Search NASA⌕ Search

SEARCH · Search NASA

Results for “Cluster Computing”

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 1,063 records · Page 59
All sourcesNASA NTRSDOE OSTI
DOE OSTI2014

IMG-ABC: An Atlas of Biosynthetic Gene Clusters to Fuel the Discovery of Novel Secondary Metabolites

In the discovery of secondary metabolites (SMs), large-scale analysis of sequence data is a promising exploration path that remains largely underutilized due to the lack of relevant computational resources. We present IMG-ABC (https://img.jgi.doe.gov/abc/) -- An Atlas of Biosynthetic gene Clusters within the Integrated Microbial Genomes (IMG) system1. IMG-ABC is a rich repository of both validated and predicted biosynthetic clusters (BCs) in cultured isolates, single-cells and metagenomes linked with the SM chemicals they produce and enhanced with focused analysis tools within IMG. The underlying scalable framework enables traversal of phylogenetic dark matter and chemical structure space -- serving as a doorway to a new era in the discovery of novel molecules.

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2025

Cluster-Graph Fingerprinting: A Framework for Quantitative Analysis of Machine-Learned Interatomic Model Training and Simulation Data

Machine-learned interatomic models represent a significant advancement in simulation methods, extending the predictive ability of first-principles methods to previously inaccessible length and time scales. However, the data-driven nature of these models can lead to difficult-to-detect errors that can compromise prediction accuracy. To address this challenge, we introduce a novel fingerprinting approach based on the Chebyshev Interaction Model for Efficient Simulation (ChIMES) ML-IAM graph-based descriptor. Our strategy enables efficient and statistically rigorous analysis of system configurations used in ML-IAM training and those generated by their application, e.g., in molecular dynamics simulations. We demonstrate that these fingerprints can effectively assess novelty of a configuration relative to an existing data set and determine dissimilarity among individual configurations, which are two key tasks in workflows for active learning-based ML-IAM training, data set curation, and on-the-fly uncertainty quantification.

36 MATERIALS SCIENCE↗
DOE OSTI2016

The chemical homogeneity of open clusters

Determining the level of chemical homogeneity in open clusters is of fundamental importance in the study of the evolution of star-forming clouds and that of the Galactic disk. Yet limiting the initial abundance spread in clusters has been hampered by difficulties in obtaining consistent spectroscopic abundances for different stellar types. Without reference to any specific model of stellar photospheres, a model for a homogeneous cluster is that it forms a one-dimensional sequence, with any differences between members due to variations in stellar mass and observational uncertainties. I present a novel method for investigating the abundance spread in open clusters that tests this one-dimensional hypothesis at the level of observed stellar spectra, rather than constraining homogeneity using derived abundances as is traditionally done. Using high-resolution APOGEE spectra for 49 giants in M67, NGC 6819, and NGC 2420 I demonstrate that these spectra form one-dimensional sequences for each cluster. With detailed forward modeling of the spectra and Approximate Bayesian Computation, I derive strong limits on the initial abundance spread of 15 elements: <0.01 (0.02) dex for C and Fe, ≲0.015 (0.03) dex for N, O, Mg, Si, and Ni, ≲0.02 (0.03) dex for Al, Ca, and Mn, and ≲0.03 (0.05) dex for Na, S, K, Ti, and V (at 68% and 95% confidence, respectively). The strong limits on C and O imply that no pollution by massive core-collapse supernovae occurred during star formation in open clusters, which, thus, need to form within ≲6 Myr. Further development of this and related techniques will bring the power of differential abundances to stars other than solar twins in large spectroscopic surveys and will help unravel the history of star formation and chemical enrichment in the Milky Way through chemical tagging.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2015

IMG-ABC. A knowledge base to fuel discovery of biosynthetic gene clusters and novel secondary metabolites

