Search

BackgroundOxygen saturation (Spo2) screening has not led to earlier detection of critical congenital heart disease (CCHD). Adding pulse oximetry features (ie, perfusion data and radiofemoral pulse delay) may improve CCHD detection, especially coarctation of the aorta (CoA). We developed and tested a machine learning (ML) pulse oximetry algorithm to enhance CCHD detection.Methods and resultsSix sites prospectively enrolled newborns with and without CCHD and recorded simultaneous pre- and postductal pulse oximetry. We focused on models at 1 versus 2 time points and with/without pulse delay for our ML algorithms. The sensitivity, specificity, and area under the receiver operating characteristic curve were compared between the Spo2-alone and ML algorithms. A total of 523 newborns were enrolled (no CHD, 317; CHD, 74; CCHD, 132, of whom 21 had isolated CoA). When applying the Spo2-alone algorithm to all patients, 26.2% of CCHD would be missed. We narrowed the sample to patients with both 2 time point measurements and pulse-delay data (no CHD, 65; CCHD, 14) to compare ML performance. Among these patients, sensitivity for CCHD detection increased with both the addition of pulse delay and a second time point. All ML models had 100% specificity. With a 2-time-points+pulse-delay model, CCHD sensitivity increased to 92.86% (P=0.25) compared with Spo2 alone (71.43%), and CoA increased to 66.67% (P=0.5) from 0. The area under the receiver operating characteristic curve for CCHD and CoA detection significantly improved (0.96 versus 0.83 for CCHD, 0.83 versus 0.48 for CoA; both P=0.03) using the 2-time-points+pulse-delay model compared with Spo2 alone.ConclusionsML pulse oximetry that combines oxygenation, perfusion data, and pulse delay at 2 time points may improve detection of CCHD and CoA within 48 hours after birth.RegistrationURL: https://www.clinicaltrials.gov/study/NCT04056104?term=NCT04056104&rank=1; Unique identifier: NCT04056104.

UNLABELLED: Invasive mold infections (IMIs) are associated with high morbidity, particularly in immunocompromised patients, with mortality rates between 40% and 80%. Early initiation of appropriate antifungal therapy can substantially improve outcomes, yet early diagnosis remains difficult to establish and often requires multidisciplinary teams evaluating clinical and radiological findings plus supportive mycological findings. Universal digital high-resolution melting (U-dHRM) analysis may enable rapid and robust diagnoses of IMI. A universal fungal assay was developed for U-dHRM and used to generate a database of melt curve signatures for 19 clinically relevant fungal pathogens. A machine learning algorithm (ML) was trained to automatically classify these pathogen curves and detect novel melt curves. Performance was assessed on 73 clinical bronchoalveolar lavage samples from patients suspected of IMI. Novel curves were identified by micropipetting U-dHRM reactions and Sanger sequencing amplicons. U-dHRM achieved 97% overall fungal organism identification accuracy and a turnaround time of ~4 hrs. U-dHRM detected pathogenic molds (Aspergillus, Mucorales, Lomentospora, and Fusarium) in 73% of 30 samples classified as IMI, including mixed infections. Specificity was optimized by requiring the number of pathogenic mold curves detected in a sample to be >8 and a sample volume to be 1 mL, which resulted in 100% specificity in 21 at-risk patients without IMI. U-dHRM showed promise as a separate or combination diagnostic approach to standard mycological tests. U-dHRMs speed, ability to simultaneously identify and quantify clinically relevant mold pathogens in polymicrobial samples, and detect emerging opportunistic pathogens may aid treatment decisions, improving patient outcomes. IMPORTANCE: Improvements in diagnostics for invasive mold infections are urgently needed. This work presents a new molecular detection approach that addresses technical and workflow challenges to provide fast pathogen detection, identification, and quantification that could inform treatment to improve patient outcomes.

Tuberculosis (TB) remains a leading cause of death from an infectious disease worldwide, partly due to a lack of effective strategies to screen and triage individuals with potential TB. Whole blood RNA signatures have been tested as biomarkers for TB, but have failed to meet the World Health Organizations (WHO) optimal target product profiles (TPP). Here, we use RNA sequencing and machine-learning to investigate the utility of plasma cell-free RNA (cfRNA) as a host-response biomarker for TB in cohorts from Uganda, Vietnam and Philippines. We report a 6-gene cfRNA signature, which differentiates TB-positive and TB-negative individuals with AUC = 0.95, 0.92, and 0.95 in test, training and validation, respectively. This signature meets WHO TPPs (sensitivity: 97.1% [95% CI: 80.9-100%], specificity: 85.2% [95% CI: 72.4-100%]) regardless of geographic location, sample collection method and HIV status. Overall, our results identify plasma cfRNA as a promising host response biomarker to diagnose TB.

MOTIVATION: Recent advances in spatial transcriptomics allow spatially resolved gene expression measurements with cellular or even sub-cellular resolution, directly characterizing the complex spatiotemporal gene expression landscape and cell-to-cell interactions in their native microenvironments. Due to technology limitations, most spatial transcriptomic technologies still yield incomplete expression measurements with excessive missing values. Therefore, gene imputation is critical to filling in missing data, enhancing resolution, and improving overall interpretability. However, existing methods either require additional matched single-cell RNA-seq data, which is rarely available, or ignore spatial proximity or expression similarity information. RESULTS: To address these issues, we introduce Impeller, a path-based heterogeneous graph learning method for spatial transcriptomic data imputation. Impeller has two unique characteristics distinct from existing approaches. First, it builds a heterogeneous graph with two types of edges representing spatial proximity and expression similarity. Therefore, Impeller can simultaneously model smooth gene expression changes across spatial dimensions and capture similar gene expression signatures of faraway cells from the same type. Moreover, Impeller incorporates both short- and long-range cell-to-cell interactions (e.g. via paracrine and endocrine) by stacking multiple GNN layers. We use a learnable path operator in Impeller to avoid the over-smoothing issue of the traditional Laplacian matrices. Extensive experiments on diverse datasets from three popular platforms and two species demonstrate the superiority of Impeller over various state-of-the-art imputation methods. AVAILABILITY AND IMPLEMENTATION: The code and preprocessed data used in this study are available at https://github.com/aicb-ZhangLabs/Impeller and https://zenodo.org/records/11212604.

