150813 research outputs found
Sort by
Computational methods for dissecting multicellular mechanisms of complex diseases
No full textSingle-cell genomics technologies have enabled unbiased characterization of cell types and cellular states. However, the high-dimensional nature of this data necessitates computational and statistical methods to uncover the biological processes that shape it. In my thesis research, I developed three computational methods to explore genetic regulatory mechanisms underlying common diseases and the resulting multicellular patterns of dysfunction. In the first project, I developed a method called scITD to investigate how cellular processes across distinct cell types coordinate in disease contexts. scITD identifies sets of genes in one or more cell types that co-vary together across biological samples. Through the application of this tool to various immune-cell datasets, we uncovered highly reproducible gene expression patterns associated with autoimmune patient phenotypes. In the second project, I characterized technical artifacts prevalent in imaging-based spatial transcriptomics data. These artifacts arise from the misassignment of transcript molecules to incorrect cells. I further demonstrated how these artifacts confound downstream analyses, including differential expression and cell-cell interaction inference. To address this, I jointly developed a correction method that mitigates these artifacts, thereby uncovering novel biological insights in cancer datasets. In the third project, I introduced a computational method to unravel the mechanisms of genetic variants identified from genome-wide association study loci. This method tests whether these same genetic variants also underly changes to gene expression in specific cell types or states. Applying this tool to autoimmune and neurodegenerative datasets uncovered new SNP-gene-phenotype links and localized their effects to specific cell populations, helping to refine our understanding of these pathologies.Ph.D
Artificial Intelligence Enhances Air Mobility Planning
No full textLincoln Laboratory is transitioning tools to the 618th Air Operations Center to streamline global transport logistics.DAF–MIT AI Accelerato
Growth-Induced Cation Order and Magnetic Anisotropy Engineering in Iron Garnet Thin Films
No full textAt its heart, Materials Science and Engineering is a discipline seeking to advance technologies by improving the materials that make them. Currently, the development of electronic devices is limited by the need for materials that enhance speed, reduce size, and improve energy efficiency. Spin-based memory devices, which encode data in a material’s magnetic state, offer a promising solution. Magnetic memory technologies are widespread today, powering devices such as memory disks, tapes, and magnetic random access memory (MRAM). While garnet materials have long been studied for these applications, they have faced challenges in becoming adoptable technologies. However, with advanced research techniques and a deeper understanding of material behaviors, magnetic garnets are experiencing a renaissance. This thesis explores the engineering of iron garnet thin films for next-generation spin-based memory applications. The work presented advances the understanding of non-equilibrium growth, characterization, and engineering of iron garnet thin films and their magnetic properties, emphasizing kinetic phenomena that govern atomic organization beyond classic ordering of unit cells and emergent magnetic anisotropy. In a composition series of europium-thulium iron garnet (EuTmIG) films, experiment confirms the 50-year-old theory of cation site preference of Eu and Tm, demonstrating that enhanced magnetic anisotropy, named ’magnetotaxial anisotropy’, is linked to cation ordering during growth. These findings lay the foundation for anisotropy engineering by cation order. Further studies investigate the effects of film formulation, growth kinetics, and post-growth annealing on structural ordering and magnetotaxial anisotropy. In bismuth-yttrium iron garnet (BiYIG) films, a linear relationship between Bi-Y ordering, magnetic anisotropy, and substrate-lattice mismatch provides deeper insight into the forces that drive cation ordering. Annealing is shown to further enhance magnetic anisotropy in these films. In lutetium-yttrium iron garnet (LuYIG) films, the laser pulse rate during growth by pulsed laser deposition is shown to influence Lu-Y ordering and magnetic anisotropy, reinforcing the kinetic nature of the cation ordering. The findings of this thesis contribute to the fundamental understanding of cation ordering in complex oxide films and provide a framework for engineering and characterizing garnet materials, enabling the future development of new spintronic devices.Ph.D
Improving Tandem Fluency Through Utilization of Deep Learning to Predict Human Motion in Exoskeleton
No full textfirst_pagesettingsOrder Article Reprints
Open AccessArticle
Improving Tandem Fluency Through Utilization of Deep Learning to Predict Human Motion in Exoskeleton
by Bon Ho Koo 1ORCID,Ho Chit Siu 2ORCID,Luke Apostolides 1ORCID,Sangbae Kim 1 andLonnie G. Petersen 3,4,*ORCID
1
Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
2
MIT Lincoln Laboratory, Lexington, MA 02421, USA
3
Department of Aeronautics and Astronautics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
4
Institute for Medical Engineering and Science, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
*
Author to whom correspondence should be addressed.
