Licheng Wen, Xuemeng Yang, Daocheng Fu, Xiaofeng Wang, Pinlong Cai, Xin Li, Tao Ma, Yingxuan Li, Linran Xu, Dengke Shang, Zheng Zhu, Shaoyan Sun, Yeqi Bai, Xinyu Cai, Min Dou, Shuanglu Hu, Botian Shi, Yu Qiao <<<

The pursuit of autonomous driving technology hinges on the sophisticatedintegration of perception, decision-making, and control systems. Traditionalapproaches, both data-driven and rule-based, have been hindered by theirinability to grasp the nuance of complex driving environments and theintentions of other road users. This has been a significant bottleneck,particularly in the development of common sense reasoning and nuanced sceneunderstanding necessary for safe and reliable autonomous driving. The advent ofVisual Language Models (VLM) represents a novel frontier in realizing fullyautonomous vehicle driving. This report provides an exhaustive evaluation ofthe latest state-of-the-art VLM, GPT-4V(ision), and its application inautonomous driving scenarios. We explore the model's abilities to understandand reason about driving scenes, make decisions, and ultimately act in thecapacity of a driver. Our comprehensive tests span from basic scene recognitionto complex causal reasoning and real-time decision-making under varyingconditions. Our findings reveal that GPT-4V demonstrates superior performancein scene understanding and causal reasoning compared to existing autonomoussystems. It showcases the potential to handle out-of-distribution scenarios,recognize intentions, and make informed decisions in real driving contexts.However, challenges remain, particularly in direction discernment, trafficlight recognition, vision grounding, and spatial reasoning tasks. Theselimitations underscore the need for further research and development. Projectis now available on GitHub for interested parties to access and utilize:\url{https://github.com/PJLab-ADG/GPT4V-AD-Exploration} <<<

Time and space are primary dimensions of human experience. Separate lines of investigation have identified neural correlates of time and space, yet little is known about how these representations converge during self-guided experience. Here, 10 subjects with intracranially implanted microelectrodes play a timed, virtual navigation game featuring object search and retrieval tasks separated by fixed delays. Time cells and place cells activate in parallel during timed navigation intervals, whereas a separate time cell sequence spans inter-task delays. The prevalence, firing rates, and behavioral coding strengths of time cells and place cells are indistinguishable-yet time cells selectively remap between search and retrieval tasks, while place cell responses remain stable. Thus, the brain can represent time and space as overlapping but dissociable dimensions. Time cells and place cells may constitute a biological basis for the cognitive map of spatiotemporal context onto which memories are written. <<<

Double-negative (DN) T cells are present at relatively low frequencies in human peripheral blood, and are characterized as expressing the alpha-beta or gamma-delta T-cell receptor (TCR), but not the CD4 nor the CD8 co-receptors. Despite their low frequencies, these cells are potent producers of cytokines and, thus, are key orchestrators of immune responses. DN T cells were initially associated with induction of peripheral immunological tolerance and immunomodulatory activities related to disease prevention. However, other studies demonstrated that these cells can also display effector functions associated with pathology development. This apparent contradiction highlighted the heterogeneity of the DN T-cell population. Here, we review phenotypic and functional characteristics of DN T cells, emphasizing their role in human diseases. The need for developing biomarkers to facilitate the translation of studies from animal models to humans will also be discussed. Finally, we will examine DN T cells as promising therapeutic targets to prevent or inhibit human disease development. <<<

Yunli Lai, Xiaofan Zhu, Sheng He, Zirui Dong, Yanqing Tang, Fuben Xu, Yun Chen, Lintao Meng, Yuli Tao, Shang Yi, Jiasun Su, Hongqian Huang, Jingsi Luo, Tak Yeung Leung, Hongwei Wei <<<