MOTIVATION: Phylogenetic placement of a query sequence on a backbone tree is increasingly used across biomedical sciences to identify the content of a sample from its DNA content. The accuracy of such analyses depends on the density of the backbone tree, making it crucial that placement methods scale to very large trees. Moreover, a new paradigm has been recently proposed to place sequences on the species tree using single-gene data. The goal is to better characterize the samples and to enable combined analyses of marker-gene (e.g., 16S rRNA gene amplicon) and genome-wide data. The recent method DEPP enables performing such analyses using metric learning. However, metric learning is hampered by a need to compute and save a quadratically growing matrix of pairwise distances during training. Thus, the training phase of DEPP does not scale to more than roughly 10 000 backbone species, a problem that we faced when trying to use our recently released Greengenes2 (GG2) reference tree containing 331 270 species. RESULTS: This paper explores divide-and-conquer for training ensembles of DEPP models, culminating in a method called C-DEPP. While divide-and-conquer has been extensively used in phylogenetics, applying divide-and-conquer to data-hungry machine-learning methods needs nuance. C-DEPP uses carefully crafted techniques to enable quasi-linear scaling while maintaining accuracy. C-DEPP enables placing 20 million 16S fragments on the GG2 reference tree in 41 h of computation. AVAILABILITY AND IMPLEMENTATION: The dataset and C-DEPP software are freely available at https://github.com/yueyujiang/dataset_cdepp/.

BACKGROUND: Language in nonmedical data sets is known to transmit human-like biases when used in natural language processing (NLP) algorithms that can reinforce disparities. It is unclear if NLP algorithms of medical notes could lead to similar transmissions of biases. RESEARCH QUESTION: Can we identify implicit bias in clinical notes, and are biases stable across time and geography? STUDY DESIGN AND METHODS: To determine whether different racial and ethnic descriptors are similar contextually to stigmatizing language in ICU notes and whether these relationships are stable across time and geography, we identified notes on critically ill adults admitted to the University of California, San Francisco (UCSF), from 2012 through 2022 and to Beth Israel Deaconess Hospital (BIDMC) from 2001 through 2012. Because word meaning is derived largely from context, we trained unsupervised word-embedding algorithms to measure the similarity (cosine similarity) quantitatively of the context between a racial or ethnic descriptor (eg, African-American) and a stigmatizing target word (eg, nonco-operative) or group of words (violence, passivity, noncompliance, nonadherence). RESULTS: In UCSF notes, Black descriptors were less likely to be similar contextually to violent words compared with White descriptors. Contrastingly, in BIDMC notes, Black descriptors were more likely to be similar contextually to violent words compared with White descriptors. The UCSF data set also showed that Black descriptors were more similar contextually to passivity and noncompliance words compared with Latinx descriptors. INTERPRETATION: Implicit bias is identifiable in ICU notes. Racial and ethnic group descriptors carry different contextual relationships to stigmatizing words, depending on when and where notes were written. Because NLP models seem able to transmit implicit bias from training data, use of NLP algorithms in clinical prediction could reinforce disparities. Active debiasing strategies may be necessary to achieve algorithmic fairness when using language models in clinical research.

Land-use control is local and highly varied. State agencies struggle to assess plan contents. Similarly, advocacy groups and planning researchers wrestle with the length of planning documents and ability to compare across plans. The goal of this research is to (1) introduce Natural Language Processing techniques that can automate qualitative coding in planning research and (2) provide policy-relevant exploratory findings. We assembled a database of 461 California city-level General Plans, extracted the text, and used topic modeling to identify areas of emphasis (clusters of co-occurring words). We find that California city general plans address more than sixty topics, including greenhouse gas mitigation and Climate Action Planning. Through spatializing results, we find that a quarter of the topics in plans are regionally specific. We also quantify the rift and convergence of planning topics. The topics focused on housing have very little overlap with other planning topics. This is likely a factor of state requirements to update and evolve the Housing Elements every five years, but not other aspects of General Plans. This finding has policy implications as housing topics evolve away from other emphasis areas such as transportation and economic development. Furthermore, the topic modeling approach reveals that many cities have had a focus on environmental justice through Health and Wellness Elements well before the state mandate in 2019. Our searchable state-level database of general plans is the first for California-and nationally. We provide a model for others that wish to comprehensively assess and compare plan contents using machine learning.

The utility of a quantum computer is highly dependent on the ability to reliably perform accurate quantum logic operations. For finding optimal control solutions, it is of particular interest to explore model-free approaches, since their quality is not constrained by the limited accuracy of theoretical models for the quantum processor—in contrast to many established gate implementation strategies. In this work, we utilize a continuous control reinforcement learning algorithm to design entangling two-qubit gates for superconducting qubits; specifically, our agent constructs cross-resonance and CNOT gates without any prior information about the physical system. Using a simulated environment of fixed-frequency fixed-coupling transmon qubits, we demonstrate the capability to generate novel pulse sequences that outperform the standard cross-resonance gates in both fidelity and gate duration, while maintaining a comparable susceptibility to stochastic unitary noise. We further showcase an augmentation in training and input information that allows our agent to adapt its pulse design abilities to drifting hardware characteristics, importantly, with little to no additional optimization. Our results exhibit clearly the advantages of unbiased adaptive-feedback learning-based optimization methods for transmon gate design.

Abstract: Active learning in semi-supervised classification involves introducing additional labels for unlabelled data to improve the accuracy of the underlying classifier. A challenge is to identify which points to label to best improve performance while limiting the number of new labels. “Model Change” active learning quantifies the resulting change incurred in the classifier by introducing the additional label(s). We pair this idea with graph-based semi-supervised learning (SSL) methods, that use the spectrum of the graph Laplacian matrix, which can be truncated to avoid prohibitively large computational and storage costs. We consider a family of convex loss functions for which the acquisition function can be efficiently approximated using the Laplace approximation of the posterior distribution. We show a variety of multiclass examples that illustrate improved performance over prior state-of-art.

