PIDiff: Physics informed diffusion model for protein pocket-specific 3D molecular generation.

Designing drugs capable of binding to the structure of target proteins for treating diseases is essential in drug development. Recent remarkable advancements in geometric deep learning have led to unprecedented progress in three-dimensional (3D) generation of ligands that can bind to the protein pocket. However, most existing methods primarily focus on modeling the geometric information of ligands in 3D space. Consequently, these methods fail to consider that the binding of proteins and ligands is a phenomenon driven by intrinsic physicochemical principles. Motivated by this understanding, we propose PIDiff, a model for generating molecules by accounting in the physicochemical principles of protein-ligand binding. Our model learns not only the structural information of proteins and ligands but also to minimize the binding free energy between them. To evaluate the proposed model, we introduce an experimental framework that surpasses traditional assessment methods by encompassing various essential aspects for the practical application of generative models to actual drug development. The results confirm that our model outperforms baseline models on the CrossDocked2020 benchmark dataset, demonstrating its superiority. Through diverse experiments, we have illustrated the promising potential of the proposed model in practical drug development.

On-line measurement of COD and nitrate in water against stochastic background interference based on ultraviolet-visible spectroscopy and physics-informed multi-task learning.

Traditional ultraviolet-visible spectroscopic quantitative analytical methods face challenges in simultaneous and long-term accurate measurement of chemical oxygen demand (COD) and nitrate due to spectral overlap and the interference from stochastic background caused by turbidity and chromaticity in water. Addressing these limitations, a compact dual optical path spectrum detection sensor is introduced, and a novel ultraviolet-visible spectroscopic quantitative analysis model based on physics-informed multi-task learning (PI-MTL) is designed. Incorporating a physics-informed block, the PI-MTL model integrates pre-existing physical knowledge for enhanced feature extraction specific to each task. A multi-task loss wrapper strategy is also employed, facilitating comprehensive loss evaluation and adaptation to stochastic backgrounds. This novel approach significantly outperforms conventional models in COD and nitrate measurement under stochastic background interference, achieving impressive prediction R2 values of 0.941 for COD and 0.9575 for nitrate, while reducing root mean squared error (RMSE) by 60.89 % for COD and 77.3 % for nitrate in comparison to the conventional chemometric model partial least squares regression (PLSR), and by 30.59 % and 65.96 %, respectively, in comparison to a benchmark convolutional neural network (CNN) model. The promising results emphasize its potential as a spectroscopic instrument designed for online multi-parameter water quality monitoring against stochastic background interference, enabling long-term accurate measurement of COD and nitrate levels.

Kinetic model-informed deep learning for multiplexed PET image separation.

BACKGROUND: Multiplexed positron emission tomography (mPET) imaging can measure physiological and pathological information from different tracers simultaneously in a single scan. Separation of the multiplexed PET signals within a single PET scan is challenging due to the fact that each tracer gives rise to indistinguishable 511 keV photon pairs, and thus no unique energy information for differentiating the source of each photon pair. METHODS: Recently, many applications of deep learning for mPET image separation have been concentrated on pure data-driven methods, e.g., training a neural network to separate mPET images into single-tracer dynamic/static images. These methods use over-parameterized networks with only a very weak inductive prior. In this work, we improve the inductive prior of the deep network by incorporating a general kinetic model based on spectral analysis. The model is incorporated, along with deep networks, into an unrolled image-space version of an iterative fully 4D PET reconstruction algorithm. RESULTS: The performance of the proposed method was evaluated on a simulated brain image dataset for dual-tracer [ 18 F]FDG+[ 11 C]MET PET image separation. The results demonstrate that the proposed method can achieve separation performance comparable to that obtained with single-tracer imaging. In addition, the proposed method outperformed the model-based separation methods (the conventional voxel-wise multi-tracer compartment modeling method (v-MTCM) and the image-space dual-tracer version of the fully 4D PET image reconstruction algorithm (IS-F4D)), as well as a pure data-driven separation [using a convolutional encoder-decoder (CED)], with fewer training examples. CONCLUSIONS: This work proposes a kinetic model-informed unrolled deep learning method for mPET image separation. In simulation studies, the method proved able to outperform both the conventional v-MTCM method and a pure data-driven CED with less training data.

A Phone in a Basket Looks Like a Knife in a Cup: Role-Filler Independence in Visual Processing.