In the discovery of secondary metabolites, analysis of sequence data is a promising exploration path that remains largely underutilized due to the lack of computational platforms that enable such a systematic approach on a large scale. In this work, we present IMG-ABC (https://img.jgi.doe.gov/abc), an atlas of biosynthetic gene clusters within the Integrated Microbial Genomes (IMG) system, which is aimed at harnessing the power of “big” genomic data for discovering small molecules. IMG-ABC relies on IMG’s comprehensive integrated structural and functional genomic data for the analysis of biosynthetic gene clusters (BCs) and associated secondary metabolites (SMs). SMs and BCs serve as the two main classes of objects in IMG-ABC, each with a rich collection of attributes. A unique feature of IMG-ABC is the incorporation of both experimentally validated and computationally predicted BCs in genomes as well as metagenomes, thus identifying BCs in uncultured populations and rare taxa. We demonstrate the strength of IMG-ABC’s focused integrated analysis tools in enabling the exploration of microbial secondary metabolism on a global scale, through the discovery of phenazine-producing clusters for the first time in lphaproteobacteria . IMG-ABC strives to fill the long-existent void of resources for computational exploration of the secondary metabolism universe; its underlying scalable framework enables traversal of uncovered phylogenetic and chemical structure space, serving as a doorway to a new era in the discovery of novel molecules. IMG-ABC is the largest publicly available database of predicted and experimental biosynthetic gene clusters and the secondary metabolites they produce. The system also includes powerful search and analysis tools that are integrated with IMG’s extensive genomic/metagenomic data and analysis tool kits. As new research on biosynthetic gene clusters and secondary metabolites is published and more genomes are sequenced, IMG-ABC will continue to expand, with the goal of becoming an essential component of any bioinformatic exploration of the secondary metabolism world.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2016

BLACK HOLE MERGERS AND BLUE STRAGGLERS FROM HIERARCHICAL TRIPLES FORMED IN GLOBULAR CLUSTERS

Hierarchical triple-star systems are expected to form frequently via close binary–binary encounters in the dense cores of globular clusters (GCs). In a sufficiently inclined triple, gravitational interactions between the inner and outer binary can cause large-amplitude oscillations in the eccentricity of the inner orbit (“Lidov–Kozai (LK) cycles”), which can lead to a collision and merger of the two inner components. In this paper we use Monte Carlo models of dense star clusters to identify all triple systems formed dynamically and we compute their evolution using a highly accurate three-body integrator which incorporates relativistic and tidal effects. We find that a large fraction of these triples evolve through a non-secular dynamical phase which can drive the inner binary to higher eccentricities than predicted by the standard secular perturbation theory (even including octupole-order terms). We place constraints on the importance of LK-induced mergers for producing: (i) gravitational wave sources detectable by Advanced LIGO (aLIGO), for triples with an inner pair of stellar black holes (BHs); and (ii) blue straggler stars, for triples with main-sequence-star components. We find a realistic aLIGO detection rate of BH mergers due to the LK mechanism of ∼1 yr{sup −1}, with about 20% of these having a finite eccentricity when they first chirp into the aLIGO frequency band. While rare, these events are likely to dominate among eccentric compact object inspirals that are potentially detectable by aLIGO. For blue stragglers, we find that the LK mechanism can contribute up to ∼10% of their total numbers in GCs.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2014

Orbital-optimized MP2.5 and its analytic gradients: Approaching CCSD(T) quality for noncovalent interactions

Orbital-optimized MP2.5 [or simply “optimized MP2.5,” OMP2.5, for short] and its analytic energy gradients are presented. The cost of the presented method is as much as that of coupled-cluster singles and doubles (CCSD) [O(N{sup 6}) scaling] for energy computations. However, for analytic gradient computations the OMP2.5 method is only half as expensive as CCSD because there is no need to solve λ{sub 2}-amplitude equations for OMP2.5. The performance of the OMP2.5 method is compared with that of the standard second-order Møller–Plesset perturbation theory (MP2), MP2.5, CCSD, and coupled-cluster singles and doubles with perturbative triples (CCSD(T)) methods for equilibrium geometries, hydrogen transfer reactions between radicals, and noncovalent interactions. For bond lengths of both closed and open-shell molecules, the OMP2.5 method improves upon MP2.5 and CCSD by 38%–43% and 31%–28%, respectively, with Dunning's cc-pCVQZ basis set. For complete basis set (CBS) predictions of hydrogen transfer reaction energies, the OMP2.5 method exhibits a substantially better performance than MP2.5, providing a mean absolute error of 1.1 kcal mol{sup −1}, which is more than 10 times lower than that of MP2.5 (11.8 kcal mol{sup −1}), and comparing to MP2 (14.6 kcal mol{sup −1}) there is a more than 12-fold reduction in errors. For noncovalent interaction energies (at CBS limits), the OMP2.5 method maintains the very good performance of MP2.5 for closed-shell systems, and for open-shell systems it significantly outperforms MP2.5 and CCSD, and approaches CCSD(T) quality. The MP2.5 errors decrease by a factor of 5 when the optimized orbitals are used for open-shell noncovalent interactions, and comparing to CCSD there is a more than 3-fold reduction in errors. Overall, the present application results indicate that the OMP2.5 method is very promising for open-shell noncovalent interactions and other chemical systems with difficult electronic structures.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

