a measurement of the Cosmic Microwave Background power spectra at 98 and 150 GHz

We present the temperature and polarization angular power spectra of the CMB measured by the Atacama Cosmology Telescope (ACT) from 5400 deg2 of the 2013–2016 survey, which covers >15000 deg2 at 98 and 150 GHz. For this analysis we adopt a blinding strategy to help avoid confirmation bias and, related to this, show numerous checks for systematic error done before unblinding. Using the likelihood for the cosmological analysis we constrain secondary sources of anisotropy and foreground emission, and derive a “CMB-only” spectrum that extends to ℓ=4000. At large angular scales, foreground emission at 150 GHz is ∼1% of TT and EE within our selected regions and consistent with that found by Planck. Using the same likelihood, we obtain the cosmological parameters for ΛCDM for the ACT data alone with a prior on the optical depth of τ=0.065±0.015. ΛCDM is a good fit. The best-fit model has a reduced χ2 of 1.07 (PTE=0.07) with H0=67.9±1.5 km/s/Mpc. We show that the lensing BB signal is consistent with ΛCDM and limit the celestial EB polarization angle to ψP =−0.07ˆ±0.09ˆ. We directly cross correlate ACT with Planck and observe generally good agreement but with some discrepancies in TE. All data on which this analysis is based will be publicly released.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Single Gaussian process method for arbitrary tokamak regimes with a statistical analysis

Abstract Gaussian process regression is a Bayesian method for inferring profiles based on input data. The technique is increasing in popularity in the fusion community due to its many advantages over traditional fitting techniques including intrinsic uncertainty quantification and robustness to over-fitting. This work investigates the use of a new method, the change-point method, for handling the varying length scales found in different tokamak regimes. The use of the Student’s t-distribution for the Bayesian likelihood probability is also investigated and shown to be advantageous in providing good fits in profiles with many outliers. To compare different methods, synthetic data generated from analytic profiles is used to create a database enabling a quantitative statistical comparison of which methods perform the best. Using a full Bayesian approach with the change-point method, Matérn kernel for the prior probability, and Student’s t-distribution for the likelihood is shown to give the best results.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2025

Challenges of standard halo models in constraining galaxy properties from cosmic infrared background anisotropies

The halo model, combined with halo occupation distribution (HOD) prescriptions, is widely used to interpret cosmic infrared background (CIB) anisotropies and extract physical information about star-forming galaxies and their connection to large-scale structures. Recent CIB-specific implementations of the halo model have adopted more physical parameterizations. However, the extent to which these models can reliably recover meaningful physical parameters remains uncertain. We assessed whether the current parameterization of CIB halo models is sufficient to recover astrophysical quantities, such as star formation efficiency, η(M h , z), and halo mass at which the peak of star formation efficiency occurs, M max , when fit to mock data. We also assessed whether discrepancies arise from assumptions about galaxy emission (the HOD ingredients) or from more fundamental components in the halo model, such as bias and matter clustering. We fit the M21 CIB HOD model, implemented within the halo model framework, to mock CIB power spectra and star formation rate density (SFRD) data generated from the SIDES-Uchuu simulation, and compared the best-fit parameters to the known simulation inputs. We then repeated the analysis using a simplified version of the simulation (SSU), explicitly designed to match the HOD assumptions. A detailed comparison of model and simulation outputs was carried out to trace the origin of observed discrepancies. While the M21 HOD model provides a good fit to the mock data, it failed to recover the intrinsic parameters accurately, particularly the halo mass at which star formation efficiency peaks. This mismatch persists even when fitting data generated with the same model assumptions. We find strong agreement (within 5%) in the emission-related components (SFRD, emissivity), but observe a scale- and redshift-dependent offset exceeding 20% in the two-halo term of the CIB power spectrum. This likely arises from limitations in the treatment of halo bias and matter clustering within the linear approximation. Additionally, incorporating scatter in the SFR–halo mass relation and the spectral energy distribution (SED) templates significantly affects the shot noise (∼50%), but has only a modest impact (less than 10%) on the clustered component. These results suggest that recovering physical parameters from CIB clustering requires improvements to the cosmological ingredients of the halo model framework, such as adopting scale-dependent halo bias and nonlinear matter power spectra in addition to careful modeling of emission physics.

cosmic background radiation↗
DOE OSTI2024

Physics-informed Estimation of the Covariance Matrix for Various Neutron Spectra

A method for estimating covariance matrices which capture the uncertainties in calculated reactor spectra has been developed. This method is based on perturbing the parameters of a physics-based analytic model fitted to a calculated spectrum. The covariance of the perturbed analytic spectra imposes energy-dependent correlations due to the physics of the neutron processes in the reactor, i.e., a fission component, a 1/E down-scatting component, and a thermal Maxwellian component. An analytic model is developed which is shown to produce good fits to several reactor environments. The covariance matrices produced via this method are then used as the prior spectrum in STAYSL least squares spectrum adjustment where it is combined with integral metrics, such as activation measurements, to produce a high-fidelity neutron spectrum characterization. It was concluded that the methodology showed agreeable results for the ACRR free-field spectrum adjustment in STAYSL resulting in a 𝜒 2 value of 2.21 (per degree of freedom), but further work is needed to describe scattering and interface regions.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2020

