Real-time simulation and the integration of digital twins necessitate the development of computationally efficient surrogate models for contemporary food drying systems, in which complex physical processes must interact dynamically with virtual representations. Traditional modeling approaches for food drying are subject to notable limitations: empirical models often lack physical fidelity, whereas high-fidelity computational fluid dynamics (CFD) simulations require the resolution of tightly coupled heat and mass transfer processes among vapor, liquid water, and air phases within unsaturated porous media, rendering them computationally prohibitive for real-time applications. This work proposes a region-aware graph neural network (GNN) surrogate model for food drying simulation by leveraging mesh-based and region-specific representations of the drying process. The proposed GNN architecture, inspired by MeshGraphNet with specialized masked message passing, handles the heterogeneous domains and sharp discontinuities characteristic of coupled heat and mass transfer in food drying. The GNN model was trained on a limited dataset comprising 28 CFD simulations of hot-air drying of a potato slice under varying ambient conditions, generated using a CFD framework validated against numerical results from the literature and laboratory-scale experimental data. Despite this small training set, the model delivers predictive performance for operating conditions beyond those seen during training. Single-step predictions achieved RMSE values of 0.17 °C for temperature ( T ), 0.26% for relative humidity ( RH ), and 0.0028 for liquid saturation ( S w ). Comparative analysis with standard message passing revealed a threefold reduction in interfacial absolute error (0.003 vs. 0.008) and elimination of nonphysical field propagation across domain boundaries. The model maintained performance on the training domain edges, with moisture ratio curves achieving R 2 > 0.97 for 10-step rollouts (R 2 = 0.989 interior domain; R 2 = 0.979 edge domain) while preserving the characteristic S-shaped drying curves in longer autonomous simulations (R 2 = 0.784 and 0.873 for full rollouts of the interior and edge-of-domain test cases, respectively). This framework is applicable to real-time process control, design optimization, and digital twin implementations in food manufacturing. • Integrated a region-aware GNN into a digital twin for potato drying. • Achieved high accuracy with minimal data in heat and mass transfer prediction. • Region-aware message passing ensured physical consistency across domains. • Model adapted to changes in temperature and geometry with low prediction error. • Fast predictions enable real-time control and food process optimization

Optimal control of inter-population disease spread via reaction-diffusion models.

Ensemble threshold Boolean modeling reveals robust attractors and regulatory drivers in pediatric leukemia.

Mortality among critically ill COVID-19 patients has varied globally. In Sweden, geographic differences in mortality have also been observed. The current study aimed to determine whether mortality differences persist after adjusting for differences in case-mix, and to identify potential independent factors contributing to regional variations in mortality. We conducted a multicenter cohort study including adult patients admitted to seven hospital ICUs across three Swedish healthcare counties between March 1, 2020 and July 31, 2021. These ICUs include one university hospital, three county hospitals and three local hospitals, and cover the intensive care infrastructure for approximately one million inhabitants. Patients were assigned to the hospital of initial ICU admission, even if transferred later during the course. Patient characteristics, disease severity, respiratory support, and treatments were registered. Primary outcome was 90-day mortality. A mixed-effects Cox proportional hazards model was used. Seven hundred and forty seven patients were included. The unadjusted 90-day mortality varied significantly, with the highest rate at 30%, and the lowest at 8.5% (p < 0.001). After adjustment for baseline confounders (Charlson comorbidity index, sex, SAPS3, age, smoking status, BMI), calendar time and healthcare county (random intercept), all hospitals were significantly associated with increased 90-day mortality compared with the lowest mortality hospital. Hazard ratios ranged from 2.38 to 5.06. Among patients admitted to ICU due to COVID-19, we observed a difference in mortality related to the hospital of first ICU admission. This difference persisted after adjustment for calendar time, baseline confounders, and healthcare county. Potential explanations are lacking within the current study. Future studies should focus on comprehensive evaluation of both organizational and contextual determinants of mortality. This analysis from 3 Swedish counties (7 hospitals) for COVID ICU cases presents factors and relations to mortality risk, including factors for first admission to university-larger-, or smaller hospital. An association was observed for higher risk if the first ICU admission was in a smaller hospital, though recognizing that this is a dataset coming from a small set of hospitals.