P SI4 1.4: Open-source software for high-throughput quantum chemistry

PSI4 is a free and open-source ab initio electronic structure program providing implementations of Hartree–Fock, density functional theory, many-body perturbation theory, configuration interaction, density cumulant theory, symmetry-adapted perturbation theory, and coupled-cluster theory. Most of the methods are quite efficient, thanks to density fitting and multi-core parallelism. The program is a hybrid of C++ and Python, and calculations may be run with very simple text files or using the Python API, facilitating post-processing and complex workflows; method developers also have access to most of PSI4 's core functionalities via Python. Job specification may be passed using The Molecular Sciences Software Institute (MolSSI) QCSCHEMA data format, facilitating interoperability. We announce a rewrite of our top-level computation driver, and concomitant adoption of the MolSSI QCARCHIVE INFRASTRUCTURE project, makes the latest version of PSI4 well suited to distributed computation of large numbers of independent tasks. The project has fostered the development of independent software components that may be reused in other quantum chemistry programs.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Discovering Hidden Geothermal Signatures using Unsupervised Machine Learning

Discovering hidden geothermal resources is a very challenging task. It requires the mining of large datasets, including various diverse data attributes representing subsurface hydrogeological and geothermal conditions. The commonly used Play Fairway Analysis (PFA) typically relies on subject-matter expertise to analyze site or regional data to estimate geothermal conditions and prospectivity. Here, we demonstrate an alternative approach based on machine learning (ML) to process a geothermal dataset of Southwest New Mexico (SWNM). The study region includes low- and medium-temperature hydrothermal systems. However, most of these systems are poorly characterized because of insufficient existing data and limited past explorative studies. This study aims to discover hidden patterns and relationships in the SWNM geothermal dataset to better understand regional hydrothermal conditions. This is achieved by applying an unsupervised machine learning algorithm based on non-negative matrix factorization coupled with customized k-means clustering (NMFk). NMFk can automatically identify (1) hidden (latent) signatures characterizing datasets, (2) the optimal number of these signatures, (3) dominant data attributes associated with each signature, and (4) spatial distribution of the extracted signatures. Here, NMFk is applied to analyze 18 geological, geophysical, hydrogeological, geothermal attributes at 44 locations in SWNM. NMFk successfully finds data patterns and identifies the spatial associations of hydrothermal signatures with the four physiographic provinces in SWNM (Colorado Plateau, Volcanic Field, Basin and Range, and the Rio Grande rift). The algorithm identified up to 5 hydrothermal signatures in the SWNM datasets that differentiate between low- and medium-temperature hydrothermal systems in different provinces. Also, the algorithm identifies two medium-temperature hydrothermal systems in SWNM that require further exploration for geothermal resource development. Based on our analyses, 12 of the attributes are important to identify medium-temperature hydrothermal systems, and the remaining six attributes are critical to characterize low-temperature hydrothermal systems. Based on the obtained results, we identify potential physiographic provinces for further exploration to characterize them as geothermal resources. The resulting NMFk model can be applied to predict geothermal conditions and their uncertainties at new SWNM locations based on limited data from unexplored areas.

58 GEOSCIENCES↗
DOE OSTI2014

Automated genomic context analysis and experimental validation platform for discovery of prokaryote transcriptional regulator functions