The Atacama Cosmology Telescope: DR4 maps and cosmological parameters

In this paper, we present new arcminute-resolution maps of the Cosmic Microwave Background temperature and polarization anisotropy from the Atacama Cosmology Telescope, using data taken from 2013–2016 at 98 and 150 GHz. The maps cover more than 17,000 deg 2 , the deepest 600 deg 2 with noise levels below 10µK-arcmin. We use the power spectrum derived from almost 6,000 deg 2 of these maps to constrain cosmology. The ACT data enable a measurement of the angular scale of features in both the divergence-like polarization and the temperature anisotropy, tracing both the velocity and density at last-scattering. From these one can derive the distance to the last-scattering surface and thus infer the local expansion rate, H 0 . By combining ACT data with large-scale information from WMAP we measure H 0 = 67.6±1.1 km/s/Mpc, at 68% confidence, in excellent agreement with the independently-measured Planck satellite estimate (from ACT alone we find H 0 = 67.9 ± 1.5 km/s/Mpc). The ΛCDM model provides a good fit to the ACT data, and we find no evidence for deviations: both the spatial curvature, and the departure from the standard lensing signal in the spectrum, are zero to within 1σ; the number of relativistic species, the primordial Helium fraction, and the running of the spectral index are consistent with ΛCDM predictions to within 1.5–2.2σ. We compare ACT, WMAP, and Planck at the parameter level and find good consistency; we investigate how the constraints on the correlated spectral index and baryon density parameters readjust when adding CMB large-scale information that ACT does not measure. The DR4 products presented here will be publicly released on the NASA Legacy Archive for Microwave Background Data Analysis.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2025

Probability Density Function for the spatial and intensity distribution of neutron-induced defects in Silicon

The ability to model semiconductor device degradation under neutron irradiation depends upon having a robust modeling capability for the neutron-induced collision cascades as well as a means to analytically fit the resulting probability distributions of defect production and ionizing energy deposition for purposes of extrapolation to low-probability, high-consequence scenarios. In this paper, the widely-utilized binary collision approximation codes MARLOWE and SRIM are deployed in conjunction with a critical examination of their parameterizations as benchmarked against higher-fidelity molecular dynamics simulations. A simple 3-parameter form described by the Generalized Logistic Distribution is shown to be a good fit to Frenkel pair and ionization intensity distributions in bulk silicon. The BCA codes are then applied to simulate cascades in 5 nm layers of a representative gate-all-around nanosheet transistor, where joint probability distributions of threshold levels of damage to multiple layers are evaluated.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Two years of optical and NIR observations of the superluminous supernova UID 30901 discovered by the UltraVISTA SN survey

ABSTRACT We present deep optical and near-infrared photometry of UID 30901, a superluminous supernova (SLSN) discovered during the UltraVISTA survey. The observations were obtained with VIRCAM (YJHKs) mounted on the VISTA telescope, DECam (griz) on the Blanco telescope, and SUBARU Hyper Suprime-Cam (HSC; grizy). These multiband observations comprise +700 d making UID 30901 one of the best photometrically followed SLSNe to date. The host galaxy of UID 30901 is detected in a deep HST F814W image with an AB magnitude of 27.3 ± 0.2. While no spectra exist for the SN or its host galaxy, we perform our analysis assuming z = 0.37, based on the photometric redshift of a possible host galaxy found at a projected distance of 7 kpc. Fitting a blackbody to the observations, the radius, temperature, and bolometric light curve are computed. We find a maximum bolometric luminosity of 5.4 ± 0.34 × 1043 erg s−1. A flattening in the light curve beyond 600 d is observed and several possible causes are discussed. We find the observations to clearly favour an SLSN type I, and plausible power sources such as the radioactive decay of 56Ni and the spin-down of a magnetar are compared to the data. We find that the magnetar model yields a good fit to the observations with the following parameters: a magnetic field B = 1.4 ± 0.3 × 1014 G, spin period of P = 6.0 ± 0.1 ms, and ejecta mass $M_{\mathrm{ ej}} = 11.9^{+4.8}_{-6.4} \,\mathrm{ M}_{\odot }$.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2020

Calibration of hyperelastic constitutive models: the role of boundary conditions, search algorithms, and experimental variability

The calibration of hyperelastic constitutive models of soft tissue and tissue surrogates is often treated as an exercise in curve-fitting to the average experimental response, and many of the complicating factors such as experimental boundary conditions and data variability are ignored. Here, we focus on three questions that arise in this area: the ramifications of ignoring the experimental boundary conditions, the use of local optimizers, and the role of data variability. Using data from a uniaxial extension experiment on a tissue surrogate, we study how these three factors affect the calibration of isotropic hyperelastic constitutive models. Our results show that even with the simplest of constitutive models, it is necessary to look beyond a "good fit" to the average.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Modeling HIV-1 Within-Host Dynamics After Passive Infusion of the Broadly Neutralizing Antibody VRC01