To evaluate the performance of noninvasive prenatal screening (NIPS) in the detection of common aneuploidies in a population-based study, a total of 86,262 single pregnancies referred for NIPS were prospectively recruited. Among 86,193 pregnancies with reportable results, follow-up was successfully conducted in 1160 fetuses reported with a high-risk result by NIPS and 82,511 cases (95.7%) with a low-risk result. The screen-positive rate (SPR) of common aneuploidies and sex chromosome abnormalities (SCAs) provided by NIPS were 0.7% (586/83,671) and 0.6% (505/83,671), respectively. The positive predictive values (PPVs) for Trisomy 21, Trisomy 18, Trisomy 13 and SCAs were calculated as 89.7%, 84.0%, 52.6% and 38.0%, respectively. In addition, less rare chromosomal abnormalities, including copy number variants (CNVs), were detected, compared with those reported by NIPS with higher read-depth. Among these rare abnormalities, only 23.2% (13/56) were confirmed by prenatal diagnosis. In total, four common trisomy cases were found to be false negative, resulting in a rate of 0.48/10,000 (4/83,671). In summary, this study conducted in an underdeveloped region with limited support for the new technology development and lack of cost-effective prenatal testing demonstrates the importance of implementing routine aneuploidy screening in the public sector for providing early detection and precise prognostic information. <<<

Joe Nassour, Lucia Gutierrez Aguiar, Adriana Correia, Tobias T Schmidt, Laura Mainz, Sara Przetocka, Candy Haggblom, Nimesha Tadepalle, April Williams, Maxim N Shokhirev, Semih C Akincilar, Vinay Tergaonkar, Gerald S Shadel, Jan Karlseder <<<

Cancers arise through the accumulation of genetic and epigenetic alterations that enable cells to evade telomere-based proliferative barriers and achieve immortality. One such barrier is replicative crisis-an autophagy-dependent program that eliminates checkpoint-deficient cells with unstable telomeres and other cancer-relevant chromosomal aberrations. However, little is known about the molecular events that regulate the onset of this important tumour-suppressive barrier. Here we identified the innate immune sensor Z-DNA binding protein 1 (ZBP1) as a regulator of the crisis program. A crisis-associated isoform of ZBP1 is induced by the cGAS-STING DNA-sensing pathway, but reaches full activation only when associated with telomeric-repeat-containing RNA (TERRA) transcripts that are synthesized from dysfunctional telomeres. TERRA-bound ZBP1 oligomerizes into filaments on the outer mitochondrial membrane of a subset of mitochondria, where it activates the innate immune adapter protein mitochondrial antiviral-signalling protein (MAVS). We propose that these oligomerization properties of ZBP1 serve as a signal amplification mechanism, where few TERRA-ZBP1 interactions are sufficient to launch a detrimental MAVS-dependent interferon response. Our study reveals a mechanism for telomere-mediated tumour suppression, whereby dysfunctional telomeres activate innate immune responses through mitochondrial TERRA-ZBP1 complexes to eliminate cells destined for neoplastic transformation. <<<

Mingrui Xia, Tianmei Si, Xiaoyi Sun, Qing Ma, Bangshan Liu, Li Wang, Jie Meng, Miao Chang, Xiaoqi Huang, Ziqi Chen, Yanqing Tang, Ke Xu, Qiyong Gong, Fei Wang, Jiang Qiu, Peng Xie, Lingjiang Li, Yong He, DIDA-Major Depressive Disorder Working Group <<<