The clustering of genes in a pathway and the co-location of functionally related genes is widely recognized in prokaryotes. We used these characteristics to predict the metabolic involvement for a Transcriptional Regulator (TR) of unknown function, identified and confirmed its biological activity. A software tool that identifies the genes encoded within a defined genomic neighborhood for the subject TR and its homologs was developed. The output lists of genes in the genetic neighborhoods, their annotated functions, the reactants/products, and identifies the metabolic pathway in which the encoded-proteins function. When a set of TRs of known function was analyzed, we observed that their homologs frequently had conserved genomic neighborhoods that co-located the metabolically related genes regulated by the subject TR. We postulate that TR effectors are metabolites in the identified pathways; indeed the known effectors were present. We analyzed Bxe_B3018 from Burkholderia xenovorans , a TR of unknown function and predicted that this TR was related to the glycine, threonine and serine degradation. We tested the binding of metabolites in these pathways and for those that bound, their ability to modulate TR binding to its specific DNA operator sequence. Using rtPCR, we confirmed that methylglyoxal was an effector of Bxe_3018.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2016

Density Functional Investigation of the Inclusion of Gold Clusters on a CH 3 S Self-Assembled Lattice on Au(111)

We employ first-principles density functional theoretical calculations to address the inclusion of gold (Au) clusters in a well-packed CH 3 S self-assembled lattice. We compute CH 3 S adsorption energies to quantify the energetic stability of the self-assembly and gold adsorption and dissolution energies to characterize the structural stability of a series of Au clusters adsorbed at the SAM-Au interface. Our results indicate that the inclusion of Au clusters with less than four Au atoms in the SAM-Au interface enhances the binding of CH 3 S species. In contrast, larger Au clusters destabilize the self-assembly. We attribute this effect to the low-coordinated gold atoms in the cluster. For small clusters, these low-coordinated sites have significantly different electronic properties compared to larger islands, which makes the binding with the self-assembly energetically more favorable. Our results further indicate that Au clusters in the SAM-Au interface are thermodynamically unstable and they will tend to dissolve, producing Au adatoms incorporated in the self-assembly in the form of CH 3 S-Au-SCH 3 species. This is due to the strong S-Au bond which stabilizes single Au adatoms in the self-assembly. Our results provide solid insight into the impact of adatom islands at the CH 3 S-Au interface.

Allen, Darnel J.↗
DOE OSTI2013

Reaction Mechanism for Direct Propylene Epoxidation by Alumina-Supported Silver Aggregates. The Role of the Particle / Support Interface

Sub-nanometer Ag aggregates on alumina supports have been found to be active toward direct propylene epoxidation to propylene oxide by molecular oxygen at low temperatures, with a negligible amount of carbon dioxide formation (Science 328, p. 224, 2010). In this work, we computationally and experimentally investigate the origin of the high reactivity of the sub-nanometer Ag aggregates. Computationally, we study O 2 dissociation and propylene epoxidation on unsupported Ag 19 and Ag 20 clusters, as well as alumina-supported Ag 19 . The O 2 dissociation and propylene epoxidation apparent barriers at the interface between the Ag aggregate and the alumina support are calculated to be 0.2 and 0.2~0.4 eV, respectively. These barriers are somewhat lower than those on sites away from the interface. The mechanism at the interface is similar to what was previously found for the silver trimer on alumina and can account for the high activity observed for the direct oxidation of propylene on the Ag aggregates. The barriers for oxygen dissociation on these model systems both at the interface and on the surfaces are small compared to crystalline surfaces, indicating that availability of oxygen will not be a rate limiting step for the aggregates, as in the case of the crystalline surfaces. Experimentally, we investigate Ultrananocrystalline Diamond (UNCD)-supported silver aggregates under reactive conditions of propylene partial oxidation. The UNCD-supported Ag clusters are found to be not measurably active toward propylene oxidation, in contrast to the alumina supported Ag clusters. This suggests that the lack of metal-oxide interfacial sites of the Ag-UNCD catalyst, limits the epoxidation catalytic activity. This combined computational and experimental study shows the importance of the metal-oxide interface as well as the non-crystalline nature of the alumina-supported sub-nanometer Ag aggregate catalysts for propylene epoxidation.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Effective many-body interactions in reduced-dimensionality spaces through neural network models

