Scientific Discovery & Drug Design - Weekly Roundup

A pragma-based OpenACC acceleration strategy delivers a 5x speedup and 3x energy reduction for the ECsim Particle-In-Cell code, proving its readiness for exascale plasma simulations.

Estimating time-varying reproduction numbers just got more robust: CIRL blends epidemiological constraints with data-driven temporal representations, outperforming traditional methods in noisy, non-stationary conditions.

Forget static embeddings: this paper shows how modeling scientific concepts as evolving complex networks reveals surprising connections between conceptual change and network topology.

This model beats clinical reports in quantitative coronary angiography, opening the door to automated, comprehensive assessment of coronary artery disease at the point of care.

Accurately modeling notoriously flexible protein-glycan interactions is now more tractable thanks to a refined HADDOCK3 protocol that achieves sub-3 angstrom RMSD.

Estimating time-varying reproduction numbers just got more robust: CIRL blends epidemiological constraints with data-driven temporal representations, outperforming traditional methods in noisy, non-stationary conditions.

Forget static embeddings: this paper shows how modeling scientific concepts as evolving complex networks reveals surprising connections between conceptual change and network topology.

This model beats clinical reports in quantitative coronary angiography, opening the door to automated, comprehensive assessment of coronary artery disease at the point of care.

Accurately modeling notoriously flexible protein-glycan interactions is now more tractable thanks to a refined HADDOCK3 protocol that achieves sub-3 angstrom RMSD.

Standard ComBat harmonization methods fall apart when applied to dMRI data containing neurological disorders, but a simple MLP-based outlier compensation fixes the problem.

Quantum computers could finally unlock the full potential of machine learning for drug discovery by directly generating the quantum chemistry data that classical computers struggle to produce.

Seemingly accurate physics-informed surrogates can fail spectacularly when integrated into power system simulations, especially under stress, highlighting the need for rigorous in-simulator validation.

Generate realistic, atom-level molecular dynamics trajectories orders of magnitude faster with a novel State Space Model that captures long-range dependencies in biomolecular systems.

Ditch costly PIDE integration: RHYME-XT learns the flow map directly, offering a continuous-time, discretization-invariant representation that beats state-of-the-art neural operators.

Virtual cell perturbation prediction gets a 12x speedup in pretraining and a 12% boost in biological fidelity with SCALE, a new foundation model that prioritizes scalable infrastructure and biologically faithful evaluation.

Injecting semantic information from related modalities early in the embedding process significantly boosts performance on multimodal medical image classification tasks.

Precisely predicting bubble formation on electrocatalysts unlocks a new level of control over electrolyzer efficiency.

NNVMC's promise for solving quantum many-body problems is currently bottlenecked by surprisingly mundane issues: low-intensity elementwise operations and data movement on GPUs.

LLMs can now automatically generate bug-detection patterns for scientific code, offering a scalable solution to the growing problem of methodology errors in AI-driven research.

See how stochastic fluctuations in a simple Hückel model can bring condensed-phase spectroscopy to life for undergrads.

Drifting offers a surprisingly effective way to distill iterative Boltzmann sampling into a single forward pass, even with unknown normalization constants.

Unlock faster simulations of complex engineering systems with piecewise linear nonlinearities using a new data-driven model order reduction technique based on dynamic mode decomposition.

Convolutional Neural Operators (CNOs) surprisingly excel at capturing translated dynamics in the FitzHugh-Nagumo model, despite other architectures achieving lower training error or faster inference.

Simulate earthquake ground motion 10,000x faster with a new latent operator flow matching method, opening the door to real-time risk assessment for critical infrastructure.

Heuristic maritime routes lead to extreme fuel waste in nearly 5% of voyages, but this RL approach cuts that risk by almost 10x.

Quantum annealing could soon accelerate protein engineering: Q-BIOLAT formulates protein fitness as a QUBO problem, directly compatible with emerging quantum annealing hardware.