Resting-state functional MRI (R-fMRI) studies have demonstrated widespread alterations in brain function in patients with major depressive disorder (MDD). However, a clear and consistent conclusion regarding a repeatable pattern of MDD-relevant alterations is still limited due to the scarcity of large-sample, multisite datasets. Here, we address this issue by including a large R-fMRI dataset with 1434 participants (709 patients with MDD and 725 healthy controls) from five centers in China. Individual functional activity maps that represent very local to long-range connections are computed using the amplitude of low-frequency fluctuations, regional homogeneity and distance-related functional connectivity strength. The reproducibility analyses involve different statistical strategies, global signal regression, across-center consistency, clinical variables, and sample size. We observed significant hypoactivity in the orbitofrontal, sensorimotor, and visual cortices and hyperactivity in the frontoparietal cortices in MDD patients compared to the controls. These alterations are not affected by different statistical analysis strategies, global signal regression and medication status and are generally reproducible across centers. However, these between-group differences are partially influenced by the episode status and the age of disease onset in patients, and the brain-clinical variable relationship exhibits poor cross-center reproducibility. Bootstrap analyses reveal that at least 400 subjects in each group are required to replicate significant alterations (an extent threshold of P < .05 and a height threshold of P < .001) at 50% reproducibility. Together, these results highlight reproducible patterns of functional alterations in MDD and relevant influencing factors, which provides crucial guidance for future neuroimaging studies of this disorder. <<<

The rapid progress in the field of deep learning has had a significant impact on protein design. Deep learning methods have recently produced a breakthrough in protein structure prediction, leading to the availability of high-quality models for millions of proteins. Along with novel architectures for generative modeling and sequence analysis, they have revolutionized the protein design field in the past few years remarkably by improving the accuracy and ability to identify novel protein sequences and structures. Deep neural networks can now learn and extract the fundamental features of protein structures, predict how they interact with other biomolecules, and have the potential to create new effective drugs for treating disease. As their applicability in protein design is rapidly growing, we review the recent developments and technology in deep learning methods and provide examples of their performance to generate novel functional proteins. <<<

Variational autoencoders are powerful algorithms for identifying dominantlatent structure in a single dataset. In many applications, however, we areinterested in modeling latent structure and variation that are enriched in atarget dataset compared to some background---e.g. enriched in patients comparedto the general population. Contrastive learning is a principled framework tocapture such enriched variation between the target and background, butstate-of-the-art contrastive methods are limited to linear models. In thispaper, we introduce the contrastive variational autoencoder (cVAE), whichcombines the benefits of contrastive learning with the power of deep generativemodels. The cVAE is designed to identify and enhance salient latent features.The cVAE is trained on two related but unpaired datasets, one of which hasminimal contribution from the salient latent features. The cVAE explicitlymodels latent features that are shared between the datasets, as well as thosethat are enriched in one dataset relative to the other, which allows thealgorithm to isolate and enhance the salient latent features. The algorithm isstraightforward to implement, has a similar run-time to the standard VAE, andis robust to noise and dataset purity. We conduct experiments across diversetypes of data, including gene expression and facial images, showing that thecVAE effectively uncovers latent structure that is salient in a particularanalysis. <<<

Danillo G Augusto, Lawton D Murdolo, Demetra S M Chatzileontiadou, Joseph J Sabatino, Tasneem Yusufali, Noah D Peyser, Xochitl Butcher, Kerry Kizer, Karoline Guthrie, Victoria W Murray, Vivian Pae, Sannidhi Sarvadhavabhatla, Fiona Beltran, Gurjot S Gill, Kara L Lynch, Cassandra Yun, Colin T Maguire, Michael J Peluso, Rebecca Hoh, Timothy J Henrich, Steven G Deeks, Michelle Davidson, Scott Lu, Sarah A Goldberg, J Daniel Kelly, Jeffrey N Martin, Cynthia A Vierra-Green, Stephen R Spellman, David J Langton, Michael J Dewar-Oldis, Corey Smith, Peter J Barnard, Sulggi Lee, Gregory M Marcus, Jeffrey E Olgin, Mark J Pletcher, Martin Maiers, Stephanie Gras, Jill A Hollenbach <<<