Person-centered care planning (PCCP) involves active collaboration between people seeking care, clinical teams, and others to co-create longitudinal treatment plans. It is a crucial part of care for people with multiple chronic conditions (MCCs) and other complex needs. Despite the widespread acceptance of the concept, the use of PCCP in the US is variable. Many effective models of PCCP exist but uptake has been limited. We sought to identify and assess current models and approaches for PCCP by performing a multi-component environmental scan. We conducted targeted literature reviews based on streamlined systematic review methods and qualitative interviews with key informants, identified for their relevant expertise and frontline knowledge of PCCP models. In all, 966 abstracts and 187 full articles met review criteria. We identified 40 models with elements fitting into 7 categories of how they differed from usual care (adding People/Roles, innovative Payment/Incentives, novel Technology, decision support Tools, Services, Functions, and Changing Focus). Most were multi-component models with evidence of effectiveness in outcomes that address the quintuple aim. Barriers to adoption and integration across sectors were primarily lack of time and appropriate payment mechanisms. A small set of measures for PCCP was identified that focused primarily on patient experience and goal setting. Similarly, a small number of PCCP models connected with social services organizations to address health-related social needs and highlighted challenges with data sharing and payment as well as growing collaborations in community care hubs to address social needs. KI interviews echoed and deepened our understanding of these findings by contextualizing experiences with barriers and factors facilitating the delivery of PCCP. PCCP is a key aspect of high-quality care for people with MCCs; uptake is limited with barriers related to time and resources. Facilitators include alignment with healthcare system objectives and practices, and ready-to-deploy resources that enable coordinated PCCP care across providers. To promote greater adoption of PCCP concerted efforts are needed to share implementation details, success stories, and build the case for PCCP with patients, their families, and healthcare system leaders.

We propose an efficient method to compute a small set of integer-constrained cone singularities, which induce a rotationally seamless conformal parameterization with low distortion. Since the problem only involves discrete variables, i.e., vertex-constrained positions, integer-constrained angles, and the number of cones, we alternately optimize these three types of variables to achieve tractable convergence. Central to high efficiency is an explicit construction algorithm that reduces the optimization problem scale to be slightly greater than the number of integer variables for determining the optimal angles with fixed positions and numbers, even for high-genus surfaces. In addition, we derive a new derivative formula that allows us to move the cones, effectively reducing distortion until convergence. Combined with other strategies, including repositioning and adding cones to decrease distortion, adaptively selecting a constrained number of integer variables for efficient optimization, and pairing cones to reduce the number, we quickly achieve a favorable tradeoff between the number of cones and the parameterization distortion. We demonstrate the effectiveness and practicability of our cones by using them to generate rotationally seamless and low-distortion parameterizations on a massive test data set. Our method demonstrates an order-of-magnitude speedup (30× faster on average) compared to state-of-the-art approaches while maintaining comparable cone numbers and parameterization distortion.

We propose a novel method to generate a small set of ruled surfaces that do not collide with the input shape for linear hot-wire rough machining. Central to our technique is a new observation: ruled surfaces constructed by vertical extrusion from planar smooth curves that approach the input shape's outer contour lines without collisions can effectively remove material during rough machining. Accordingly, we develop an iterative algorithm that alternates in each iteration between computing a viewpoint to determine an outer contour line and optimizing a smooth curve to approximate that contour line under the collision-free constraint. Specifically, a view selection approach based on a genetic algorithm is used to optimize the viewpoint for removing materials as much as possible, and an adaptive fitting algorithm is presented to find the constrained curves. The feasibility and practicability of our method are demonstrated through 10 physical examples. Compared with manual designs, our method obtains lower errors with the same number of cuts.