When a piece of fruit is in a bowl, and the bowl is on a table, we appreciate not only the individual objects and their features, but also the relations containment and support, which abstract away from the particular objects involved. Independent representation of roles (e.g., containers vs. supporters) and "fillers" of those roles (e.g., bowls vs. cups, tables vs. chairs) is a core principle of language and higher-level reasoning. But does such role-filler independence also arise in automatic visual processing? Here, we show that it does, by exploring a surprising error that such independence can produce. In four experiments, participants saw a stream of images containing different objects arranged in force-dynamic relations-e.g., a phone contained in a basket, a marker resting on a garbage can, or a knife sitting in a cup. Participants had to respond to a single target image (e.g., a phone in a basket) within a stream of distractors presented under time constraints. Surprisingly, even though participants completed this task quickly and accurately, they false-alarmed more often to images matching the target's relational category than to those that did not-even when those images involved completely different objects. In other words, participants searching for a phone in a basket were more likely to mistakenly respond to a knife in a cup than to a marker on a garbage can. Follow-up experiments ruled out strategic responses and also controlled for various confounding image features. We suggest that visual processing represents relations abstractly, in ways that separate roles from fillers.

Application of physics encoded neural networks to improve predictability of properties of complex multi-scale systems.

Predicting physical properties of complex multi-scale systems is a common challenge and demands analysis of various temporal and spatial scales. However, physics alone is often not sufficient due to lack of knowledge on certain details of the system. With sufficient data, however, machine learning techniques may aid. If data are yet relatively cumbersome to obtain, hybrid methods may come to the rescue. We focus in this report on using various types of neural networks (NN) including NN's into which physics information is encoded (PeNN's) and also studied effects of NN's hyperparameters. We apply the networks to predict the viscosity of an emulsion as a function of shear rate. We show that using various network performance metrics as the mean squared error and the coefficient of determination ( R 2 ) that the PeNN's always perform better than the NN's, as also confirmed by a Friedman test with a p-value smaller than 0.0002. The PeNN's capture extrapolation and interpolation very well, contrary to the NN's. In addition, we have found that the NN's hyperparameters including network complexity and optimization methods do not have any effect on the above conclusions. We suggest that encoding NN's with any disciplinary system based information yields promise to better predict properties of complex systems than NN's alone, which will be in particular advantageous for small numbers of data. Such encoding would also be scalable, allowing different properties to be combined, without repetitive training of the NN's.

Breakthrough-induced loop formation in evolving transport networks.

Transport networks, such as vasculature or river networks, provide key functions in organisms and the environment. They usually contain loops whose significance for the stability and robustness of the network is well documented. However, the dynamics of their formation is usually not considered. Such structures often grow in response to the gradient of an external field. During evolution, extending branches compete for the available flux of the field, which leads to effective repulsion between them and screening of the shorter ones. Yet, in remarkably diverse processes, from unstable fluid flows to the canal system of jellyfish, loops suddenly form near the breakthrough when the longest branch reaches the boundary of the system. We provide a physical explanation for this universal behavior. Using a 1D model, we explain that the appearance of effective attractive forces results from the field drop inside the leading finger as it approaches the outlet. Furthermore, we numerically study the interactions between two fingers, including screening in the system and its disappearance near the breakthrough. Finally, we perform simulations of the temporal evolution of the fingers to show how revival and attraction to the longest finger leads to dynamic loop formation. We compare the simulations to the experiments and find that the dynamics of the shorter finger are well reproduced. Our results demonstrate that reconnection is a prevalent phenomenon in systems driven by diffusive fluxes, occurring both when the ratio of the mobility inside the growing structure to the mobility outside is low and near the breakthrough.

Towards interpretable Cryo-EM: disentangling latent spaces of molecular conformations.