Immunohistochemistry (IHC) is used to guide treatment decisions in multiple cancer types. For treatment with checkpoint inhibitors, programmed death ligand 1 (PD-L1) IHC is used as a companion diagnostic. However, the scoring of PD-L1 is complicated by its expression in cancer and immune cells. Separation of cancer and noncancer regions is needed to calculate tumor proportion scores (TPS) of PD-L1, which is based on the percentage of PD-L1-positive cancer cells. Evaluation of PD-L1 expression requires highly experienced pathologists and is often challenging and time-consuming. Here, we used a multi-institutional cohort of 77 lung cancer cases stained centrally with the PD-L1 22C3 clone. We developed a 4-step pipeline for measuring TPS that includes the coregistration of hematoxylin and eosin, PD-L1, and negative control (NC) digital slides for exclusion of necrosis, segmentation of cancer regions, and quantification of PD-L1+ cells. As cancer segmentation is a challenging step for TPS generation, we trained DeepLab V3 in the Visiopharm software package to outline cancer regions in PD-L1 and NC images and evaluated the model performance by mean intersection over union (mIoU) against manual outlines. Only 14 cases were required to accomplish a mIoU of 0.82 for cancer segmentation in hematoxylin-stained NC cases. For PD-L1-stained slides, a model trained on PD-L1 tiles augmented by registered NC tiles achieved a mIoU of 0.79. In segmented cancer regions from whole slide images, the digital TPS achieved an accuracy of 75% against the manual TPS scores from the pathology report. Major reasons for algorithmic inaccuracies include the inclusion of immune cells in cancer outlines and poor nuclear segmentation of cancer cells. Our transparent and stepwise approach and performance metrics can be applied to any IHC assay to provide pathologists with important insights on when to apply and how to evaluate commercial automated IHC scoring systems.

BACKGROUND: Maternal cigarette smoking during pregnancy (MSDP) is associated with numerous adverse health outcomes in infants and children with potential lifelong consequences. Negative effects of MSDP on placental DNA methylation (DNAm), placental structure, and function are well established. OBJECTIVE: Our aim was to develop biomarkers of MSDP using DNAm measured in placentas (N=96), collected as part of the Vitamin C to Decrease the Effects of Smoking in Pregnancy on Infant Lung Function double-blind, placebo-controlled randomized clinical trial conducted between 2012 and 2016. We also aimed to develop a digital polymerase chain reaction (PCR) assay for the top ranking cytosine-guanine dinucleotide (CpG) so that large numbers of samples can be screened for exposure at low cost. METHODS: We compared the ability of four machine learning methods [logistic least absolute shrinkage and selection operator (LASSO) regression, logistic elastic net regression, random forest, and gradient boosting machine] to classify MSDP based on placental DNAm signatures. We developed separate models using the complete EPIC array dataset and on the subset of probes also found on the 450K array so that models exist for both platforms. For comparison, we developed a model using CpGs previously associated with MSDP in placenta. For each final model, we used model coefficients and normalized beta values to calculate placental smoking index (PSI) scores for each sample. Final models were validated in two external datasets: the Extremely Low Gestational Age Newborn observational study, N=426; and the Rhode Island Childrens Health Study, N=237. RESULTS: Logistic LASSO regression demonstrated the highest performance in cross-validation testing with the lowest number of input CpGs. Accuracy was greatest in external datasets when using models developed for the same platform. PSI scores in smokers only (n=72) were moderately correlated with maternal plasma cotinine levels. One CpG (cg27402634), with the largest coefficient in two models, was measured accurately by digital PCR compared with measurement by EPIC array (R2=0.98). DISCUSSION: To our knowledge, we have developed the first placental DNAm-based biomarkers of MSDP with broad utility to studies of prenatal disease origins. https://doi.org/10.1289/EHP13838.

Cell atlases serve as vital references for automating cell labeling in new samples, yet existing classification algorithms struggle with accuracy. Here we introduce SIMS (scalable, interpretable machine learning for single cell), a low-code data-efficient pipeline for single-cell RNA classification. We benchmark SIMS against datasets from different tissues and species. We demonstrate SIMS's efficacy in classifying cells in the brain, achieving high accuracy even with small training sets (

The demand for specialized hardware to train AI models has increased in tandem with the increase in the model complexity over the recent years. Graphics processing unit (GPU) is one such hardware that is capable of parallelizing operations performed on a large chunk of data. Companies like Nvidia, AMD, and Google have been constantly scaling-up the hardware performance as fast as they can. Nevertheless, there is still a gap between the required processing power and processing capacity of the hardware. To increase the hardware utilization, the software has to be optimized too. In this paper, we present some general GPU optimization techniques we used to efficiently train the optiGAN model, a Generative Adversarial Network that is capable of generating multidimensional probability distributions of optical photons at the photodetector face in radiation detectors, on an 8GB Nvidia Quadro RTX 4000 GPU. We analyze and compare the performances of all the optimizations based on the execution time and the memory consumed using the Nvidia Nsight Systems profiler tool. The optimizations gave approximately a 4.5x increase in the runtime performance when compared to a naive training on the GPU, without compromising the model performance. Finally we discuss optiGANs future work and how we are planning to scale the model on GPUs.

The efficacy of supervised deep learning in medical image analyses, particularly in pathology, is hindered by the necessity for extensive manual annotations. Annotating images at the gigapixel level manually proves to be a highly labor-intensive and time-consuming task. Semi-supervised learning (SSL) has emerged as a promising approach that leverages unlabeled data to reduce labeling efforts. In this work, we introduce Semi-Path, a practical SSL framework enhanced with active learning (AL) for gigapixel pathology tasks. Unlike existing methods that treat SSL and AL as independent components where AL incurs significant computational complexity to SSL, we propose a deep fusion of SSL and AL into a unified framework. Our framework introduces Informative Active Annotation (IAA) that employs a SSL-AL iterative structure to effectively extract knowledge from unlabeled pathology data. This structure significantly minimizes labeling efforts and computational complexity. Then, we propose Adaptive Pseudo-Labeling (APL) to address heterogeneity in class distribution, and prediction difficulty that are often observed in real-world pathology tasks. We evaluate Semi-Path on pathology image classification and segmentation tasks over three datasets that include WSIs from breast, colorectal, and brain tissues. The experimental results demonstrate the consistent superiority of Semi-Path over state-of-the-art methods.