Distant metastasis, predominantly to the liver, remains the leading cause of death in colorectal cancer (CRC), yet biomarkers that capture metastatic competence remain limited. Ferroptosis is an iron-dependent, lipid peroxidation-driven form of regulated cell death that can restrain tumor progression, but whether primary CRC from patients with liver metastasis shows ferroptosis-resistance-related features remains incompletely understood. In a small exploratory set of T-stage-matched primary CRC tumors with or without liver metastasis, we quantified glutathione redox and lipid peroxidation-related readouts and assessed SLC7A11 and GPX4 expression. We integrated GSE62321 transcriptomic profiles with a FerrDb ferroptosis gene set, evaluated prognosis in TCGA-COAD/READ, and performed genetic knockdown, MDA assays, C11-BODIPY lipid ROS staining, ferrostatin-1 rescue assays, and Transwell assays in CRC cell models. Primary tumors from patients with liver metastasis showed a more reduced redox profile and increased expression of core ferroptosis-suppressive proteins, consistent with enhanced ferroptosis resistance potential but not direct evidence of lower in vivo ferroptotic cell death. Integrative discovery highlighted fatty acid binding protein 4 (FABP4), α-synuclein (SNCA), and discoidin domain receptor 2 (DDR2) as CRC-LM-associated ferroptosis-related candidates. High expression of each gene was associated with unfavorable disease-free survival. In CRC cell models, including the lymph-node-metastasis-derived SW620 line and additional validation lines, silencing FABP4, SNCA, or DDR2 increased bulk MDA and/or C11-BODIPY-detected lipid ROS, altered ferroptosis susceptibility, and suppressed migratory and/or invasive phenotypes. Ferrostatin-1 partially rescued knockdown-induced viability loss, lipid ROS accumulation, and migratory/invasive defects, supporting involvement of ferroptosis-associated lipid peroxidation while not excluding broader stress-response mechanisms. FABP4, SNCA, and DDR2 are CRC-LM-associated ferroptosis-related candidates that modulate lipid peroxidation, ferroptosis susceptibility, and migratory/invasive phenotypes in CRC cell models, warranting further validation in viability-controlled and liver metastasis-specific models.

Heterogeneity of plasmids containing OXA-48-like and NDM-5 carbapenemases and emergence of OXA-181 and NDM-5 co-carrying strains and plasmids in Escherichia coli from veterinary settings.

Tensor-based multi-view clustering has been widely studied to capture high-order correlations among multiple views. Nevertheless, existing tensorial methods still exhibit several limitations. First, many approaches rely on full similarity graphs, leading to quadratic or cubic complexity in the number of samples and poor scalability. Second, view-specific anchor graphs are often tensorized without cross-view anchor alignment, yielding structurally inconsistent tensor representations and reduced cross-view comparability. Third, low-rank regularization is typically imposed via the tensor nuclear norm (TNN), which uniformly shrinks singular values and may bias the estimation of the intrinsic tensor rank. To this end, we propose a novel framework, named Align then Tensorize: Multi-level Consistent Anchor Graph Learning for Scalable Multi-View Clustering (ATTMVC). It adopts an anchor-based graph learning framework in which each view is reconstructed from a small set of anchors with sample-wise sparse noise, substantially reducing computational complexity. Unlike existing tensor-based methods that directly tensorize unaligned view-wise anchor graphs, ATTMVC first aligns view-specific anchor graphs into a shared latent space, thereby enforcing structural consistency across views and enabling more reliable modeling of cross-view higher-order correlations. Furthermore, we introduce a Threshold Tensor Rank (TTR) surrogate on the aligned anchor graph tensor, which effectively promotes low-rank structure while mitigating the over-shrinking effect commonly caused by TNN-based regularization. Finally, extensive experiments demonstrate that ATTMVC outperforms state-of-the-art multi-view clustering methods. The code is publicly available at https://github.com/tangchuan2000/ATTMVC.

Myeloproliferative neoplasms (MPNs) present a fundamental paradox: despite sharing a small set of canonical driver mutations in JAK2, CALR, or MPL, patients exhibit striking heterogeneity in disease latency, clinical presentation, and evolutionary trajectories to myelofibrosis or secondary acute myeloid leukemia. This review synthesizes recent advances in single-cell and spatial multi-omic technologies that are resolving this paradox by moving analysis from bulk averages to individual cells and their microenvironmental ecosystems. We examine how targeted single-cell DNA sequencing reconstructs clonal architectures and phylogenies, revealing that driver mutations arise within complex mosaics where mutation order, co-mutation context, and cellular ancestry determine phenotypic outcomes. Integrated single-cell transcriptomic and epigenomic profiling exposes within-clone heterogeneity, lineage biases, and functional states that explain variable penetrance and therapy responses. Spatial transcriptomics, especially when integrated with single-cell transcriptomics, histopathology, and multiplex proteomics, further demonstrates that malignant hematopoietic stem and progenitor cells actively remodel bone marrow niches, creating localized inflammatory and fibrotic microenvironments that select for aggressive subclones. Together, these approaches support a new ecological model of MPN pathogenesis in which early epigenetic hits create permissive stem cell reservoirs, clonal competition and cooperation shape disease progression, and non-cell-autonomous niche and immune signals drive malignant metamorphosis. We discuss how this framework refines prognostication, informs rational combination therapies targeting both malignant cells and their ecosystem, and enables real-time monitoring of clonal dynamics, ultimately charting a course from descriptive atlases to actionable clinical strategies.