Studies have demonstrated that at least 20% of individuals infected with SARS-CoV-2 remain asymptomatic. Although most global efforts have focused on severe illness in COVID-19, examining asymptomatic infection provides a unique opportunity to consider early immunological features that promote rapid viral clearance. Here, postulating that variation in the human leukocyte antigen (HLA) loci may underly processes mediating asymptomatic infection, we enrolled 29,947 individuals, for whom high-resolution HLA genotyping data were available, in a smartphone-based study designed to track COVID-19 symptoms and outcomes. Our discovery cohort (n = 1,428) comprised unvaccinated individuals who reported a positive test result for SARS-CoV-2. We tested for association of five HLA loci with disease course and identified a strong association between HLA-B*15:01 and asymptomatic infection, observed in two independent cohorts. Suggesting that this genetic association is due to pre-existing T cell immunity, we show that T cells from pre-pandemic samples from individuals carrying HLA-B*15:01 were reactive to the immunodominant SARS-CoV-2 S-derived peptide NQKLIANQF. The majority of the reactive T cells displayed a memory phenotype, were highly polyfunctional and were cross-reactive to a peptide derived from seasonal coronaviruses. The crystal structure of HLA-B*15:01-peptide complexes demonstrates that the peptides NQKLIANQF and NQKLIANAF (from OC43-CoV and HKU1-CoV) share a similar ability to be stabilized and presented by HLA-B*15:01. Finally, we show that the structural similarity of the peptides underpins T cell cross-reactivity of high-affinity public T cell receptors, providing the molecular basis for HLA-B*15:01-mediated pre-existing immunity. <<<

PURPOSE: Deformable image registration (DIR) plays an important role in assisting disease diagnosis. The emergence of the Transformer enables the DIR framework to extract long-range dependencies, which relieves the limitations of intrinsic locality caused by convolution operation. However, suffering from the interference of missing or spurious connections, it is a challenging task for Transformer-based methods to capture the high-quality long-range dependencies.METHODS: In this paper, by staking the graph convolution Transformer (Graformer) layer at the bottom of the feature extraction network, we propose a Graformer-based DIR framework, named GraformerDIR. The Graformer layer is consist of the Graformer module and the Cheby-shev graph convolution module. Among them, the Graformer module is designed to capture high-quality long-range dependencies. Cheby-shev graph convolution module is employed to further enlarge the receptive field.RESULTS: The performance and generalizability of GraformerDIR have been evaluated on publicly available brain datasets including the OASIS, LPBA40, and MGH10 datasets. Compared with VoxelMorph, the GraformerDIR has obtained performance improvements of 4.6% in Dice similarity coefficient (DSC) and 0.055 mm in the average symmetric surface distance (ASD) while reducing the non-positive rate of Jacobin determinant (Npr.Jac) index about 60 times on publicly available OASIS dataset. On unseen dataset MGH10, the GraformerDIR has obtained the performance improvements of 4.1% in DSC and 0.084 mm in ASD compared with VoxelMorph, which demonstrates the GraformerDIR with better generalizability. The promising performance on the clinical cardiac dataset ACDC indicates the GraformerDIR is practicable.CONCLUSION: With the advantage of Transformer and graph convolution, the GraformerDIR has obtained comparable performance with the state-of-the-art method VoxelMorph. <<<

Chunhui Li, Honghui Guan, Xin Jing, Yaoyao Li, Baobao Wang, Yongxiang Li, Xuyang Liu, Dengfeng Zhang, Cheng Liu, Xiaoqing Xie, Haiyan Zhao, Yanbo Wang, Jingbao Liu, Panpan Zhang, Guanghui Hu, Guoliang Li, Suiyan Li, Dequan Sun, Xiaoming Wang, Yunsu Shi, Yanchun Song, Chengzhi Jiao, Jeffrey Ross-Ibarra, Yu Li, Tianyu Wang, Haiyang Wang <<<