Actuators 2025, 14(6), 260; https://doi.org/10.3390/act14060260
Submission received: 8 April 2025 / Revised: 13 May 2025 / Accepted: 20 May 2025 / Published: 23 May 2025
(This article belongs to the Special Issue Recent Advances in Soft Actuators, Robotics and Intelligence)
Downloadkeyboard_arrow_down Browse Figures Versions Notes
Abstract
Today’s exoskeletons face challenges with low fluency (a quantifiable alternative to “seamlessness”), hypothesized to be caused by a lag in active control innate in many leader–follower paradigms seen in contemporary systems, leading to inefficiencies and discomfort. Furthermore, tandem fluency, a variation of fluency specific for tandem robots systems as exoskeletons, is yet to be rigorously tested in practice. This study aims to utilize metrics of tandem fluency in order to demonstrate improved human–robot interaction (HRI) in exoskeletons through human subject testing of a prototype 1 degree of freedom (DoF) exoskeleton using a motion prediction bidirectional long short-term memory (bi-LSTM) deep learning network. Subjects were recruited to conduct various upper body exercises about the elbow joint, and the collected sEMG, goniometer, and gas exchange data was used to design, test, optimize, and assess the performance of the 1 DoF exoskeleton using tandem fluency metrics. We found that the correlation between I-ACT, a metric of tandem fluency, the subjective survey responses, and metabolic data suggest that the use of a predictive bi-LSTM network to control a 1 DoF exoskeleton about the elbow results in an overall positive trend, which may correlate to high tandem fluency
Carbon flow and food web structure in the mesopelagic zone of the North Atlantic Ocean
No full textMesopelagic ecosystems are vital habitats that link the euphotic zone and the deep ocean through food web interactions and carbon flow pathways. In this dissertation, I use carbon compound-specific stable carbon isotope analysis of amino acids (CSIA-AA) and DNA gut metabarcoding methodologies to provide a broad ecological outlook on mesopelagic carbon flow coupled with finer scale taxonomic details. In Chapter 2, I analyze the diets of seven abundant mesopelagic fish species by combining the integrative power of CSIA-AA with the instantaneous, taxonomic aspects of DNA gut metabarcoding. Three primary diet types were identified: copepod-based, fish-based, and generalist. Additionally, carbon sources were variable across the two years, but cyanobacteria were consistently an important carbon source - evidence that mesopelagic fish are essential exporters in weaker biological pump systems. Finally, this chapter includes cyanobacteria CSIA-AA signature data that was previously missing from the literature. In Chapter 3, I augment the CSIA-AA data by adding genus-level zooplankton data and samples from the winter season. Zooplankton were more dispersed among all the end members than fish, particularly in the winter. Fish, however, still relied the most on cyanobacteria-sourced carbon. This chapter supplies the first zooplankton carbon CSIA-AA data set at such a fine taxonomic resolution. In Chapter 4, I examine the effect of phytoplankton community structure on fish and zooplankton carbon sources by sampling before and during diatom a diatom bloom. Zooplankton, and to a lesser extent fish, showed a shift to diatom-based carbon sources during the bloom. As a whole, this dissertation advances our knowledge of mesopelagic food webs by providing a baseline carbon-CSIA-AA data set for key zooplankton and fish species across several seasons that will inform ecological models to understand how the mesopelagic will react to anthropogenic pressure.Ph.D
Search for heavy long-lived charged particles with large ionization energy loss in proton-proton collisions at s = 13 TeV
No full textA search for heavy, long-lived, charged particles with large ionization energy loss within the silicon tracker of the CMS experiment is presented. A data set of proton-proton collisions at a center of mass energy at s = 13 TeV, collected in 2017 and 2018 at the CERN LHC, corresponding to an integrated luminosity of 101 fb−1, is used in this analysis. Two different approaches for the search are taken. A new method exploits the independence of the silicon pixel and strips measurements, while the second method improves on previous techniques using ionization to determine a mass selection. No significant excess of events above the background expectation is observed. The results are interpreted in the context of the pair production of supersymmetric particles, namely gluinos, top squarks, and tau sleptons, and of the Drell-Yan pair production of fourth generation (τ′) leptons with an electric charge equal to or twice the absolute value of the electron charge (e). An interpretation of a Z’ boson decaying to two τ′ leptons with an electric charge equal to 2e is presented for the first time. The 95% confidence upper limits on the production cross section are extracted for each of these hypothetical particles