The accurate characterization of nucleophilicity has attracted the attention of research groups worldwide for many years. So far, no universally applicable descriptor that correlates closely with experimental observations has been established. In this study, the behavior of the two conceptual DFT nucleophilicity indices is examined for a broad set of chemical compounds using semiempirical methods. The presented analysis reveals that the tested indices do not reliably capture experimental nucleophilicity trends when applied outside of small reference sets, leading to weak correlations with Mayr's reactivity scale. To address these shortcomings, an empirical nucleophilicity index, Nemp, is constructed through symbolic regression, incorporating frontier orbital energies, CDFT-based reactivity descriptors, atom-specific Fukui functions, local charge information, and solvent characteristics. Although the derived global model outperforms conventional descriptors (R2 = 0.737), substantially higher accuracy (R2 = 0.811) is obtained only when the models are further refined according to molecular scaffold. Collectively, these findings expose the inherent limitations of existing global CDFT nucleophilicity indices and demonstrate the potential of data-driven strategies to build context-aware reactivity descriptors.

Extreme cold snaps pose significant risks to buildings, infrastructure, energy systems, and occupants, yet standardized climatic datasets tailored for resilience-focused building performance modeling remain limited. This study presents a methodology and corresponding dataset of cold snap events for 217 U.S. cities, derived from 24 years of historical hourly temperature data obtained from the NASA POWER project. Cold snaps were detected using a percentile-based, location-specific threshold that identifies periods of "abnormal cold" with additional constraints to ensure that events reflect meaningful differences from local winter conditions. Each event was characterized using a suite of metrics, including event duration, heating degree hours, and overcooling degree. Events were further classified into four categories based on the mean outdoor air dry-bulb temperature, analogous to intensity scales used in other hazard domains. A selection procedure was applied to ensure that each city is represented by a small set of short, medium, and long-duration events, resulting in a curated dataset of 880 cold snaps suitable for building energy simulations and resilience assessments. The dataset is provided as EnergyPlus Weather (EPW) files accompanied by a summary spreadsheet containing all events and their metrics. This dataset supports the systematic evaluation of building performance under extreme cold weather conditions and provides a foundation for thermal and energy resilience modeling across the U.S. climates.

We study normal-only few-shot medical image anomaly detection, where only a small support set of normal images is available for adaptation and no real abnormal samples are used during training. While recent vision-language models (VLMs) such as MediCLIP show promise through synthetic anomaly generation and adapter mechanisms, their linear projection architectures struggle to capture fine-grained spatial features, and static learnable prompts lack adaptivity to heterogeneous lesion patterns. We introduce SpatioPrompt, a parameter-efficient VLM adaptation framework that addresses these limitations through two complementary innovations. First, we incorporate a spatial attention mechanism inspired by CBAM to explicitly model local region dependencies, enhancing lesion-focused representations while suppressing background interference. Second, we propose a FewShotEnhancedRWKV decoder that fuses GRU-style gating with RWKV temporal recurrence, enabling dynamic generation of image-conditioned prompts. Experiments across three medical imaging modalities demonstrate substantial improvements over MediCLIP: 6.6% gain on BrainMRI (99.9% vs. 93.3% AUROC at 8-shot) and 3.9% gain on BUSI (92.0% vs. 88.1% at 4-shot), with consistent improvements on CheXpert (73.8% vs. 70.7% at 16/32-shot). These results show that combining spatial attention with dynamic prompt generation can improve normal-only few-shot medical anomaly detection.