Single-cross maize hybrids display superior heterosis and are produced from crossing two parental inbred lines belonging to genetically different heterotic groups. Here we assembled 1,604 historically utilized maize inbred lines belonging to various female heterotic groups (FHGs) and male heterotic groups (MHGs), and conducted phenotyping and genomic sequencing analyses. We found that the FHGs and MHGs have undergone both convergent and divergent changes for different sets of agronomic traits. Using genome-wide selection scans and association analyses, we identified a large number of candidate genes that contributed to the improvement of agronomic traits of the FHGs and MHGs. Moreover, we observed increased genetic differentiation between the FHGs and MHGs across the breeding eras, and we found a positive correlation between increasing heterozygosity levels in the differentiated genes and heterosis in hybrids. Furthermore, we validated the function of two selected genes and a differentiated gene. This study provides insights into the genomic basis of modern hybrid maize breeding. <<<

Isothermal autocatalytic DNA circuits have been proven to be versatile and powerful biocomputing platforms by virtue of their self-sustainable and self-accelerating reaction profiles, yet they are currently constrained by their complicated designs, severe signal leakages, and unclear reaction mechanisms. Herein, we developed a simpler-yet-efficient autocatalytic assembly circuit (AAC) for highly robust bioimaging in live cells and mice. The scalable and sustainable AAC system was composed of a mere catalytic DNA assembly reaction with minimal strand complexity and, upon specific stimulation, could reproduce numerous new triggers to expedite the whole reaction. Through in-depth theoretical simulations and systematic experimental demonstrations, the catalytic efficiency of these reproduced triggers was found to play a vital role in the autocatalytic profile and thus could be facilely improved to achieve more efficient and characteristic autocatalytic signal amplification. Due to its exponentially high signal amplification and minimal reaction components, our self-stacking AAC facilitated the efficient detection of trace biomolecules with low signal leakage, thus providing great clinical diagnosis and therapeutic assessment potential. <<<

Long term particulate matter (PM) exposure has been associated with an increased incidence of respiratory diseases. Here, an in vitro model was developed to study how long term diesel exhaust particle (DEP) exposure might predispose to the development of allergic reactions. Airway epithelial (16HBE) cells were exposed to low concentrations of diesel exhaust particle (DEP) for 4 days after which they were challenged with house dust mite (HDM) extract (24 h). Compared to acute exposure (24 h), 4 days DEP exposure to 16HBE cells further reduced the transepithelial electrical resistance (TEER) and increased CXCL-8 release. DEP pre-exposure aggravated HDM-induced loss of TEER, increased tracer flux across the barrier and reduced CLDN-3 expression in these 16HBE cells. HDM-induced cytokine (IL-6, CCL-22, IL-10 and CXCL-8) release was significantly increased after DEP pre-exposure. In the current study an in vitro model with long term PM exposure was presented, which might be helpful for further understanding the interplay between long term PM exposure and allergic responses. <<<

Judit Kumuthini, Michael Chimenti, Sven Nahnsen, Alexander Peltzer, Rebone Meraba, Ross McFadyen, Gordon Wells, Deanne Taylor, Mark Maienschein-Cline, Jian-Liang Li, Jyothi Thimmapuram, Radha Murthy-Karuturi, Lyndon Zass <<<

Life scientists are increasingly turning to high-throughput sequencing technologies in their research programs, owing to the enormous potential of these methods. In a parallel manner, the number of core facilities that provide bioinformatics support are also increasing. Notably, the generation of complex large datasets has necessitated the development of bioinformatics support core facilities that aid laboratory scientists with cost-effective and efficient data management, analysis, and interpretation. In this article, we address the challenges-related to communication, good laboratory practice, and data handling-that may be encountered in core support facilities when providing bioinformatics support, drawing on our own experiences working as support bioinformaticians on multidisciplinary research projects. Most importantly, the article proposes a list of guidelines that outline how these challenges can be preemptively avoided and effectively managed to increase the value of outputs to the end user, covering the entire research project lifecycle, including experimental design, data analysis, and management (i.e., sharing and storage). In addition, we highlight the importance of clear and transparent communication, comprehensive preparation, appropriate handling of samples and data using monitoring systems, and the employment of appropriate tools and standard operating procedures to provide effective bioinformatics support. <<<