Molecular coatings can restructure the electromagnetic vacuum around nanoparticles, enabling strong coupling and Rabi oscillations where they were previously impossible.

Ditch the temperature ladder: Generative Replica Exchange (GREX) uses normalizing flows to generate high-temperature configurations on-demand, slashing the computational cost of replica exchange simulations.

LLMs can disentangle Long COVID pathology from confounding factors like menopause, achieving high precision in predicting disease severity using wearable sensor data.

Forget rigid circuits - this new method seamlessly weaves stretchable sensors directly into clothing using a clever combo of 3D printing and embroidery.

Precise calculations of electron-impact rotational excitation rates for water ion isotopologues provide critical data for modeling astrophysical plasma environments.

A national center focused on AI and robotics in medicine could be the key to unlocking the transformative potential of these technologies in healthcare.

Continuous, high-resolution shape sensing in steerable drilling robots is now possible without directly embedding sensors on the instrument surface, thanks to a clever OFDR-based assembly.

Cavity quantum electrodynamics offers a new knob for tuning bond-breaking processes in molecular materials, as demonstrated by a new method that explicitly includes quantized cavity photons in spin-flip configuration interaction calculations.

A specific Zn-Al alloy (Zn$_{0.9}$Al$_{0.1}$) achieves OER performance rivaling transition-metal catalysts, offering a low-cost alternative for alkaline water splitting.

Tackle previously intractable open quantum systems simulations with TENSO, a new open-source package that efficiently handles complex environments via tree tensor networks.

Unlock higher PEM fuel cell efficiency: in-phase current and temperature oscillations can slash resistivity by eliminating proton transport losses.

Diamond interfaces dramatically alter the phase diagram of high-pressure ice, suggesting that many experimental measurements may be skewed by interface effects.

Quantum field theory offers a fresh perspective on chemical interactions, promising to unlock new insights into reaction manipulation, large-scale system modeling, and unconventional scaling laws beyond the reach of standard quantum mechanics.

Isotopic substitution in water reveals intricate details of vibrational energy transfer and dephasing, offering a new lens into the interplay between intramolecular modes and their thermal environment.

A new prompting strategy closes the gap between general-purpose and specialized cell segmentation models, suggesting a path to more efficient adaptation.

Achieve oracle-level performance in multi-armed sequential hypothesis testing by betting, even without knowing which arm provides the most evidence.

Exploiting geometric symmetries in tensegrity structures slashes computational cost and boosts accuracy in physics-informed neural networks.

Accurately predict urban pollutant dispersion in real-time with a novel data-driven model that's orders of magnitude faster than traditional CFD.

By explicitly modeling pollutant propagation delays with neural delay differential equations, AirDDE significantly improves air quality forecasting accuracy.

Generative models can fail to produce globally consistent counterfactuals when causal graphs have complex topologies, but a novel sheaf-theoretic framework with entropic regularization can overcome these limitations.

Imagine seeing your tongue move in real-time based on the sounds you make – AURORA brings that closer to reality.

LLMs can now recommend drugs with state-of-the-art accuracy by synthesizing individual patient context with the prescribing tendencies of similar cases, outperforming guideline-based and similar-patient retrieval methods.

Neural networks can accurately predict polymer free energies, even when traditional methods like Bennett Acceptance Ratio fail due to poor phase-space overlap.

An AI model can estimate legal age from clavicle CT scans with higher accuracy than human experts, potentially revolutionizing forensic age assessment.

A new prompt-free medical image segmentation model achieves impressive zero-shot and cross-modal transfer performance by explicitly disentangling geometric and semantic anatomical knowledge.

AI spots a hidden pattern in lung scans of lupus patients, revealing that specific airway dilations in the upper lobes could be a telltale sign of interstitial lung disease.

Standard PCA, despite its widespread use in CAD, struggles to directly reveal the original design parameters of a geometry, but this paper identifies conditions for accurate parameter estimation.

