ArXiv最新文献

Surface meshes are a favoured domain for representing structural and functional information on the human cortex, but their complex topology and geometry pose significant challenges for deep learning analysis. While Transformers have excelled as domainagnostic architectures for sequence-to-sequence learning, the quadratic cost of the self-attention operation remains an obstacle for many dense prediction tasks. Inspired by some of the latest advances in hierarchical modelling with vision transformers, we introduce the Multiscale Surface Vision Transformer (MS-SiT) as a backbone architecture for surface deep learning. The self-attention mechanism is applied within local-mesh-windows to allow for high-resolution sampling of the underlying data, while a shifted-window strategy improves the sharing of information between windows. Neighbouring patches are successively merged, allowing the MS-SiT to learn hierarchical representations suitable for any prediction task. Results demonstrate that the MS-SiT outperforms existing surface deep learning methods for neonatal phenotyping prediction tasks using the Developing Human Connectome Project (dHCP) dataset. Furthermore, building the MS-SiT backbone into a U-shaped architecture for surface segmentation demonstrates competitive results on cortical parcellation using the UK Biobank (UKB) and manually-annotated MindBoggle datasets. Code and trained models are publicly available at https://github.com/metrics-lab/surface-vision-transformers.

Motivation: Generative deep learning methods have recently been proposed for generating 3D molecules using equivariant graph neural networks (GNNs) within a denoising diffusion framework. However, such methods are unable to learn important geometric properties of 3D molecules, as they adopt molecule-agnostic and non-geometric GNNs as their 3D graph denoising networks, which notably hinders their ability to generate valid large 3D molecules.

Results: In this work, we address these gaps by introducing the Geometry-Complete Diffusion Model (GCDM) for 3D molecule generation, which outperforms existing 3D molecular diffusion models by significant margins across conditional and unconditional settings for the QM9 dataset and the larger GEOM-Drugs dataset, respectively. Importantly, we demonstrate that GCDM's generative denoising process enables the model to generate a significant proportion of valid and energetically-stable large molecules at the scale of GEOM-Drugs, whereas previous methods fail to do so with the features they learn. Additionally, we show that extensions of GCDM can not only effectively design 3D molecules for specific protein pockets but can be repurposed to consistently optimize the geometry and chemical composition of existing 3D molecules for molecular stability and property specificity, demonstrating new versatility of molecular diffusion models.

Availability: Code and data are freely available on GitHub.

Pediatric tumors of the central nervous system are the most common cause of cancer-related death in children. The five-year survival rate for high-grade gliomas in children is less than 20%. Due to their rarity, the diagnosis of these entities is often delayed, their treatment is mainly based on historic treatment concepts, and clinical trials require multi-institutional collaborations. The MICCAI Brain Tumor Segmentation (BraTS) Challenge is a landmark community benchmark event with a successful history of 12 years of resource creation for the segmentation and analysis of adult glioma. Here we present the CBTN-CONNECT-DIPGR-ASNR-MICCAI BraTS-PEDs 2023 challenge, which represents the first BraTS challenge focused on pediatric brain tumors with data acquired across multiple international consortia dedicated to pediatric neuro-oncology and clinical trials. The BraTS-PEDs 2023 challenge focuses on benchmarking the development of volumentric segmentation algorithms for pediatric brain glioma through standardized quantitative performance evaluation metrics utilized across the BraTS 2023 cluster of challenges. Models gaining knowledge from the BraTS-PEDs multi-parametric structural MRI (mpMRI) training data will be evaluated on separate validation and unseen test mpMRI dataof high-grade pediatric glioma. The CBTN-CONNECT-DIPGR-ASNR-MICCAI BraTS-PEDs 2023 challenge brings together clinicians and AI/imaging scientists to lead to faster development of automated segmentation techniques that could benefit clinical trials, and ultimately the care of children with brain tumors.

Capturing how the Caenorhabditis elegans connectome structure gives rise to its neuron functionality remains unclear. It is through fiber symmetries found in its neuronal connectivity that synchronization of a group of neurons can be determined. To understand these we investigate graph symmetries and search for such in the symmetrized versions of the forward and backward locomotive sub-networks of the Caenorhabditi elegans worm neuron network. The use of ordinarily differential equations simulations admissible to these graphs are used to validate the predictions of these fiber symmetries and are compared to the more restrictive orbit symmetries. Additionally fibration symmetries are used to decompose these graphs into their fundamental building blocks which reveal units formed by nested loops or multilayered fibers. It is found that fiber symmetries of the connectome can accurately predict neuronal synchronization even under not idealized connectivity as long as the dynamics are within stable regimes of simulations.

Many physics-based and machine-learned scoring functions (SFs) used to predict protein-ligand binding free energies have been trained on the PDBBind dataset. However, it is controversial as to whether new SFs are actually improving since the general, refined, and core datasets of PDBBind are cross-contaminated with proteins and ligands with high similarity, and hence they may not perform comparably well in binding prediction of new protein-ligand complexes. In this work we have carefully prepared a cleaned PDBBind data set of non-covalent binders that are split into training, validation, and test datasets to control for data leakage, defined as proteins and ligands with high sequence and structural similarity. The resulting leak-proof (LP)-PDBBind data is used to retrain four popular SFs: AutoDock Vina, Random Forest (RF)-Score, InteractionGraphNet (IGN), and DeepDTA, to better test their capabilities when applied to new protein-ligand complexes. In particular we have formulated a new independent data set, BDB2020+, by matching high quality binding free energies from BindingDB with co-crystalized ligand-protein complexes from the PDB that have been deposited since 2020. Based on all the benchmark results, the retrained models using LP-PDBBind consistently perform better, with IGN especially being recommended for scoring and ranking applications for new protein-ligand systems.