Building good machine-learning (ML) models to predict the bioactivity of novel chemical matter remains a challenging task. Accurate models require a training set with a large number of diverse compounds and a low level of noise. When extracting data from public databases such as ChEMBL, different levels of curation rigor may be applied, resulting in training sets of varying size, diversity, and, presumably, noise levels. It is not possible to know a priori whether increasing the size of the data set at the cost of adding more noise improves model generalization. To assess this trade-off, we compare three data curation and modeling approaches: (1) models trained on data for a single target, (2) models trained on target-specific data further restricted to a single set of assay conditions, and (3) multitask learning (MTL) models where each assay condition is treated as a separate task. This MTL approach was designed to bridge the gap between data quantity and quality. Graph neural networks (GNN) and random forests (RF) regressors are evaluated via a leave-assay-out strategy to minimize noise in the test sets. Our results show no meaningful performance differences between these curation strategies, suggesting that for lead-optimization tasks, increasing data quantity at the expense of label consistency does not improve generalization. Notably, the MTL approach also failed to provide a performance advantage. Additionally, we find that GNNs exhibit high seed-dependent variability in connection with the comparatively small training sets common for bioactivity measurements, highlighting the necessity of multiseed evaluation for robust model assessment.

Unitary Coupled Cluster (UCC) theory is a promising variational method for electronic structure calculations, particularly for systems that exhibit strong electronic correlation and for implementation on quantum computers. However, its practical application is limited to small chemical systems with small basis sets due to its steep computational scaling, which results from its nonterminating Baker-Campbell-Hausdorff expansion. Here, we introduce an active space UCCSD(4)/MP2 approach that leverages a fourth-order many-body perturbation theory truncation of UCCSD within a selected active space while treating external excitations at the MP2 level. We explore two variants: a composite method that sums separate internal and external contributions and an interacting method that couples the amplitudes for potentially greater accuracy. We test our approach on a range of systems, including molecules from the GW100 data set in their equilibrium geometries, a moderately correlated metaphosphate hydrolysis reaction, and the strongly correlated torsion of ethylene. Our results suggest that the interacting method with canonical orbitals is robust and stable for both weakly and moderately correlated systems and accurately reproduces the full UCCSD(4) potential energy curves, including only 15-25% of the virtual orbitals in its active space. In comparison, the composite formulation exhibits greater sensitivity to the choice of orbital basis and active space size, leading to less systematic behavior across the benchmark set. For ethylene torsion, a system dominated by strong static correlation, both the composite and interacting formulations employing canonical orbitals closely track the full UCCSD(4) reference while preserving the qualitative behavior of the parent method, including the breakdown in strongly multireference regimes. This active space framework offers a computationally tractable approach for modeling correlated molecules and reactions on classical computers and provides a viable path for scaling UCC calculations for resource-constrained quantum hardware.

Chikungunya fever (CHIKF) and dengue fever are mosquito-borne viral diseases. These infections often circulate in the same regions at the same time. Early symptoms can look very similar between the two diseases. This overlap makes early and accurate diagnosis difficult. Many endemic clinics do not have easy access to molecular tests such as RT-PCR [1, 2]. Clinicians therefore need other practical tools for early decision-making. In this study, we aimed to develop an interpretable machine learning model. The model uses routine clinical signs and standard laboratory results. We also aimed to validate the model for differentiation of CHIKF from dengue fever at initial hospital-based assessment. This retrospective observational study analyzed 1,058 laboratory-confirmed arboviral infections, including 366 patients with CHIKF and 692 patients with dengue fever. The dataset was stratified by diagnosis and randomly divided into a training set (n = 742) and a held-out test set (n = 316) at a 7:3 ratio. The team collected clinical symptoms, complete blood count (CBC) results, and inflammatory marker data. The team then used these variables to build eight machine learning models. The study evaluated model performance with discrimination metrics. The study also assessed calibration. The study further used decision curve analysis to estimate clinical usefulness. The team examined feature importance with Shapley Additive Explanations (SHAP). The team deployed the best-performing model as a web-based clinical decision-support tool. Among the tested approaches, the gradient boosting model (GBM) showed the best and most consistent performance. The GBM achieved a high area under the ROC curve (AUC) in both the training and test sets. The GBM also delivered strong sensitivity and specificity across both cohorts. The SHAP analysis repeatedly highlighted platelet count (PLT) and rash as the most important predictors of CHIKF. These findings match well with known clinical patterns. The online deployment integrated the final model into a simple platform. The platform provides automated, real-time risk estimates using only a small set of routinely available variables. This study shows that interpretable machine learning models can help clinicians distinguish CHIKF from dengue fever early. The models rely on routine clinical information and standard laboratory tests. These inputs are widely available in many settings. The study also presents a web-based tool that applies the best model at the bedside. The tool may be especially useful in resource-limited clinics. However, the tool still needs external validation. Future studies should test the model in multicenter, prospective cohorts.