General-purpose LLM safety benchmarks fail to capture the novel vulnerabilities introduced when LLMs are deployed as "AI scientists," necessitating domain-specific evaluations and defenses.

Predicting permeability tensors from microstructure images just got 33% more accurate thanks to a physics-informed CNN-Transformer that learns faster and generalizes better via pretraining and differentiable constraints.

By explicitly modeling cardiac pathology, this ECG reconstruction method achieves a 76% reduction in error compared to existing techniques, promising more accurate diagnoses from portable devices.

Medical vision-language models perform better when the modality gap is tuned to an intermediate level, challenging the assumption that minimizing it is always optimal.

Genetic programming can discover unconventional multigrid cycles that outperform hand-tuned methods, suggesting automated algorithm design can unlock untapped performance in classical numerical solvers.

ML's seismic and volcanic signal analysis needs more than just accuracy; it demands reliability under changing conditions, uncertainty quantification, and physically meaningful constraints.

Ultrafast X-ray spectroscopy reveals the hidden choreography of electronic state transitions that drive Norrish Type-I reactions, pinpointing the long-lived $^3n\pi^*$ state as the key player.

By handling input noise directly through Wasserstein distances, \PWAGPs offer a more robust and transparent approach to uncertainty quantification in GP regression compared to latent-input models.

Unlocking insights from massive molecular dynamics simulations just got easier: covariance matrix comparisons reveal key physical properties and phase transitions with remarkable data efficiency.

By explicitly modeling tooth relationships, TCATSeg achieves state-of-the-art accuracy in 3D dental model segmentation, even in challenging pre-orthodontic cases.

DINOv2's powerful visual features come with a hidden flaw: strong positional biases that ALiBi positional encoding can effectively mitigate.

Finally, a unified software framework promises to tame the wild west of quantum dot device tuning, enabling researchers to share and adapt characterization routines across labs.

Early-career researchers in experimental physics report significant gaps in training for software and machine learning tools crucial to their work, highlighting a critical need for improved educational resources.

By injecting biological heuristics into a deep learning pipeline, this method achieves state-of-the-art performance in classifying rare white blood cell subtypes, a task where standard deep learning models often fail.

Achieve expert-level accuracy in wasp identification with a YOLO-based model that also shows *why* it makes its classifications, thanks to integrated HiResCAM explainability.

You can now train graph transformers that generalize across different mesh resolutions, thanks to a new architecture that maintains gauge invariance while scaling linearly.

By forcing a model to reconstruct aggressively masked EEG spectrograms, SpecMoE learns intricate neural patterns across both high- and low-frequency domains, leading to state-of-the-art cross-species EEG decoding.

By blending geometry with classification, this new Finsler metric lets you trace trajectories more accurately through complex systems, like cell development, where you have both spatial data and lineage trees.

Unsupervised pretraining of drug-response models offers clear gains when adapting to patient tumors with very limited labeled data, despite providing limited benefit when source and target domains overlap substantially.

Solve Fokker-Planck equations on manifolds without meshes by pushing forward samples with neural networks.

Fuzzy logic and deep learning join forces to make radio astronomy ML pipelines less black-box.

Achieve near-linear scaling and 40x speedup for MP2 calculations on large molecules by unleashing multi-GPU parallelism for local correlation methods.

Deep learning slashes design time for high-efficiency Doherty power amplifiers, enabling complex pixelated combiners that extend the back-off efficiency range.

Directly modeling 3D geometry in dental scans unlocks a 9.58% accuracy boost in multi-disease diagnosis compared to methods relying on 2D or multi-view image representations.

Most scientific claims in NLP die in obscurity, and even the survivors are more likely to be subtly reshaped than outright validated or debunked.

Unlock a deeper understanding of cryptographic security by bridging the gap between torsor-based reasoning and the power of sheaf and topos theory.