We construct non-linear machine learning (ML) prediction models for systolic and diastolic blood pressure (SBP, DBP) using demographic and clinical variables and polygenic risk scores (PRSs). We developed a two-model ensemble, consisting of a baseline model, where prediction is based on demographic and clinical variables only, and a genetic model, where we also include PRSs. We evaluate the use of a linear versus a non-linear model at both the baseline and the genetic model levels and assess the improvement in performance when incorporating multiple PRSs. We report the ensemble models performance as percentage variance explained (PVE) on a held-out test dataset. A non-linear baseline model improved the PVEs from 28.1 to 30.1% (SBP) and 14.3% to 17.4% (DBP) compared with a linear baseline model. Including seven PRSs in the genetic model computed based on the largest available GWAS of SBP/DBP improved the genetic model PVE from 4.8 to 5.1% (SBP) and 4.7 to 5% (DBP) compared to using a single PRS. Adding additional 14 PRSs computed based on two independent GWASs further increased the genetic model PVE to 6.3% (SBP) and 5.7% (DBP). PVE differed across self-reported race/ethnicity groups, with primarily all non-White groups benefitting from the inclusion of additional PRSs. In summary, non-linear ML models improves BP prediction in models incorporating diverse populations.

SummaryThe increasing development of sequence-based machine learning models has raised the demand for manipulating sequences for this application. However, existing approaches to edit and evaluate genome sequences using models have limitations, such as incompatibility with structural variants, challenges in identifying responsible sequence perturbations, and the need for vcf file inputs and phased data. To address these bottlenecks, we present Sequence Mutator for Predictive Models (SuPreMo), a scalable and comprehensive tool for performing and supporting in silico mutagenesis experiments. We then demonstrate how pairs of reference and perturbed sequences can be used with machine learning models to prioritize pathogenic variants or discover new functional sequences.Availability and implementationSuPreMo was written in Python, and can be run using only one line of code to generate both sequences and 3D genome disruption scores. The codebase, instructions for installation and use, and tutorials are on the GitHub page: https://github.com/ketringjoni/SuPreMo.

Prompts to large language models (LLMs) have evolved beyond simple user questions. For LLMs to solve complex problems, today's practices are to include domain-specific instructions, illustration of tool usages, and long context such as textbook chapters in prompts. As such, many parts of prompts are repetitive across requests, and their attention computation results can be reused. However, today's LLM serving systems treat every request in isolation, missing the opportunity of computation reuse.This paper proposes Preble, the first distributed LLM serving platform that targets and optimizes for prompt sharing. We perform a study on five popular LLM workloads. Based on our study results, we designed a distributed scheduling system that co-optimizes computation reuse and load balancing. Our evaluation of Preble on two to 8 GPUs with real workloads and request arrival patterns on two open-source LLM models shows that Preble outperforms the state of the art avg latency by 1.5x to 14.5x and p99 by 2x to 10x.

The Snapchat ads dataset contains political ad data for ads on Snapchat, oneof the largest social media networks in the world. A key feature of the datasetis how much money an organization spends on a particular ad, found in the`Spend` column. It is reasonable to assume that this amount varies based oncertain factors, but can we use those factors to figure out how much is spenton an ad? We can explore this by predicting ad spending through machinelearning. After feature analysis and engineering, we arrive at a linearregression model with $R^2$ = .85 and perform a fairness evaluation of thealgorithm.

Surveillance for genetic variation of microbial pathogens, both within and among species, plays an important role in informing research, diagnostic, prevention, and treatment activities for disease control. However, large-scale systematic screening for novel genotypes remains challenging in part due to technological limitations. Towards addressing this challenge, we present an advancement in universal microbial high resolution melting (HRM) analysis that is capable of accomplishing both known genotype identification and novel genotype detection. Specifically, this novel surveillance functionality is achieved through time-series modeling of sequence-defined HRM curves, which is uniquely enabled by the large-scale melt curve datasets generated using our high-throughput digital HRM platform. Taking the detection of bacterial genotypes as a model application, we demonstrate that our algorithms accomplish an overall classification accuracy over 99.7% and perform novelty detection with a sensitivity of 0.96, specificity of 0.96 and Youden index of 0.92. Since HRM-based DNA profiling is an inexpensive and rapid technique, our results add support for the feasibility of its use in surveillance applications.

Machine learning (ML) methods are proliferating in scientific research. However, the adoption of these methods has been accompanied by failures of validity, reproducibility, and generalizability. These failures can hinder scientific progress, lead to false consensus around invalid claims, and undermine the credibility of ML-based science. ML methods are often applied and fail in similar ways across disciplines. Motivated by this observation, our goal is to provide clear recommendations for conducting and reporting ML-based science. Drawing from an extensive review of past literature, we present the REFORMS checklist (recommendations for machine-learning-based science). It consists of 32 questions and a paired set of guidelines. REFORMS was developed on the basis of a consensus of 19 researchers across computer science, data science, mathematics, social sciences, and biomedical sciences. REFORMS can serve as a resource for researchers when designing and implementing a study, for referees when reviewing papers, and for journals when enforcing standards for transparency and reproducibility.

Machine learning approaches using structural magnetic resonance imaging (sMRI) can be informative for disease classification, although their ability to predict psychosis is largely unknown. We created a model with individuals at CHR who developed psychosis later (CHR-PS+) from healthy controls (HCs) that can differentiate each other. We also evaluated whether we could distinguish CHR-PS+ individuals from those who did not develop psychosis later (CHR-PS-) and those with uncertain follow-up status (CHR-UNK). T1-weighted structural brain MRI scans from 1165 individuals at CHR (CHR-PS+, n = 144; CHR-PS-, n = 793; and CHR-UNK, n = 228), and 1029 HCs, were obtained from 21 sites. We used ComBat to harmonize measures of subcortical volume, cortical thickness and surface area data and corrected for non-linear effects of age and sex using a general additive model. CHR-PS+ (n = 120) and HC (n = 799) data from 20 sites served as a training dataset, which we used to build a classifier. The remaining samples were used external validation datasets to evaluate classifier performance (test, independent confirmatory, and independent group [CHR-PS- and CHR-UNK] datasets). The accuracy of the classifier on the training and independent confirmatory datasets was 85% and 73% respectively. Regional cortical surface area measures-including those from the right superior frontal, right superior temporal, and bilateral insular cortices strongly contributed to classifying CHR-PS+ from HC. CHR-PS- and CHR-UNK individuals were more likely to be classified as HC compared to CHR-PS+ (classification rate to HC: CHR-PS+, 30%; CHR-PS-, 73%; CHR-UNK, 80%). We used multisite sMRI to train a classifier to predict psychosis onset in CHR individuals, and it showed promise predicting CHR-PS+ in an independent sample. The results suggest that when considering adolescent brain development, baseline MRI scans for CHR individuals may be helpful to identify their prognosis. Future prospective studies are required about whether the classifier could be actually helpful in the clinical settings.