Molecules are essential building blocks of life and their different conformations (i.e., shapes) crucially determine the functional role that they play in living organisms. Cryogenic Electron Microscopy (cryo-EM) allows for acquisition of large image datasets of individual molecules. Recent advances in computational cryo-EM have made it possible to learn latent variable models of conformation landscapes. However, interpreting these latent spaces remains a challenge as their individual dimensions are often arbitrary. The key message of our work is that this interpretation challenge can be viewed as an Independent Component Analysis (ICA) problem where we seek models that have the property of identifiability. That means, they have an essentially unique solution, representing a conformational latent space that separates the different degrees of freedom a molecule is equipped with in nature. Thus, we aim to advance the computational field of cryo-EM beyond visualizations as we connect it with the theoretical framework of (nonlinear) ICA and discuss the need for identifiable models, improved metrics, and benchmarks. Moving forward, we propose future directions for enhancing the disentanglement of latent spaces in cryo-EM, refining evaluation metrics and exploring techniques that leverage physics-based decoders of biomolecular systems. Moreover, we discuss how future technological developments in time-resolved single particle imaging may enable the application of nonlinear ICA models that can discover the true conformation changes of molecules in nature. The pursuit of interpretable conformational latent spaces will empower researchers to unravel complex biological processes and facilitate targeted interventions. This has significant implications for drug discovery and structural biology more broadly. More generally, latent variable models are deployed widely across many scientific disciplines. Thus, the argument we present in this work has much broader applications in AI for science if we want to move from impressive nonlinear neural network models to mathematically grounded methods that can help us learn something new about nature.

Gendered hiring and attrition on the path to parity for academic faculty.

Despite long-running efforts to increase gender diversity among tenured and tenure-track faculty in the U.S., women remain underrepresented in most academic fields, sometimes dramatically so. Here, we quantify the relative importance of faculty hiring and faculty attrition for both past and future faculty gender diversity using comprehensive data on the training and employment of 268,769 tenured and tenure-track faculty rostered at 12,112U.S. PhD-granting departments, spanning 111 academic fields between 2011 and 2020. Over this time, we find that hiring had a far greater impact on women's representation among faculty than attrition in the majority (90.1%) of academic fields, even as academia loses a higher share of women faculty relative to men at every career stage. Finally, we model the impact of five specific policy interventions on women's representation, and project that eliminating attrition differences between women and men only leads to a marginal increase in women's overall representation-in most fields, successful interventions will need to make substantial and sustained changes to hiring in order to reach gender parity.

Feasibility and flexibility of a novel multi-dose level avoidance reirradiation technical methodology in recurrent head and neck cancer.

Reirradiation in recurrent head and neck cancer presents a considerable clinical challenge in radiation oncology. Though technically feasible due to advanced treatment delivery and planning techniques, confidence in delivering such treatments is not universal and patient selection is critical. Radiotherapy planning in reirradiation cases presents a complex technical challenge owing to the often-considerable overlap of dose from a patient's first treatment plan. This technical note describes three clinical case studies of recurrent head and neck cancer and the technical details of how their multidose level reirradiation was planned. Each patient had confirmed recurrence of squamous cell carcinoma and was referred for reirradiation to a previously irradiated area. The clinical details for each patient are provided before a detailed description of the treatment planning methodology is presented, which specifies how to approach such complex overlapping treatment volumes. The patient outcomes are described and a discussion is presented outlining the clinical challenges associated with these cases and the variables that must be accounted for when considering patients for potential reirradiation.

Free-Docking and Template-Based Docking: Physics Versus Knowledge-Based Docking.

Docking methods can be used to predict the orientations of two or more molecules with respect of each other using a plethora of various algorithms, which can be based on the physics of interactions or can use information from databases and templates. The usability of these approaches depends on the type and size of the molecules, whose relative orientation will be estimated. The two most important limitations are (i) the computational cost of the prediction and (ii) the availability of the structural information for similar complexes. In general, if there is enough information about similar systems, knowledge-based and template-based methods can significantly reduce the computational cost while providing high accuracy of the prediction. However, if the information about the system topology and interactions between its partners is scarce, physics-based methods are more reliable or even the only choice. In this chapter, knowledge-, template-, and physics-based methods will be compared and briefly discussed providing examples of their usability with a special emphasis on physics-based protein-protein, protein-peptide, and protein-fullerene docking in the UNRES coarse-grained model.

A method for empirically validating FMEA RPN scores in a radiation oncology clinic using physics QC data.

In failure modes and effects analysis (FMEA), the components of the risk priority number (RPN) for a failure mode (FM) are often chosen by consensus. We describe an empirical method for estimating the occurrence (O) and detectability (D) components of a RPN. The method requires for a given FM that its associated quality control measure be performed twice as is the case when a FM is checked for in an initial physics check and again during a weekly physics check. If instances of the FM caught by these checks are recorded, O and D can be computed. Incorporation of the remaining RPN component, Severity, is discussed. This method can be used as part of quality management design ahead of an anticipated FMEA or afterwards to validate consensus values.

