More Like this

1. Supramolecular Approach to Electron Paramagnetic Resonance Distance Measurement of Spin-Labeled Proteins

   https://doi.org/10.1021/acs.jpcb.0c00743  

   Yang, Zhimin; Stein, Richard A.; Ngendahimana, Thacien; Pink, Maren; Rajca, Suchada; Jeschke, Gunnar; Eaton, Sandra S; Eaton, Gareth R.; Mchaourab, Hassane S; Rajca, Andrzej (March 2020, The Journal of Physical Chemistry B)

   We demonstrate a host-guest molecular recognition approach to advance double electron-electron resonance (DEER) distance measurements of spin-labeled proteins. We synthesized an iodoacetamide (IA) derivative of 2,6-diazaadamantane nitroxide (DZD) spin label that could be doubly incorporated into T4 Lysozyme (T4L) by site-directed spin labeling (SDSL) with efficiency up to 50% per cysteine. The rigidity of the fused ring structure and absence of mobile methyl groups increase the spin echo dephasing time (Tm) at temperatures above 80 K. This enables DEER measurements of distances >4 nm in DZD labeled-T4L in glycerol/water at temperatures up to 150 K, with increased sensitivity compared to common spin label such as MTSL. Addition of β-cyclodextrin (β-CD) reduces the rotational correlation time of the label, slightly increases Tm, and most importantly, narrows (and slightly lengthens) the inter-spin distance distributions. The distance distributions are in good agreement with simulated distance distributions obtained by rotamer libraries. These results provide a foundation for developing supramolecular recognition to facilitate long-distance DEER measurements at near physiological temperatures. 

   more » « less
2. Solving Bayesian Inverse Problems via Variational Autoencoders

   Goh, H. (October 2021, Proceeding of Machine Learning Research, 2nd Annual Conference on Mathematical and Scientific Machine Learning)

   In recent years, the field of machine learning has made phenomenal progress in the pursuit of simulating real-world data generation processes. One notable example of such success is the variational autoencoder (VAE). In this work, with a small shift in perspective, we leverage and adapt VAEs for a different purpose: uncertainty quantification in scientific inverse problems. We introduce UQ-VAE: a flexible, adaptive, hybrid data/model-informed framework for training neural networks capable of rapid modelling of the posterior distribution representing the unknown parameter of interest. Specifically, from divergence-based variational inference, our framework is derived such that most of the information usually present in scientific inverse problems is fully utilized in the training procedure. Additionally, this framework includes an adjustable hyperparameter that allows selection of the notion of distance between the posterior model and the target distribution. This introduces more flexibility in controlling how optimization directs the learning of the posterior model. Further, this framework possesses an inherent adaptive optimization property that emerges through the learning of the posterior uncertainty. 

   more » « less
3. Solving Bayesian Inverse Problems via Variational Autoencoders

   Goh, H. (January 2021, Proceeding of Machine Learning Research, 2nd Annual Conference on Mathematical and Scientific Machine Learning)

   Joan Bruna, Jan S (Ed.)

   In recent years, the field of machine learning has made phenomenal progress in the pursuit of simulating real-world data generation processes. One notable example of such success is the variational autoencoder (VAE). In this work, with a small shift in perspective, we leverage and adapt VAEs for a different purpose: uncertainty quantification in scientific inverse problems. We introduce UQ-VAE: a flexible, adaptive, hybrid data/model-constrained framework for training neural networks capable of rapid modelling of the posterior distribution representing the unknown parameter of interest. Specifically, from divergence-based variational inference, our framework is derived such that most of the information usually present in scientific inverse problems is fully utilized in the training procedure. Additionally, this framework includes an adjustable hyperparameter that allows selection of the notion of distance between the posterior model and the target distribution. This introduces more flexibility in controlling how optimization directs the learning of the posterior model. Further, this framework possesses an inherent adaptive optimization property that emerges through the learning of the posterior uncertainty. Numerical results for an elliptic PDE-constrained Bayesian inverse problem are provided to verify the proposed framework. 

   more » « less
4. An optimal acquisition scheme for Q-band EPR distance measurements using Cu ²⁺ -based protein labels

   https://doi.org/10.1039/d2cp01032a  

   Bogetti, Xiaowei; Hasanbasri, Zikri; Hunter, Hannah R.; Saxena, Sunil (January 2022, Physical Chemistry Chemical Physics)

   Recent advances in site-directed Cu 2+ labeling of proteins and nucleic acids have added an attractive new methodology to measure the structure-function relationship in biomolecules. Despite the promise, accessing the higher sensitivity of Q-band Double Electron Electron Resonance (DEER) has been challenging for Cu 2+ labels designed for proteins. Q-band DEER experiments on this label typically require many measurements at different magnetic fields, since the pulses can excite only a few orientations at a given magnetic field. Herein, we analyze such orientational effects through simulations and show that three DEER measurements, at strategically selected magnetic fields, are generally sufficient to acquire an orientational-averaged DEER time trace for this spin label at Q-band. The modeling results are experimentally verified on Cu 2+ labeled human glutathione S-transferase (hGSTA1-1). The DEER distance distribution measured at the Q-band shows good agreement with the distance distribution sampled by molecular dynamics (MD) simulations and X-band experiments. The concordance of MD sampled distances and experimentally measured distances adds growing evidence that MD simulations can accurately predict distances for the Cu 2+ labels, which remains a key bottleneck for the commonly used nitroxide label. In all, this minimal collection scheme reduces data collection time by as much as six-fold and is generally applicable to many octahedrally coordinated Cu 2+ systems. Furthermore, the concepts presented here may be applied to other metals and pulsed EPR experiments. 

   more » « less
5. Active multi-fidelity Bayesian online changepoint detection

   Gundersen, Gregory W.; Cai, Diana; Zhou, Chuteng; Engelhardt, Barbara E.; Adams, Ryan P. (January 2021, Uncertainty in artificial intelligence)

   null (Ed.)

   Online algorithms for detecting changepoints, or abrupt shifts in the behavior of a time series, are often deployed with limited resources, e.g., to edge computing settings such as mobile phones or industrial sensors. In these scenarios it may be beneficial to trade the cost of collecting an environmental measurement against the quality or "fidelity" of this measurement and how the measurement affects changepoint estimation. For instance, one might decide between inertial measurements or GPS to determine changepoints for motion. A Bayesian approach to changepoint detection is particularly appealing because we can represent our posterior uncertainty about changepoints and make active, cost-sensitive decisions about data fidelity to reduce this posterior uncertainty. Moreover, the total cost could be dramatically lowered through active fidelity switching, while remaining robust to changes in data distribution. We propose a multi-fidelity approach that makes cost-sensitive decisions about which data fidelity to collect based on maximizing information gain with respect to changepoints. We evaluate this framework on synthetic, video, and audio data and show that this information-based approach results in accurate predictions while reducing total cost. 

   more » « less