Deep learning is transforming bioimage analysis, but its application in single-cell segmentation is limited by the lack of large, diverse annotated datasets. We addressed this by introducing a CycleGAN-based architecture, cGAN-Seg, that enhances the training of cell segmentation models with limited annotated datasets. During training, cGAN-Seg generates annotated synthetic phase-contrast or fluorescent images with morphological details and nuances closely mimicking real images. This increases the variability seen by the segmentation model, enhancing the authenticity of synthetic samples and thereby improving predictive accuracy and generalization. Experimental results show that cGAN-Seg significantly improves the performance of widely used segmentation models over conventional training techniques. Our approach has the potential to accelerate the development of foundation models for microscopy image analysis, indicating its significance in advancing bioimage analysis with efficient training methodologies.

PURPOSE: The purpose of this review was to evaluate the effectiveness of patient-specific rods for adult spinal deformity. METHODS: A systematic review of the literature was performed through an electronic search of the PubMed, Scopus, and Web of Science databases. Human studies between 2012 and 2023 were included. Sample size, sagittal vertical axis (SVA), pelvic incidence-lumbar lordosis (PI-LL), pelvic tilt (PT), operation time, blood loss, follow-up duration, and complications were recorded for each study when available. RESULTS: Seven studies with a total of 304 adult spinal deformity patients of various etiologies were included. All studies reported SVA, and PT; two studies did not report PI-LL. Four studies reported planned radiographic outcomes. Two found a significant association between preoperative plan and postoperative outcome in all three outcomes. One found a significant association for PI-LL alone. The fourth found no significant associations. SVA improved in six of seven studies, PI-LL improved in all five, and three of seven studies found improved postoperative PT. Significance of these results varied greatly by study. CONCLUSION: Preliminary evidence suggests potential benefits of PSRs in achieving optimal spino-pelvic parameters in ASD surgery. Nevertheless, conclusions regarding the superiority of PSRs over traditional rods must be judiciously drawn, given the heterogeneity of patients and study methodologies, potential confounding variables, and the absence of robust randomized controlled trials. Future investigations should concentrate on enhancing preoperative planning, standardizing surgical methodologies, isolating specific patient subgroups, and head-to-head comparisons with traditional rods to fully elucidate the impact of PSRs in ASD surgery.

BackgroundWhole brain delineation (WBD) is utilized in neuroimaging analysis for data preprocessing and deriving whole brain image metrics. Current automated WBD techniques for analysis of preclinical brain MRI data show limited accuracy when images present with significant neuropathology and anatomical deformations, such as that resulting from organophosphate intoxication (OPI) and Alzheimer's Disease (AD), and inadequate generalizability.MethodsA modified 2D U-Net framework was employed for WBD of MRI rodent brains, consisting of 27 convolutional layers, batch normalization, two dropout layers and data augmentation, after training parameter optimization. A total of 265 T2-weighted 7.0 T MRI scans were utilized for the study, including 125 scans of an OPI rat model for neural network training. For testing and validation, 20 OPI rat scans and 120 scans of an AD rat model were utilized. U-Net performance was evaluated using Dice coefficients (DC) and Hausdorff distances (HD) between the U-Net-generated and manually segmented WBDs.ResultsThe U-Net achieved a DC (median[range]) of 0.984[0.936-0.990] and HD of 1.69[1.01-6.78] mm for OPI rat model scans, and a DC (mean[range]) of 0.975[0.898-0.991] and HD of 1.49[0.86-3.89] for the AD rat model scans.Comparison with existing methodsThe proposed approach is fully automated and robust across two rat strains and longitudinal brain changes with a computational speed of 8 seconds/scan, overcoming limitations of manual segmentation.ConclusionsThe modified 2D U-Net provided a fully automated, efficient, and generalizable segmentation approach that achieved high accuracy across two disparate rat models of neurological diseases.

Today, colorectal cancer (CRC) diagnosis is performed using colonoscopy, which is the current, most effective screening method. However, colonoscopy poses risks of harm to the patient and is an invasive process. Recent research has proven metabolomics as a potential, non-invasive detection method, which can use identified biomarkers to detect potential cancer in a patients body. The aim of this study is to develop a machine-learning (ML) model based on chemical descriptors that will recognize CRC-associated metabolites. We selected a set of metabolites found as the biomarkers of CRC, confirmed that they participate in cancer-related pathways, and used them for training a machine-learning model for the diagnostics of CRC. Using a set of selective metabolites and random compounds, we developed a range of ML models. The best performing ML model trained on Stage 0-2 CRC metabolite data predicted a metabolite class with 89.55% accuracy. The best performing ML model trained on Stage 3-4 CRC metabolite data predicted a metabolite class with 95.21% accuracy. Lastly, the best-performing ML model trained on Stage 0-4 CRC metabolite data predicted a metabolite class with 93.04% accuracy. These models were then tested on independent datasets, including random and unrelated-disease metabolites. In addition, six pathways related to these CRC metabolites were also distinguished: aminoacyl-tRNA biosynthesis; glyoxylate and dicarboxylate metabolism; glycine, serine, and threonine metabolism; phenylalanine, tyrosine, and tryptophan biosynthesis; arginine biosynthesis; and alanine, aspartate, and glutamate metabolism. Thus, in this research study, we created machine-learning models based on metabolite-related descriptors that may be helpful in developing a non-invasive diagnosis method for CRC.

The movement of organic anionic drugs across cell membranes is partly governed by interactions with SLC and ABC transporters in the intestine, liver, kidney, blood-brain barrier, placenta, breast, and other tissues. Major transporters involved include organic anion transporters (OATs, SLC22 family), organic anion transporting polypeptides (OATPs, SLCO family), and multidrug resistance proteins (MRPs, ABCC family). However, the sets of molecular properties of drugs that are necessary for interactions with OATs (OAT1, OAT3) vs. OATPs (OATP1B1, OATP1B3) vs. MRPs (MRP2, MRP4) are not well-understood. Defining these molecular properties is necessary for a better understanding of drug and metabolite handling across the gut-liver-kidney axis, gut-brain axis, and other multi-organ axes. It is also useful for tissue targeting of small molecule drugs and predicting drug-drug interactions and drug-metabolite interactions. Here, we curated a database of drugs shown to interact with these transporters in vitro and used chemoinformatic approaches to describe their molecular properties. We then sought to define sets of molecular properties that distinguish drugs interacting with OATs, OATPs, and MRPs in binary classifications using machine learning and artificial intelligence approaches. We identified sets of key molecular properties (e.g., rotatable bond count, lipophilicity, number of ringed structures) for classifying OATs vs. MRPs and OATs vs. OATPs. However, sets of molecular properties differentiating OATP vs. MRP substrates were less evident, as drugs interacting with MRP2 and MRP4 do not form a tight group owing to differing hydrophobicity and molecular complexity for interactions with the two transporters. If the results also hold for endogenous metabolites, they may deepen our knowledge of organ crosstalk, as described in the Remote Sensing and Signaling Theory. The results also provide a molecular basis for understanding how small organic molecules differentially interact with OATs, OATPs, and MRPs.