Phase diagram simulations incorporating the gap anisotropy with AFM spin and charge density wave under spin-orbital coupling in Fe-based superconductors.

For over a decade, iron-based superconductors (IBSCs) have been the subject of intense scientific research, yet the underlying principle of their pairing mechanism remains elusive. To address this, we have developed a simulation tool that reasonably predicts the regional superconducting phase diagrams of key IBSCs, incorporating factors such as anisotropic superconducting gap, spin-orbital coupling, electron-phonon coupling, antiferromagnetism, spin density wave, and charge transfer. Our focus has been on bulk FeSe, LiFeAs, NaFeAs, and FeSe films on SrTiO3 substrates. By incorporating angle-resolved photoemission spectroscopy (ARPES) data to fine-tune the electron concentration in the superconducting state, our simulations have successfully predicted the theoretical superconducting transition temperature (Tc) of these compounds, closely matching experimental results. Our research not only aids in identifying patterns and establishing correlations with Tc but also provides a simulation tool for potentially predicting high-pressure phase diagrams.

On-chip omnidirectional electromagnetic-thermal cloak.

Simultaneously guiding electromagnetic waves and heat flow at any incidence angle to smoothly bypass some electromagnetic/thermal sensitive elements is a key factor to ensure efficient communication and thermal protection for an on-chip system. In this study, an omnidirectional on-chip electromagnetic-thermal cloak is proposed. Firstly, a holey metallic plate with periodic array of subwavelength apertures is designed by optical surface transformation to realize an omnidirectional electromagnetic cloaking module for on-chip electromagnetic signal. Secondly, a two-layer ring-shaped engineered thermal structure is designed by solving Laplace equation to realize an omnidirectional thermal cloaking module for in-chip heat flow. Finally, these two cloaking modules are combined to achieve cloaking effect for both the electromagnetic waves and thermal fields simultaneously, thus protecting the build-in electromagnetic/thermal sensitive elements without disturbing the external fields. The proposed electromagnetic-thermal cloak may have potential advantage in dealing with omnidirectional electromagnetic compatibility/shielding and multi-directional thermal management/dissipation of an on-chip system.

Discontinuous Shear Thickening in Biological Tissue Rheology.

During embryonic morphogenesis, tissues undergo dramatic deformations in order to form functional organs. Similarly, in adult animals, living cells and tissues are continually subjected to forces and deformations. Therefore, the success of embryonic development and the proper maintenance of physiological functions rely on the ability of cells to withstand mechanical stresses as well as their ability to flow in a collective manner. During these events, mechanical perturbations can originate from active processes at the single-cell level, competing with external stresses exerted by surrounding tissues and organs. However, the study of tissue mechanics has been somewhat limited to either the response to external forces or to intrinsic ones. In this work, we use an active vertex model of a 2D confluent tissue to study the interplay of external deformations that are applied globally to a tissue with internal active stresses that arise locally at the cellular level due to cell motility. We elucidate, in particular, the way in which this interplay between globally external and locally internal active driving determines the emergent mechanical properties of the tissue as a whole. For a tissue in the vicinity of a solid-fluid jamming or unjamming transition, we uncover a host of fascinating rheological phenomena, including yielding, shear thinning, continuous shear thickening, and discontinuous shear thickening. These model predictions provide a framework for understanding the recently observed nonlinear rheological behaviors in vivo.

Gouy phase effects on photocurrents in plasmonic nanogaps driven by single-cycle pulses.

The investigation of optical phenomena in the strong-field regime requires few-cycle laser pulses at field strengths exceeding gigavolts per meter (GV/m). Surprisingly, such conditions can be reached by tightly focusing pJ-level pulses with nearly octave spanning optical bandwidth onto plasmonic nanostructures, exploiting the field-enhancement effect. In this situation, the Gouy phase of the focused beam can deviate significantly from the monochromatic scenario. Here, we study the effect of the Gouy phase of a pulse exploited to drive coherent strong-field photocurrents within a plasmonic gap nanoantenna. While the influence of the specific Gouy phase profile in the experiment approaches the monochromatic case closely, this scheme may be utilized to identify more intricate phase profiles at sub-diffraction scale. Our results pave the way for Gouy phase engineering at picojoule (pJ) pulse energy levels, enabling the optimization of strong-field optical phenomena.