The COVID-19 pandemic has emphasized the need for novel drug discovery process. However, the journey from conceptualizing a drug to its eventual implementation in clinical settings is a long, complex, and expensive process, with many potential points of failure. Over the past decade, a vast growth in medical information has coincided with advances in computational hardware (cloud computing, GPUs, and TPUs) and the rise of deep learning. Medical data generated from large molecular screening profiles, personal health or pathology records, and public health organizations could benefit from analysis by Artificial Intelligence (AI) approaches to speed up and prevent failures in the drug discovery pipeline. We present applications of AI at various stages of drug discovery pipelines, including the inherently computational approaches of design and prediction of a drug's likely properties. Open-source databases and AI-based software tools that facilitate drug design are discussed along with their associated problems of molecule representation, data collection, complexity, labeling, and disparities among labels. How contemporary AI methods, such as graph neural networks, reinforcement learning, and generated models, along with structure-based methods, (i.e., molecular dynamics simulations and molecular docking) can contribute to drug discovery applications and analysis of drug responses is also explored. Finally, recent developments and investments in AI-based start-up companies for biotechnology, drug design and their current progress, hopes and promotions are discussed in this article. <<<

Ales Varabyou, Markus J Sommer, Beril Erdogdu, Ida Shinder, Ilia Minkin, Kuan-Hao Chao, Sukhwan Park, Jakob Heinz, Christopher Pockrandt, Alaina Shumate, Natalia Rincon, Daniela Puiu, Martin Steinegger, Steven L Salzberg, Mihaela Pertea <<<

CHESS 3 represents an improved human gene catalog based on nearly 10,000 RNA-seq experiments across 54 body sites. It significantly improves current genome annotation by integrating the latest reference data and algorithms, machine learning techniques for noise filtering, and new protein structure prediction methods. CHESS 3 contains 41,356 genes, including 19,839 protein-coding genes and 158,377 transcripts, with 14,863 protein-coding transcripts not in other catalogs. It includes all MANE transcripts and at least one transcript for most RefSeq and GENCODE genes. On the CHM13 human genome, the CHESS 3 catalog contains an additional 129 protein-coding genes. CHESS 3 is available at http://ccb.jhu.edu/chess . <<<

Our ability to enact successful goal-directed actions involves multiple learning processes. Among these processes, implicit motor adaptation ensures that the sensorimotor system remains finely tuned in response to changes in the body and environment. Whether Parkinson's disease impacts implicit motor adaptation remains a contentious area of research: whereas multiple reports show impaired performance in this population, many others show intact performance. While there is a range of methodological differences across studies, one critical issue is that performance in many of the studies may reflect a combination of implicit adaptation and strategic re-aiming. Here, we revisited this controversy using a visuomotor task designed to isolate implicit adaptation. In two experiments, we found that adaptation in response to a wide range of visual perturbations was similar in Parkinson's disease and matched control participants. Moreover, in a meta-analysis of previously published and unpublished work, we found that the mean effect size contrasting Parkinson's disease and controls across 16 experiments involving over 200 participants was not significant. Together, these analyses indicate that implicit adaptation is preserved in Parkinson's disease, offering a fresh perspective on the role of the basal ganglia in sensorimotor learning. <<<