Achieve intention-driven start-stop control of a rehabilitation exoskeleton from non-invasive EEG by fixing a common bias in task-based recentering.

Diffusion models can now solve full-waveform inversion problems more robustly than traditional optimization or standard diffusion posterior sampling, even with noisy data.

You don't need predefined features: a Transformer trained on raw microscopy videos can predict cancer cell fate with high accuracy, revealing the temporal dynamics of cellular decision-making.

A new loss function lets you train a deep learning model to detect rare bee and wasp brood cells with minimal labeling effort, even when data is highly imbalanced.

Federated learning can match or beat centralized models for predicting postoperative complications, all while keeping patient data siloed at each hospital.

By jointly estimating the mapping from calibration parameters to VAE-encoded image representations, this work achieves a 2x reduction in error when calibrating electron microscopes, demonstrating the power of bridging simulation and reality.

Ditch the manual tuning: SuCor automatically corrects EPI image distortions using optimal transport, outperforming FSL TOPUP in accuracy and speed.

Surgical AI gets a major data boost: Surg$Σ$ unifies millions of surgical conversations with multimodal annotations, paving the way for more generalizable and interpretable models.

Forget reasoning, AI's next frontier is "scientific taste"—and fine-tuning on publication records beats both LLMs and expert panels at judging research ideas.

Ditch the D-score: a new Bayesian model boosts IAT accuracy for mental health assessments, rivaling complex methods without task-specific tuning.

Accurately model complex nonlinear forces in mechanical systems using a surprisingly simple combination of min/max functions, unlocking better forced response analysis.

Ditch signal reconstruction for EEG foundation models: LeJEPA-based Laya learns better, more transferable representations by predicting latent features instead.

By penalizing treatment plans that lead to trajectory distributions far from observed patient data, this method provides a more robust approach to treatment optimization than standard model-based methods.

Optimizing time reparameterization for smoothness unlocks stable and accurate explicit ML-ROMs for stiff dynamical systems, outperforming existing methods by orders of magnitude.

Forget retraining for every new physics problem – pADAM learns a single generative model that handles forward prediction, inverse inference, and even identifies governing laws across different PDEs.

Ditch the manual tracing: a new graph-based algorithm automates the segmentation and measurement of complex 3D detonation cells with high accuracy.

Automated microscopy can now actively discover new scientific information by searching for diverse functional responses, rather than being limited to optimizing for known objectives.

An interpretable machine learning framework leveraging XGBoost and DeepSeek reveals key genetic factors driving drug response in lung cancer, offering a path towards personalized treatment strategies.

Positrons don't always behave as expected: they delay electron ionization in PsH but enhance it in PsCl when exposed to lasers.

A pragma-based OpenACC acceleration strategy delivers a 5x speedup and 3x energy reduction for the ECsim Particle-In-Cell code, proving its readiness for exascale plasma simulations.

Forget curated datasets – this work shows you can bootstrap AI scientists by training them on automatically generated, self-verified ML tasks, leading to significant performance gains on MLGym.

A tabular LLM, TAP-GPT, rivals state-of-the-art general-purpose LLMs in few-shot Alzheimer's prediction while offering interpretable reasoning and robustness to missing data, opening the door to more transparent and reliable clinical AI.

A new rank-based phenotypic characterization scheme slashes the computational cost of genetic programming for dynamic project scheduling, enabling faster discovery of high-quality heuristic rules.

Quantum protons get a speed boost: NEO-ELMD slashes computational cost for simulating proton transfer, opening doors to simulating larger, more complex chemical systems.

By amortizing sequential design into a neural network, this method achieves real-time model-based design of experiments, unlocking new possibilities for efficient parameter estimation in complex dynamical systems.

Time gating and kinetic-energy filtering in photoelectron-detected 2D spectroscopy can now disentangle exciton-exciton annihilation, unlocking previously obscured energy transfer dynamics.