In recent years, there has been a surge in artificial intelligence (AI)-based drug-protein (or pesticide-protein) affinity (DPA) prediction tools. The field has been primarily driven by evolving deep-learning architectures and increasingly complex representations, leading to a growing demand for training data sourced from limited experimental data. In this work, we present DeepDOX1, a dual-drive DPA prediction tool featuring the tight integration of a concise AI architecture and an interpretable, quantum chemistry-based representation of protein-ligand interactions. To be more specific, the first-principles quantum chemistry-generated features incorporating the interactions between the drug and the protein binding pocket allow a relatively simple convolutional neural network (CNN) model trained on a relatively small training set (9,938 binders) to exhibit exceptional generalization capabilities across extensive testing involving 1,281 binders. Notably, DeepDOX1 outperforms popular AI-based DPA prediction methods in the tests simulating real-world hit-to-lead optimization (HLO) scenarios and a highly challenging test set featuring covalent ligands, halogenated ligands, and metalloproteins, even though its training set does not contain any covalent ligands. To further validate its practical utility, we designed a series of novel covalent inhibitors targeting the diabetes target molecule Hu-FBPase using DeepDOX1. Subsequent experimental validation, including enzyme-level bioactivity assays and crystal structure determination, revealed strengthened activity of the newly designed compound and confirmed DeepDOX1's effectiveness in real-world drug design applications. It is conceivable that the combination of deep-learning architecture and first-principles quantum chemistry might be one of the next breakthroughs in DPA prediction.

The uniformity of tobacco blend mixing is a critical factor influencing cigarette quality. However, due to the dynamic nature of the blending process and the complex characteristics of the materials involved, current detection methods primarily rely on manual sampling or offline analysis. These approaches suffer from poor real-time performance and lack representativeness. To address these limitations, this study explores methods for generating tobacco-leaf image data and proposes a real-time semantic segmentation technique for tobacco-leaf images based on U-Net. First, a small set of tobacco leaf images was acquired to establish a raw dataset of composite tobacco leaf images. Three generative adversarial networks-CycleGAN, WGAN, and DCGAN-were employed to generate tobacco leaf images from this dataset. Based on image quality, the optimal image generation network was selected. Experimental results showed that CycleGAN produced the highest-quality tobacco leaf images. Next, the Squeeze Excitation Attention (SE) mechanism was integrated into the VGG16 backbone network. This enhancement improved the network's ability to extract features from various types of tobacco leaf images while suppressing interference from irrelevant pixels. The results demonstrated that, compared to mainstream models such as Segformer, DeepLabV3, PSPNet, and the unmodified U-Net, the proposed SE-UNet model achieved superior segmentation accuracy, excellent real-time performance, and overall optimal segmentation capabilities. Compared to Segformer, DeepLabV3, PSPNet, and the original model, the MIOU in the evaluation metrics improved by 17.46, 8.7, 13.39, and 2.77 percentage points, respectively. Finally, the pixel areas of different tobacco leaf types were extracted and calculated from the segmented images to determine the proportion of each leaf type within the images. This research offers a novel approach for practical tobacco production and quality inspection, while also providing a new pathway for real-time online detection of other agricultural products.

Organic semiconductors require both high carrier mobility and structural diversity, but direct first-principles evaluation is costly and brute-force exploration of chemical space is infeasible. We propose a data-driven framework that combines a hierarchical variational autoencoder (HVAE), Gaussian mixture regression (GMR), and Bayesian optimization to design small molecules exhibiting low reorganization energy and high carrier mobility. An HVAE was trained to learn a latent representation of organic semiconductor-like molecules, and GMR linked latent variables to hole and electron reorganization energies. By (i) adding random noise around low-reorganization energy molecules in the latent space and (ii) sampling from a Gaussian mixture model fitted to their latent distribution, we generated structurally reasonable candidates under constraints on ring number and molecular size, including a molecule with a new minimum hole reorganization energy obtained via sulfur-to-nitrogen substitution. Subsequent Bayesian optimization using reorganization energy and a small set of structural descriptors identified molecules with high mobilities and confirmed reorganization energy as a key descriptor for both hole and electron mobility.