Since the onset of the coronavirus disease 2019 (COVID-19) pandemic, SARS-CoV-2 variants capable of breakthrough infections have attracted global attention. These variants have significant mutations in the receptor-binding domain (RBD) of the spike protein and the membrane (M) protein, which may imply an enhanced ability to evade immune responses. In this study, an examination of co-mutations within the spike RBD and their potential correlation with mutations in the M protein was conducted. The EVmutation method was utilized to analyze the distribution of the mutations to elucidate the relationship between the mutations in the spike RBD and the alterations in the M protein. Additionally, the Sequence-to-Sequence Transformer Model (S2STM) was employed to establish mapping between the amino acid sequences of the spike RBD and M proteins, offering a novel and efficient approach for streamlined sequence analysis and the exploration of their interrelationship. Certain mutations in the spike RBD, G339D-S373P-S375F and Q493R-Q498R-Y505, are associated with a heightened propensity for inducing mutations at specific sites within the M protein, especially sites 3 and 19/63. These results shed light on the concept of mutational synergy between the spike RBD and M proteins, illuminating a potential mechanism that could be driving the evolution of SARS-CoV-2.

Small non-coding RNAs can be secreted through a variety of mechanisms, including exosomal sorting, in small extracellular vesicles, and within lipoprotein complexes. However, the mechanisms that govern their sorting and secretion are not well understood. Here, we present ExoGRU, a machine learning model that predicts small RNA secretion probabilities from primary RNA sequences. We experimentally validated the performance of this model through ExoGRU-guided mutagenesis and synthetic RNA sequence analysis. Additionally, we used ExoGRU to reveal cis and trans factors that underlie small RNA secretion, including known and novel RNA-binding proteins (RBPs), e.g., YBX1, HNRNPA2B1, and RBM24. We also developed a novel technique called exoCLIP, which reveals the RNA interactome of RBPs within the cell-free space. Together, our results demonstrate the power of machine learning in revealing novel biological mechanisms. In addition to providing deeper insight into small RNA secretion, this knowledge can be leveraged in therapeutic and synthetic biology applications.

The capture-recapture method is a common tool used in epidemiology to estimate the size of hidden populations and correct the underascertainment of cases, based on incomplete and overlapping lists of the target population. Log-linear models are often used to estimate the population size yet may produce implausible and unreliable estimates due to model misspecification and small cell sizes. A novel targeted minimum loss-based estimation (TMLE) model developed for capture-recapture makes several notable improvements to conventional modeling: targeting the parameter of interest, flexibly fitting the data to alternative functional forms, and limiting bias from small cell sizes. Using simulations and empirical data from the San Francisco, California, Department of Public Healths human immunodeficiency virus (HIV) surveillance registry, we evaluated the performance of the TMLE model and compared results with those of other common models. Based on 2,584 people observed on 3 lists reportable to the surveillance registry, the TMLE model estimated the number of San Francisco residents living with HIV as of December 31, 2019, to be 13,523 (95% confidence interval: 12,222, 14,824). This estimate, compared with a ground truth of 12,507, was the most accurate and precise of all models examined. The TMLE model is a significant advancement in capture-recapture studies, leveraging modern statistical methods to improve estimation of the sizes of hidden populations.

Fungi use the accessory gene content of their pangenomes to adapt to their environments. While gene presence-absence variation contributes to shaping accessory gene reservoirs, the genomic contexts that shape these events remain unclear. Since pangenome studies are typically species-wide and do not analyze different populations separately, it is yet to be uncovered whether presence-absence variation patterns and mechanisms are consistent across populations. Fungal plant pathogens are useful models for studying presence-absence variation because they rely on it to adapt to their hosts, and members of a species often infect distinct hosts. We analyzed gene presence-absence variation in the blast fungus, Magnaporthe oryzae (syn. Pyricularia oryzae), and found that presence-absence variation genes involved in host-pathogen and microbe-microbe interactions may drive the adaptation of the fungus to its environment. We then analyzed genomic and epigenomic features of presence-absence variation and observed that proximity to transposable elements, gene GC content, gene length, expression level in the host, and histone H3K27me3 marks were different between presence-absence variation genes and conserved genes. We used these features to construct a model that was able to predict whether a gene is likely to experience presence-absence variation with high precision (86.06%) and recall (92.88%) in M. oryzae. Finally, we found that presence-absence variation genes in the rice and wheat pathotypes of M. oryzae differed in their number and their genomic context. Our results suggest that genomic and epigenomic features of gene presence-absence variation can be used to better understand and predict fungal pangenome evolution. We also show that substantial intra-species variation can exist in these features.