Accurately describing properties of challenging problems in physical sciences often requires complex mathematical models that are unmanageable to tackle head on. Therefore, developing reduced-dimensionality representations that encapsulate complex correlation effects in many-body systems is crucial to advance the understanding of these complicated problems. However, a numerical evaluation of these predictive models can still be associated with a significant computational overhead. To address this challenge, in this paper we discuss a combined framework that integrates recent advances in the development of active-space representations of coupled cluster (CC) downfolded Hamiltonians with neural network approaches. The primary objective of this effort is to train neural networks to eliminate the computationally expensive steps required for evaluating hundreds or thousands of Hugenholtz diagrams, which correspond to multidimensional tensor contractions necessary for evaluating a many-body form of downfolded effective Hamiltonians. Using small molecular systems (the H 2 O and HF molecules) as examples, we demonstrate that training neural networks employing effective Hamiltonians for a few nuclear geometries of molecules can accurately interpolate or extrapolate their forms to other geometrical configurations characterized by different intensities of correlation effects. We also discuss differences between effective interactions that define CC downfolded Hamiltonians with those of bare Hamiltonians defined by Coulomb interactions in the active spaces. Published by the American Physical Society 2024

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Motility-induced phase separation and frustration in active matter swarmalators

Here we introduce a two dimensional system of active matter swarmalators composed of elastically interacting run-and-tumble active disks with an internal parameter Φ i . The disks experience an additional attractive or repulsive force with neighboring disks depending upon their relative difference in Φ i , making them similar to swarmalators used in robotic systems. In the absence of the internal parameter, the system forms a motility-induced phase separated (MIPS) state, but when the swarmalator interactions are present, a wide variety of other active phases appear depending upon whether the interaction is attractive or repulsive and whether the particles act to synchronize or ant-synchronize their internal parameter values. These phases include a gas-free gel regime, arrested clusters, a labyrinthine state, a regular MIPS state, a frustrated MIPS state for attractive antisynchronization, and a superlattice MIPS state for attractive synchronization.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2019

Global Sensitivity Analysis of Bulk Properties of an Atomic Nucleus

We perform a global sensitivity analysis of the binding energy and the charge radius of the nucleus 16 O to identify the most influential low-energy constants in the next-to-next-to-leading order chiral Hamiltonian with two- and three-nucleon forces. For this purpose, we develop a subspace-projected coupled-cluster method using eigenvector continuation [Frame D. et al., Phys. Rev. Lett. 121, 032501 (2018)]. With this method, we compute the binding energy and charge radius of 16 O at more than 106 different values of the 16 low-energy constants in one hour on a standard laptop computer. For relatively small subspace projections, the root-mean-square error is about 1% compared to full-space coupled-cluster results. We find that 58(1)% of the variance in energy can be apportioned to a single contact term in the 3 S 1 wave, whereas the radius depends sensitively on several low-energy constants and their higher-order correlations. The results identify the most important parameters for describing nuclear saturation and help prioritize efforts for uncertainty reduction of theoretical predictions. The achieved acceleration opens up an array of computational statistics analyses of the underlying description of the strong nuclear interaction in nuclei across the Segrè chart.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2016

Computing and Applying Atomic Regulons to Understand Gene Expression and Regulation

Understanding gene function and regulation is essential for the interpretation, prediction, and ultimate design of cell responses to changes in the environment. A multitude of technologies, abstractions, and interpretive frameworks have emerged to answer the challenges presented by genome function and regulatory network inference. Here, we propose a new approach for producing biologically meaningful clusters of coexpressed genes, called Atomic Regulons (ARs), based on expression data, gene context, and functional relationships. We demonstrate this new approach by computing ARs for Escherichia coli , which we compare with the coexpressed gene clusters predicted by two prevalent existing methods: hierarchical clustering and k-means clustering. We test the consistency of ARs predicted by all methods against expected interactions predicted by the Context Likelihood of Relatedness (CLR) mutual information based method, finding that the ARs produced by our approach show better agreement with CLR interactions. We then apply our method to compute ARs for four other genomes: Shewanella oneidensis , Pseudomonas aeruginosa , Thermus thermophilus , and Staphylococcus aureus . We compare the AR clusters from all genomes to study the similarity of coexpression among a phylogenetically diverse set of species, identifying subsystems that show remarkable similarity over wide phylogenetic distances. We also study the sensitivity of our method for computing ARs to the expression data used in the computation, showing that our new approach requires less data than competing approaches to converge to a near final configuration of ARs. We go on to use our sensitivity analysis to identify the specific experiments that lead most rapidly to the final set of ARs for E. coli . As a result, this analysis produces insights into improving the design of gene expression experiments.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2019

RELICS: Strong Lensing Analysis of MACS J0417.5–1154 and Predictions for Observing the Magnified High-redshift Universe with JWST