VRC01 is a broadly neutralizing antibody that targets the CD4 binding site of HIV-1 gp120. Passive administration of VRC01 in humans has assessed the safety and the effect on plasma viremia of this monoclonal antibody (mAb) in a phase 1 clinical trial. After VRC01 infusion, the plasma viral load in most of the participants was reduced but had particular dynamics not observed during antiretroviral therapy. In this paper, we introduce different mathematical models to explain the observed dynamics and fit them to the plasma viral load data. Based on the fitting results we argue that a model containing reversible Ab binding to virions and clearance of virus-VRC01 complexes by a two-step process that includes (1) saturable capture followed by (2) internalization/degradation by phagocytes, best explains the data. This model predicts that VRC01 may enhance the clearance of Ab-virus complexes, explaining the initial viral decay observed immediately after antibody infusion in some participants. Because Ab-virus complexes are assumed to be unable to infect cells, i.e., contain neutralized virus, the model predicts a longer-term viral decay consistent with that observed in the VRC01 treated participants. By assuming a homogeneous viral population sensitive to VRC01, the model provides good fits to all of the participant data. However, the fits are improved by assuming that there were two populations of virus, one more susceptible to antibody-mediated neutralization than the other.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Modelling the stellar halo with RR-Lyrae stars

ABSTRACT A seven-parameter distribution function (DF) is fitted to $20\, 000$ RR-Lyrae stars for which only astrometric data are available. The observational data are predicted by the DF in conjunction with the gravitational potential of a self-consistent model Galaxy defined by DFs for the dark halo, the bulge, and a four-component disc. Tests of the technique developed to deal with missing line-of-sight velocities show that adding such velocities tightens constraints on the DF only slightly. The recovered model of the RR-Lyrae population confirms that the population is flattened and has a strongly radially biased velocity distribution. At large radii, its density profile tends to ρ ∼ r−4.5 but no power law provides a good fit inside the solar sphere. The model is shown to provide an excellent fit to the data for stars brighter than r = 16.5 but at certain longitudes it predicts too few faint stars at Galactocentric radii $\sim 20\, \mathrm{kpc}$, possibly signalling that the halo is not axisymmetric. The DF is used to predict the velocity distribution of BHB stars for which space velocities are available. The z components are predicted successfully but too much anisotropy in the vRvϕ plane is expected.

Li, Chengdong↗
DOE OSTI2022

Relative Controls of Vapor Pressure Deficit and Soil Water Stress on Canopy Conductance in Global Simulations by an Earth System Model

Climate change impacts drought and vapor pressure deficit (VPD) that critically influence land-atmosphere water vapor and CO 2 exchange. There is a lack of model-based global evaluation of the relative roles of the two types of water stresses in limiting canopy conductance (G c ). Using the Energy Exascale Earth System Model (E3SM), we conducted four land-atmosphere coupled simulations for the historical period with and without a plant hydraulics scheme (PHS) and under present-day and quadrupling of CO 2 concentrations. By fitting an empirical model of G c to the E3SM simulations, we evaluated the sensitivity of G c to water stresses caused by VPD and soil water (SW) in the growing season. The empirical model is based on a multiplicative algorithm that adjusts G c according to VPD, soil water saturation, and their interaction. Our results show positive trends of VPD over land during the historical period, with large differences between simulations with and without PHS. Despite the differences, a majority of land pixels that experience both SW and VPD stresses are in semi-arid and tropical savanna ecosystems. G c is more sensitive to VPD than SW in less than 20% of land pixels where the empirical model provides a good fit for the simulated G c . Ignoring SW and VPD interaction may underestimate the relative role of VPD on G c by as much as 28%. Including PHS accentuates SW stress limitation on G c by simulating wetter soil and higher ET that lower VPD. Elevated CO 2 reduces G c by increasing water use efficiency and ameliorates SW stress on G c .

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Kinetic Deep Learning v0.1

Here, we present a method that uses protein levels to predict times series of metabolite concentrations. Understanding this type of pathway dynamics is important in order to predict the behavior of the pathway and, more pragmatically, to be able to design biological systems (such as strains bioengineered to produce chemical products) reliably. Typically, for this purpose, kinetic models consisting of differential equations based on the Michaelis-Menten dynamics have been used in the past. However, these methods can rarely produce good fits to measured data time series. Possibly, this happens because the kinetic constants are unknown or are different from the ones measured in vivo, or perhaps because Michaelis-Menten dynamics is not a satisfactory description. In order to improve the predictive nature of these kinetic models we have eliminated the Michaelis-Menten description of pathway dynamics and we have substituted it by algorithms that automatically learn these dynamics from previously obtained metabolomics and proteomics data using machine learning approaches. Specifically, kinetic deep learning uses deep learning to map proteomics time series to metabolite concentration time series, instead of learning the first metabolite derivative and integrating in (as in the first version of kinetic learning). This approach is shown to provide good to excellent results with a data set specifically collected for this purpose.

Garcia Martin, Hector [Joint BioEnergy Institute (↗