UNLABELLED: Small sample sizes and loss of sequencing reads during the microbiome data preprocessing can limit the statistical power of differentiating fresh produce phenotypes and prevent the detection of important bacterial species associated with produce contamination or quality reduction. Here, we explored a machine learning-based k-mer hash analysis strategy to identify DNA signatures predictive of produce safety (PS) and produce quality (PQ) and compared it against the amplicon sequence variant (ASV) strategy that uses a typical denoising step and ASV-based taxonomy strategy. Random forest-based classifiers for PS and PQ using 7-mer hash data sets had significantly higher classification accuracy than those using the ASV data sets. We also demonstrated that the proposed combination of integrating multiple data sets and leveraging a 7-mer hash strategy leads to better classification performance for PS and PQ compared to the ASV method but presents lower PS classification accuracy compared to the feature-selected ASV-based taxonomy strategy. Due to the current limitation of generating taxonomy using the 7-mer hash strategy, the ASV-based taxonomy strategy with remarkably less computing time and memory usage is more efficient for PS and PQ classification and applicable for important taxa identification. Results generated from this study lay the foundation for future studies that wish and need to incorporate and/or compare different microbiome sequencing data sets for the application of machine learning in the area of microbial safety and quality of food. IMPORTANCE: Identification of generalizable indicators for produce safety (PS) and produce quality (PQ) improves the detection of produce contamination and quality decline. However, effective sequencing read loss during microbiome data preprocessing and the limited sample size of individual studies restrain statistical power to identify important features contributing to differentiating PS and PQ phenotypes. We applied machine learning-based models using individual and integrated k-mer hash and amplicon sequence variant (ASV) data sets for PS and PQ classification and evaluated their classification performance and found that random forest (RF)-based models using integrated 7-mer hash data sets achieved significantly higher PS and PQ classification accuracy. Due to the limitation of taxonomic analysis for the 7-mer hash, we also developed RF-based models using feature-selected ASV-based taxonomic data sets, which performed better PS classification than those using the integrated 7-mer hash data set. The RF feature selection method identified 480 PS indicators and 263 PQ indicators with a positive contribution to the PS and PQ classification.

Understanding thermal dynamics and obtaining the computational model of residential buildings enable its scaled application in energy retrofits, control optimization and decarbonization. In this paper, we present a deep learning approach to model building thermal dynamics with smart thermostat data collected from residential buildings, with the goal to investigate model generalizability. In the first stage, we developed and compared different Deep Learning architectures including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) models and CNN-LSTM to predict indoor air temperature in a multi-step time horizon. In the second stage, we implemented a Transfer Learning (TL) process, which aims to improve the prediction performance on a new set of buildings (targets), exploiting the knowledge of related or similar buildings (sources). Different TL strategies and source model identification methods were investigated. The study showed that the CNN-LSTM performed the best among the architectures compared, with an average Mean Absolute Error (MAE) of 0.26 °C for one-hour-ahead (twelve 5-min future steps) predictions. Furthermore, the results showed that freezing the LSTM layer and fine-tuning the other layers of the CNN-LSTM achieved the best performance among four TL strategies, which further improved the performance with respect to a machine learning approach by 10%, and proving the effectiveness and generalizability of the proposed approach. A comparison of three different source model identification methods showed that randomly selecting source models constrained by similar building characteristics can provide good TL performance while retaining simplicity comparing with other quantitative source identification methods.

Autonomous vehicles (AVs) heavily rely on machine learning-based perception models to accurately interpret their surroundings. However, these crucial perception components are vulnerable to a range of malicious attacks. Even though individual attacks can be highly successful, the actual security risks such attacks can pose to our daily life are unclear. Various factors, such as lack of stealthiness, cost-effectiveness, and ease of deployment, can deter potential attackers from employing certain attacks, thereby reducing the actual risk. This research report presents the first quantitative risk assessment for physical adversarial attacks on AVs. The specific focus is on attacks on AV’s perception components due to their highly critical function and representation in existing research. The report defines the daily-life risk as the likelihood that a given type of attack will be employed in real life and the authors develop a problem-specific risk scoring system and accompanying metrics. They perform an initial evaluation of the proposed risk assessment method for all the reported attacks on AVs from 2017 to 2023. They quantitatively rank the daily-life risks posed by each of eight different categories of attacks s and find three attacks with the highest risks: 2D printed images, 2D patches, and coated camouflage stickers, which deserve more focused attention for potential future mitigation strategy development and policy making.

The evolution of pathway enzymes enhances the biosynthesis of high-value chemicals, crucial for pharmaceutical, and agrochemical applications. However, unpredictable evolutionary landscapes of pathway genes often hinder successful evolution. Here, the presence of complex epistasis is identifued within the representative naringenin biosynthetic pathway enzymes, hampering straightforward directed evolution. Subsequently, a biofoundry-assisted strategy is developed for pathway bottlenecking and debottlenecking, enabling the parallel evolution of all pathway enzymes along a predictable evolutionary trajectory in six weeks. This study then utilizes a machine learning model, ProEnsemble, to further balance the pathway by optimizing the transcription of individual genes. The broad applicability of this strategy is demonstrated by constructing an Escherichia coli chassis with evolved and balanced pathway genes, resulting in 3.65 g L-1 naringenin. The optimized naringenin chassis also demonstrates enhanced production of other flavonoids. This approach can be readily adapted for any given number of enzymes in the specific metabolic pathway, paving the way for automated chassis construction in contemporary biofoundries.

Global change is currently impacting ecosystems and their contributions to people (i.e. ecosystem services). These impacts have consequences for societies and human well-being, especially in Africa. Historically, efforts have focused on assessing global change from a social or biophysical perspective, treating them as separate entities. Yet, our understanding of impacts to social-ecological systems remains limited, particularly in the Global South, due to a lack of data, tools, and approaches accounting for social and ecological aspects of ecosystem services. This is especially relevant for cultural ecosystem services as they are less tangible. We use a simple indicator and important provider of a multitude of cultural ecosystem services, birding, to understand how climate, biodiversity, and land use change will impact cultural ecosystem services across Africa. We explore how emerging tools and data can overcome limitations in mapping and modeling cultural ecosystem services, particularly in analyzing human preferences and behavior at large spatiotemporal scales and in data-poor regions. Leveraging crowdsourced data from eBird and using machine learning techniques we map and model recreational birding to assess the underlying social-ecological relationships and the impact of future climate and environmental change. We show that bird species richness, protected areas, accessibility, and max temperature contribute most to birding suitability across the continent. Further, we show spatial shifts in the suitability of birding under three future climate scenarios (SSP126, 370, and 585). Models suggest climate and biodiversity change will increasingly constrain the flow of birding related cultural ecosystem services across Africa. This has implications for human-nature interactions, development of countries, management of protected areas, and overall human well-being in the future. More generally, we highlight opportunities for crowdsourced datasets and machine learning to integrate non-material ecosystem services in models and thus, enhance the understanding of future impacts to ecosystem services and human well-being.