Mapping facade materials utilizing zero-shot segmentation for applications in urban microclimate research
No full textTo address the Urban Heat Island (UHI) effect-a significant urban climate challenge-detailed urban microclimate modeling is essential. Such modeling typically requires data on urban surface properties and morphologies from street canyons and buildings. Most urban surveying efforts have focused on morphological attributes such as sky view factor, vegetation or building surface ratio, while the mass-collection of facade materials has been hindered by the complexity of the segmentation task and the need for large and diverse labeled datasets. Recognizing the importance of mapping facade materials for urban thermal comfort, envelope heat emissions, and building energy studies, we employ computer vision-based state-of-the-art zero-shot learning paradigms for high-fidelity facade material extraction. Our approach circumvents the traditional need for extensive labeled training data, allowing for adaptation to a variety of urban contexts and material types. Tested in Dubai, Amsterdam, and Boston (three architecturally diverse cities), our algorithm successfully detects the predominant facade material in 68% of cases and identifies the top three present material classes in 85% of cases. Additionally, we show how material coverage identification is crucial for assessing outdoor thermal comfort, as evident in shifts in annual cold and heat stress hours across the climates of the three cities in a sample urban canyon
Neutronic Performance and Thermal Hydraulic Analysis of the MIT Reactor Fission Converter Experimental Facility Using High-Density U-10Mo Low-Enriched Uranium Fuel Elements
No full textThe MITR fission converter (FC) is a core-driven subcritical assembly at the MIT Nuclear Reactor Laboratory, located on the MIT campus in Cambridge, MA. The assembly is made of eleven partially-depleted MITR-II fuel elements in a separate cooling tank attached to the side of the core-tank graphite reflector. The FC serves to boost the thermal flux from the core and send a hardened neutron spectrum to an irradiation target, providing a fission energy flux spectrum without the need to put a sample inside the core tank. It was previously used for boron-neutron capture therapy clinical trials before its decommissioning in the 2010s. Recently, it has been modified from a medical beamline to a general-use engineering and materials testing facility. The new FC-based experimental facility has roughly one cubic meter of empty space downstream intended to contain large experiments, called the m³. This work is a safety and performance study aimed at quantifying the impact of modifying the facility’s geometry as part of the FC’s recommissioning, as well as the impact of changing its fuel from HEU to LEU fuel as part of the MITR LEU conversion project. Neutronics and thermal hydraulics analysis of the renovated facility have been performed using the codes MCNP5 and STAT7, respectively. This analysis quantified the FC’s k_eff, power distribution, multi-group neutron flux, and conditions which cause onset of nucleate boiling (ONB). It was determined that the FC assembly will remain subcritical (k
_eff < 0.9) and low power (≤200 kW) under a wide range of performance conditions, including with both types of fuel and a variety of materials on the target-side of the FC tank. The HEU-fueled FC is expected to require no changes to the limiting safety system settings (LSSS) outlined in the original technical specifications document. The LEU fuel is expected to increase the FC performance, but as a tradeoff, will require minor changes to the LSSS setpoints to maintain margin to ONB under the most limiting thermal-hydraulic conditions. Additionally, this study evaluates the feasibility of using the FC for in-assembly fuel experiments, particularly as a pathway for testing the new LEU fuel elements at low power. This study indicated that this proposed FC configuration with one LEU and ten HEU elements is feasible and maintains wide safety margins.S.M