BACKGROUND: Deep learning is a promising way to improve health care. Image-processing medical disciplines, such as pathology, are expected to be transformed by deep learning. The first clinically applicable deep-learning diagnostic support tools are already available in cancer pathology, and their number is increasing. However, data on the environmental sustainability of these tools are scarce. We aimed to conduct an environmental-sustainability analysis of a theoretical implementation of deep learning in patient-care pathology.METHODS: For this modelling study, we first assembled and calculated relevant data and parameters of a digital-pathology workflow. Data were breast and prostate specimens from the university clinic at the Institute of Pathology of the Rheinisch-Westfälische Technische Hochschule Aachen (Aachen, Germany), for which commercially available deep learning was already available. Only specimens collected between Jan 1 and Dec 31, 2019 were used, to omit potential biases due to the COVID-19 pandemic. Our final selection was based on 2 representative weeks outside holidays, covering different types of specimens. To calculate carbon dioxide (CO2) or CO2 equivalent (CO2 eq) emissions of deep learning in pathology, we gathered relevant data for exact numbers and sizes of whole-slide images (WSIs), which were generated by scanning histopathology samples of prostate and breast specimens. We also evaluated different data input scenarios (including all slide tiles, only tiles containing tissue, or only tiles containing regions of interest). To convert estimated energy consumption from kWh to CO2 eq, we used the internet protocol address of the computational server and the Electricity Maps database to obtain information on the sources of the local electricity grid (ie, renewable vs non-renewable), and estimated the number of trees and proportion of the local and world's forests needed to sequester the CO2 eq emissions. We calculated the computational requirements and CO2 eq emissions of 30 deep-learning models that varied in task and size. The first scenario represented the use of one commercially available deep-learning model for one task in one case (1-task), the second scenario considered two deep-learning models for two tasks per case (2-task), the third scenario represented a future, potentially automated workflow that could handle 7 tasks per case (7-task), and the fourth scenario represented the use of a single potential, large, computer-vision model that could conduct multiple tasks (multitask). We also compared the performance (ie, accuracy) and CO2 eq emissions of different deep-learning models for the classification of renal cell carcinoma on WSIs, also from Rheinisch-Westfälische Technische Hochschule Aachen. We also tested other approaches to reducing CO2 eq emissions, including model pruning and an alternative method for histopathology analysis (pathomics).FINDINGS: The pathology database contained 35 552 specimens (237 179 slides), 6420 of which were prostate specimens (10 115 slides) and 11 801 of which were breast specimens (19 763 slides). We selected and subsequently digitised 140 slides from eight breast-cancer cases and 223 slides from five prostate-cancer cases. Applying large deep-learning models on all WSI tiles of prostate and breast pathology cases would result in yearly CO2 eq emissions of 7·65 metric tons (t; 95% CI 7·62-7·68) with the use of a single deep-learning model per case; yearly CO2 eq emissions were up to 100·56 t (100·21-100·99) with the use of seven deep-learning models per case. CO2 eq emissions for different deep-learning model scenarios, data inputs, and deep-learning model sizes for all slides varied from 3·61 t (3·59-3·63) to 2795·30 t (1177·51-6482·13. For the estimated number of overall pathology cases worldwide, the yearly CO2 eq emissions varied, reaching up to 16 megatons (Mt) of CO2 eq, requiring up to 86 590 km2 (0·22%) of world forest to sequester the CO2 eq emissions. Use of the 7-task scenario and small deep-learning models on slides containing tissue only could substantially reduce CO2 eq emissions worldwide by up to 141 times (0·1 Mt, 95% CI 0·1-0·1). Considering the local environment in Aachen, Germany, the maximum CO2 eq emission from the use of deep learning in digital pathology only would require 32·8% (95% CI 13·8-76·6) of the local forest to sequester the CO2 eq emissions. A single pathomics run on a tissue could provide information that was comparable to or even better than the output of multitask deep-learning models, but with 147 times reduced CO2 eq emissions.INTERPRETATION: Our findings suggest that widespread use of deep learning in pathology might have considerable global-warming potential. The medical community, policy decision makers, and the public should be aware of this potential and encourage the use of CO2 eq emissions reduction strategies where possible.FUNDING: German Research Foundation, European Research Council, German Federal Ministry of Education and Research, Health, Economic Affairs and Climate Action, and the Innovation Fund of the Federal Joint Committee. <<<