The article discusses the potential benefits of using artificial intelligence (AI) and machine learning (ML) tools and algorithms in physical activity and sedentary behavior research. While AI and ML have been widely utilized in healthcare and medicine, their application in studying physical activity and sedentary behavior is still underutilized. The article highlights the capabilities of AI and ML in generating novel hypotheses, improving measurement of behaviors, predicting health indicators, and advancing precision medicine. However, there are challenges in implementing these techniques, including ethical concerns and limitations in identifying cause-effect relationships. Overall, integrating AI and ML tools can contribute to a deeper understanding of physical activity and sedentary behavior. [Extracted from the article]

The carbon nano-peapod is a representative structure with interlayer van der Waals (vdW) interactions, in which encapsulated fullerene molecules play a critical role in modulating the transport properties of the carbon nanotubes (CNTs). In particular, their influence on the thermal transport characteristics has been the focal point of considerable attention. In this study, we trained an accurate machine learning potential for fullerene-encapsulated CNTs based on the efficient NEP model to investigate their thermal properties. Using equilibrium molecular dynamics simulation along with the spectral decomposition method for thermal conductivity, we find that the thermal conductivity of fullerene-encapsulated CNTs is roughly 55 % lower than that of empty CNTs, aligning with experimental observations for CNT bundles with fullerene encapsulation [Kodama et al., Nat. Mater. 16, 892 (2017)]. The research suggests that weak vdW interactions between both the fullerene and CNTs, as well as between fullerene molecules themselves, hinder phonon propagation. The encapsulated fullerene contributes to an increase in phonon scattering within the CNTs, ultimately leading to a reduction in thermal conductivity. We utilized machine learning potential to investigate the structure of fullerene-encapsulated CNTs and their heat transport property. This approach provides valuable insights for performance research of complex systems featuring interlayer vdW interactions. [ABSTRACT FROM AUTHOR]

Machine learning (ML) of kinetic energy functionals (KEFs), in particular kinetic energy density (KED) functionals, is a promising way to construct KEFs for orbital-free density functional theory (DFT). Neural networks and kernel methods including Gaussian process regression (GPR) have been used to learn Kohn–Sham (KS) KED from density-based descriptors derived from KS DFT calculations. The descriptors are typically expressed as functions of different powers and derivatives of the electron density. This can generate large and extremely unevenly distributed datasets, which complicates effective application of ML techniques. Very uneven data distributions require many training datapoints, can cause overfitting, and can ultimately lower the quality of an ML KED model. We show that one can produce more accurate ML models from fewer data by working with smoothed density-dependent variables and KED. Smoothing palliates the issue of very uneven data distributions and associated difficulties of sampling while retaining enough spatial structure necessary for working within the paradigm of KEDF. We use GPR as a function of smoothed terms of the fourth order gradient expansion and KS effective potential and obtain accurate and stable (with respect to different random choices of training points) kinetic energy models for Al, Mg, and Si simultaneously from as few as 2000 samples (about 0.3% of the total KS DFT data). In particular, accuracies on the order of 1% in a measure of the quality of energy–volume dependence B ′ = E V 0 − Δ V − 2 E V 0 + E ( V 0 + Δ V) Δ V / V 0 2 (where V0 is the equilibrium volume and ΔV is a deviation from it) are obtained simultaneously for all three materials. [ABSTRACT FROM AUTHOR]

Understanding interparticle interactions has been one of the most important topics of research in the field of micro/nanoscale materials. Many significant characteristics of such materials directly stem from the way their building blocks interact with each other. In this work, we investigate the efficacy of a specific category of Machine Learning (ML) methods known as interaction networks in predicting interparticle interactions within colloidal systems. We introduce and study Local Neighborhood Graph Neural Networks (LN-GNNs), defined according to the local environment of colloidal particles derived from particle trajectory data. The LN-GNN framework is trained for unique categories of particle neighborhood environments in order to predict interparticle interactions. We compare the performance of the LN-GNN to a baseline interaction network with a simpler architecture and to an Instance-Based ML algorithm, which is computationally more expensive. We find that the prediction performance of LN-GNN measured as an average normalized mean absolute error outperforms the baseline interaction network by a factor of 2–10 for different local neighborhood configurations. Furthermore, LN-GNN's performance turns out to be very comparable to the instance-based ML framework while being an order of magnitude less expensive in terms of the required computation time. The results of this work can provide the foundations for establishing accurate models of colloidal particle interactions that are derived from real particle trajectory data. [ABSTRACT FROM AUTHOR]

Coarse-grained (CG) molecular dynamics can be a powerful method for probing complex processes. However, most CG force fields use pairwise nonbonded interaction potentials sets, which can limit their ability to capture complex multi-body phenomena such as the hydrophobic effect. As the hydrophobic effect primarily manifests itself due to the nonpolar solute affecting the nearby hydrogen bonding network in water, capturing such effects using a simple one CG site or "bead" water model is a challenge. In this work, we systematically test the ability of CG one site water models for capturing critical features of the solvent environment around a hydrophobe as well as the potential of mean force (PMF) of neopentane association. We study two bottom-up models: a simple pairwise (SP) force-matched water model constructed using the multiscale coarse-graining method and the Bottom-Up Many-Body Projected Water (BUMPer) model, which has implicit three-body correlations. We also test the top-down monatomic (mW) and the Machine Learned mW (ML-mW) water models. The mW models perform well in capturing structural correlations but not the energetics of the PMF. BUMPer outperforms SP in capturing structural correlations and also gives an accurate PMF in contrast to the two mW models. Our study highlights the importance of including three-body interactions in CG water models, either explicitly or implicitly, while in general highlighting the applicability of bottom-up CG water models for studying hydrophobic effects in a quantitative fashion. This assertion comes with a caveat, however, regarding the accuracy of the enthalpy–entropy decomposition of the PMF of hydrophobe association. [ABSTRACT FROM AUTHOR]

We investigated the accelerated prediction of the thermal conductivity of materials through end-to-end structure-based approaches employing machine learning methods. Due to the non-availability of high-quality thermal conductivity data, we first performed high-throughput calculations based on first principles and the Boltzmann transport equation for 225 materials, effectively more than doubling the size of the existing dataset. We assessed the performance of state-of-the-art machine learning models for thermal conductivity prediction on this expanded dataset and observed that all these models suffered from overfitting. To address this issue, we introduced a different graph-based neural network model, which demonstrated more consistent and regularized performance across all evaluated datasets. Nevertheless, the best mean absolute percentage error achieved on the test dataset remained in the range of 50–60%. This suggests that while these models are valuable for expediting material screening, their current accuracy is still limited. [ABSTRACT FROM AUTHOR]