Clinical trials are often hindered by the challenge of patient recruitment. In this work, we introduce TrialGPT, a first-of-its-kind large language model (LLM) framework to assist patient-to-trial matching. Given a patient note, TrialGPT predicts the patient's eligibility on a criterion-by-criterion basis and then consolidates these predictions to assess the patient's eligibility for the target trial. We evaluate the trial-level prediction performance of TrialGPT on three publicly available cohorts of 184 patients with over 18,000 trial annotations. We also engaged three physicians to label over 1,000 patient-criterion pairs to assess its criterion-level prediction accuracy. Experimental results show that TrialGPT achieves a criterion-level accuracy of 87.3% with faithful explanations, close to the expert performance (88.7%-90.0%). The aggregated TrialGPT scores are highly correlated with human eligibility judgments, and they outperform the best-competing models by 32.6% to 57.2% in ranking and excluding clinical trials. Furthermore, our user study reveals that TrialGPT can significantly reduce the screening time (by 42.6%) in a real-life clinical trial matching task. These results and analyses have demonstrated promising opportunities for clinical trial matching with LLMs such as TrialGPT.

Animal behavior is shaped by a myriad of mechanisms acting on a wide range of scales, which hampers quantitative reasoning and the identification of general principles. Here, we combine data analysis and theory to investigate the relationship between behavioral plasticity and heavy-tailed statistics often observed in animal behavior. Specifically, we first leverage high-resolution recordings of C. elegans locomotion to show that stochastic transitions among long-lived behaviors exhibit heavy-tailed first passage time distributions and correlation functions. Such heavy tails can be explained by slow adaptation of behavior over time. This particular result motivates our second step of introducing a general model where we separate fast dynamics on a quasi-stationary multi-well potential, from non-ergodic, slowly varying modes. We then show that heavy tails generically emerge in such a model, and we provide a theoretical derivation of the resulting functional form, which can become a power law with exponents that depend on the strength of the fluctuations. Finally, we provide direct support for the generality of our findings by testing them in a C. elegans mutant where adaptation is suppressed and heavy tails thus disappear, and recordings of larval zebrafish swimming behavior where heavy tails are again prevalent.

Tuberculosis (TB) is the world's deadliest infectious disease, with over 1.5 million deaths annually and 10 million new cases reported each year¹. The causative organism, Mycobacterium tuberculosis (Mtb) can take nearly 40 days to culture^2,3, a required step to determine the pathogen's antibiotic susceptibility. Both rapid identification of Mtb and rapid antibiotic susceptibility testing (AST) are essential for effective patient treatment and combating antimicrobial resistance. Here, we demonstrate a rapid, culture-free, and antibiotic incubation-free drug susceptibility test for TB using Raman spectroscopy and machine learning. We collect few-to-single-cell Raman spectra from over 25,000 cells of the MtB complex strain Bacillus Calmette-Guérin (BCG) resistant to one of the four mainstay anti-TB drugs, isoniazid, rifampicin, moxifloxacin and amikacin, as well as a pan-susceptible wildtype strain. By training a neural network on this data, we classify the antibiotic resistance profile of each strain, both on dried samples and in patient sputum samples. On dried samples, we achieve >98% resistant versus susceptible classification accuracy across all 5 BCG strains. In patient sputum samples, we achieve ~79% average classification accuracy. We develop a feature recognition algorithm in order to verify that our machine learning model is using biologically relevant spectral features to assess the resistance profiles of our mycobacterial strains. Finally, we demonstrate how this approach can be deployed in resource-limited settings by developing a low-cost, portable Raman microscope that costs <$5000. We show how this instrument and our machine learning model enables combined microscopy and spectroscopy for accurate few-to-single-cell drug susceptibility testing of BCG.

Since the inception of magnetization transfer (MT) imaging, it has been widely assumed that Henkelman's two spin pools have similar longitudinal relaxation times, which motivated many researchers to constrain them to each other. However, several recent publications reported a $T_1^s$ of the semi-solid spin pool that is much shorter than $T_1^f$ of the free pool. While these studies tailored experiments for robust proofs-of-concept, we here aim to quantify the disentangled relaxation processes on a voxel-by-voxel basis in a clinical imaging setting, i.e., with an effective resolution of 1.24mm isotropic and full brain coverage in 12min. To this end, we optimized a hybrid-state pulse sequence for mapping the parameters of an unconstrained MT model. We scanned four people with relapsing-remitting multiple sclerosis (MS) and four healthy controls with this pulse sequence and estimated $T_1^f\approx 1.84s$ and $T_1^s\approx 0.34s$ in healthy white matter. Our results confirm the reports that $T_1^s\ll T_1^f$ and we argue that this finding identifies MT as an inherent driver of longitudinal relaxation in brain tissue. Moreover, we estimated a fractional size of the semi-solid spin pool of $m_0^s\approx 0.212$, which is larger than previously assumed. An analysis of $T_1^f$ in normal-appearing white matter revealed statistically significant differences between individuals with MS and controls.