Comment on ‘Physics-based representations for machine learning properties of chemical reactions’
No full textIn a recent article in this journal, van Gerwen et al (2022 Mach. Learn.: Sci. Technol. 3 045005)
presented a kernel ridge regression model to predict reaction barrier heights. Here, we comment on
the utility of that model and present references and results that contradict several statements made
in that article. Our primary interest is to offer a broader perspective by presenting three aspects
that are essential for researchers to consider when creating models for chemical kinetics: (1) are the
model’s prediction targets and associated errors sufficient for practical applications? (2) Does the
model prioritize user-friendly inputs so it is practical for others to integrate into prediction
workflows? (3) Does the analysis report performance on both interpolative and more challenging
extrapolative data splits so users have a realistic idea of the likely errors in the model’s predictions
Leveraging Competitive Sorption in Microporous Polymer Membranes to Enhance Gas Separation Performance
No full textChemical separations account for roughly half of the United States’ industrial energy consumption, 49% of which is attributed to distillation alone. Membrane-based systems, on the other hand, offer a more energy-efficient alternative to conventional separation processes because they do not require thermally intensive phase changes to operate. Specifically, polymer membranes with a more permanent porosity (termed “microporous”) have gained attention due to their impressive combination of permeability (throughput) and permselectivity (separation efficiency) relative to the empirically defined “upper bound” for membrane materials.
Traditionally, the permeability of a membrane for a gas is defined by the product of the gas’s diffusivity and sorption coefficient in the material. By extension, a membrane’s permselectivity can be broken down into the product of its diffusion selectivity and sorption selectivity. Microporous polymer membranes exhibit impressive diffusion selectivity due to their small free volume elements (< 2 nm) and rigid backbones. However, separating gases based primarily on size can become exceedingly difficult given that some gases differ in kinetic diameter by less than an angstrom. Instead, recent advancements in the design of microporous polymers have indicated that a phenomenon known as competitive sorption can be used to enhance separation performance by leveraging gas–polymer interactions instead of differences in gas diffusivity. This thesis investigates how the increase of sorption selectivity through competition between gases can be exploited to enhance the permselectivity of microporous polymer membranes. Specific focus is placed on the archetypal polymer of intrinsic microporosity (PIM-1) and its amine-functional analog (PIM-NH₂) to study how enhanced acid-gas (CO₂ and H₂S) sorption brought on by amine functionality positively impacts separation performance. To confirm the generalizability of these trends, competition effects in the microporous poly(arylene ether) (PAE) backbone were studied as well. To investigate more industrially viable membranes while retaining strong gas–polymer interactions afforded by the amine group, this PAE backbone was also used to develop 8 solution-processable tertiary-amine-functional analogs. Lastly, in an effort to study the effects of water vapor on CO₂-focused separations in amine-functional microporous polymer membranes, a humidified gas permeation apparatus was developed and used to measure dry and humidified CO₂ transport in PIM-1, PIM-NH₂, and a novel secondary-amine-functional analog, PIM-NHiPr. Taken together, this thesis focuses on the fundamentals and practical implications of leveraging competitive sorption to enhance performance in application-relevant and multi-component gas mixtures. More specifically, this work provides valuable insight regarding amine functionalization and its strong effects on sorption energetics and humidified gas transport that will help to inform future design of polymer membranes for gas separations.Ph.D