Strong gravitational lensing by clusters of galaxies probes the mass distribution at the core of each cluster and magnifies the universe behind it. MACS J0417.5–1154 at z = 0.443 is one of the most massive clusters known based on weak lensing, X-ray, and Sunyaev–Zel'dovich analyses. In this work we compute a strong lens model of MACS J0417 based on Hubble Space Telescope imaging observations collected, in part, by the Reionization Lensing Cluster Survey (RELICS), and recently reported spectroscopic redshifts from the MUSE instrument on the Very Large Telescope (VLT). We measure an Einstein radius of ${\theta }_{E}\simeq 36^{\prime\prime} $ at z = 9 and a mass projected within 200 kpc of ${M}_{(200\mathrm{kpc})}={1.78}_{-0.03}^{+0.01}\times {10}^{14}$ M ${}_{\odot }$. Using this model, we measure a ratio between the mass attributed to cluster-member galaxy halos and the main cluster halo of order 1:100. We assess the probability to detect magnified high-redshift galaxies in the field of this cluster, both for comparison with RELICS HST results and as a prediction for the James Webb Space Telescope (JWST) Guaranteed Time Observations upcoming for this cluster. Our lensing analysis suggests that this cluster has similar lensing strength to other clusters in the RELICS program. Our lensing analysis predicts a detection of at least a few z ~ 6–8 galaxies behind this cluster, at odds with a recent analysis that yielded no such candidates in this field. Reliable strong lensing models are important for accurately predicting the intrinsic properties of lensed galaxies. As part of the RELICS program, our strong lensing model produced with the Lenstool parametric method is publicly available through the Mikulski Archive for Space Telescopes.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2015

Visual exploration of high-dimensional data through subspace analysis and dynamic projections

Here, we introduce a novel interactive framework for visualizing and exploring high-dimensional datasets based on subspace analysis and dynamic projections. We assume the high-dimensional dataset can be represented by a mixture of low-dimensional linear subspaces with mixed dimensions, and provide a method to reliably estimate the intrinsic dimension and linear basis of each subspace extracted from the subspace clustering. Subsequently, we use these bases to define unique 2D linear projections as viewpoints from which to visualize the data. To understand the relationships among the different projections and to discover hidden patterns, we connect these projections through dynamic projections that create smooth animated transitions between pairs of projections. We introduce the view transition graph, which provides flexible navigation among these projections to facilitate an intuitive exploration. Finally, we provide detailed comparisons with related systems, and use real-world examples to demonstrate the novelty and usability of our proposed framework.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Improving Medication Regimen Recommendation for Parkinson’s Disease Using Sensor Technology

Parkinson’s disease medication treatment planning is generally based on subjective data obtained through clinical, physician-patient interactions. The Personal KinetiGraph™ (PKG) and similar wearable sensors have shown promise in enabling objective, continuous remote health monitoring for Parkinson’s patients. In this proof-of-concept study, we propose to use objective sensor data from the PKG and apply machine learning to cluster patients based on levodopa regimens and response. The resulting clusters are then used to enhance treatment planning by providing improved initial treatment estimates to supplement a physician’s initial assessment. We apply k-means clustering to a dataset of within-subject Parkinson’s medication changes—clinically assessed by the MDS-Unified Parkinson’s Disease Rating Scale-III (MDS-UPDRS-III) and the PKG sensor for movement staging. A random forest classification model was then used to predict patients’ cluster allocation based on their respective demographic information, MDS-UPDRS-III scores, and PKG time-series data. Clinically relevant clusters were partitioned by levodopa dose, medication administration frequency, and total levodopa equivalent daily dose—with the PKG providing similar symptomatic assessments to physician MDS-UPDRS-III scores. A random forest classifier trained on demographic information, MDS-UPDRS-III scores, and PKG time-series data was able to accurately classify subjects of the two most demographically similar clusters with an accuracy of 86.9%, an F1 score of 90.7%, and an AUC of 0.871. A model that relied solely on demographic information and PKG time-series data provided the next best performance with an accuracy of 83.8%, an F1 score of 88.5%, and an AUC of 0.831, hence further enabling fully remote assessments. These computational methods demonstrate the feasibility of using sensor-based data to cluster patients based on their medication responses with further potential to assist with medication recommendations.

59 BASIC BIOLOGICAL SCIENCES↗