Prediagnosis recognition of acute ischemic stroke by artificial intelligence from facial images.

Stroke is a major threat to life and health in modern society, especially in the aging population. Stroke may cause sudden death or severe sequela-like hemiplegia. Although computed tomography (CT) and magnetic resonance imaging (MRI) are standard diagnosis methods, and artificial intelligence models have been built based on these images, shortage in medical resources and the time and cost of CT/MRI imaging hamper fast detection, thus increasing the severity of stroke. Here, we developed a convolutional neural network model by integrating four networks, Xception, ResNet50, VGG19, and EfficientNetb1, to recognize stroke based on 2D facial images with a cross-validation area under curve (AUC) of 0.91 within the training set of 185 acute ischemic stroke patients and 551 age- and sex-matched controls, and AUC of 0.82 in an independent data set regardless of age and sex. The model computed stroke probability was quantitatively associated with facial features, various clinical parameters of blood clotting indicators and leukocyte counts, and, more importantly, stroke incidence in the near future. Our real-time facial image artificial intelligence model can be used to rapidly screen and prediagnose stroke before CT scanning, thus meeting the urgent need in emergency clinics, potentially translatable to routine monitoring.

Coordinate-wise monotonic transformations enable privacy-preserving age estimation with 3D face point cloud.

The human face is a valuable biomarker of aging, but the collection and use of its image raise significant privacy concerns. Here we present an approach for facial data masking that preserves age-related features using coordinate-wise monotonic transformations. We first develop a deep learning model that estimates age directly from non-registered face point clouds with high accuracy and generalizability. We show that the model learns a highly indistinguishable mapping using faces treated with coordinate-wise monotonic transformations, indicating that the relative positioning of facial information is a low-level biomarker of facial aging. Through visual perception tests and computational 3D face verification experiments, we demonstrate that transformed faces are significantly more difficult to perceive for human but not for machines, except when only the face shape information is accessible. Our study leads to a facial data protection guideline that has the potential to broaden public access to face datasets with minimized privacy risks.

Single-cell senescence identification reveals senescence heterogeneity, trajectory, and modulators.

Cellular senescence underlies many aging-related pathologies, but its heterogeneity poses challenges for studying and targeting senescent cells. We present here a machine learning program senescent cell identification (SenCID), which accurately identifies senescent cells in both bulk and single-cell transcriptome. Trained on 602 samples from 52 senescence transcriptome datasets spanning 30 cell types, SenCID identifies six major senescence identities (SIDs). Different SIDs exhibit different senescence baselines, stemness, gene functions, and responses to senolytics. SenCID enables the reconstruction of senescent trajectories under normal aging, chronic diseases, and COVID-19. Additionally, when applied to single-cell Perturb-seq data, SenCID helps reveal a hierarchy of senescence modulators. Overall, SenCID is an essential tool for precise single-cell analysis of cellular senescence, enabling targeted interventions against senescent cells.

Identification of the metaphyseal skeletal stem cell building trabecular bone.

Skeletal stem cells (SSCs) that are capable of self-renewal and multipotent differentiation contribute to bone development and homeostasis. Several populations of SSCs at different skeletal sites have been reported. Here, we identify a metaphyseal SSC (mpSSC) population whose transcriptional landscape is distinct from other bone mesenchymal stromal cells (BMSCs). These mpSSCs are marked by Sstr2 or Pdgfrb+Kitl-, located just underneath the growth plate, and exclusively derived from hypertrophic chondrocytes (HCs). These HC-derived mpSSCs have properties of self-renewal and multipotency in vitro and in vivo, producing most HC offspring postnatally. HC-specific deletion of Hgs, a component of the endosomal sorting complex required for transport, impairs the HC-to-mpSSC conversion and compromises trabecular bone formation. Thus, mpSSC is the major source of BMSCs and osteoblasts in bone marrow, supporting the postnatal trabecular bone formation.

Tissue-specific profiling of age-dependent miRNAomic changes in Caenorhabditis elegans.

Ageing exhibits common and distinct features in various tissues, making it critical to decipher the tissue-specific ageing mechanisms. MiRNAs are essential regulators in ageing and are recently highlighted as a class of intercellular messengers. However, little is known about the tissue-specific transcriptomic changes of miRNAs during ageing. C. elegans is a well-established model organism in ageing research. Here, we profile the age-dependent miRNAomic changes in five isolated worm tissues. Besides the diverse ageing-regulated miRNA expression across tissues, we discover numerous miRNAs in the tissues without their transcription. We further profile miRNAs in the extracellular vesicles and find that worm miRNAs undergo inter-tissue trafficking via these vesicles in an age-dependent manner. Using these datasets, we uncover the interaction between body wall muscle-derived mir-1 and DAF-16/FOXO in the intestine, suggesting mir-1 as a messenger in inter-tissue signalling. Taken together, we systematically investigate worm miRNAs in the somatic tissues and extracellular vesicles during ageing, providing a valuable resource to study tissue-autonomous and nonautonomous functions of miRNAs in ageing.

Application of AI in biological age prediction.

The development of anti-aging interventions requires quantitative measurement of biological age. Machine learning models, known as "aging clocks," are built by leveraging diverse aging biomarkers that vary across lifespan to predict biological age. In addition to traditional aging clocks harnessing epigenetic signatures derived from bulk samples, emerging technologies allow the biological age estimating at single-cell level to dissect cellular diversity in aging tissues. Moreover, imaging-based aging clocks are increasingly employed with the advantage of non-invasive measurement, making it suitable for large-scale human cohort studies. To fully capture the features in the ever-growing multi-modal and high-dimensional aging-related data and uncover disease associations, deep-learning based approaches, which are effective to learn complex and non-linear relationships without relying on pre-defined features, are increasingly applied. The use of big data and AI-based aging clocks has achieved high accuracy, interpretability and generalizability, guiding clinical applications to delay age-related diseases and extend healthy lifespans.

The ticking of aging clocks.

Computational models that measure biological age and aging rate regardless of chronological age are called aging clocks. The underlying counting mechanisms of the intrinsic timers of these clocks are still unclear. Molecular mediators and determinants of aging rate point to the key roles of DNA damage, epigenetic drift, and inflammation. Persistent DNA damage leads to cellular senescence and the senescence-associated secretory phenotype (SASP), which induces cytotoxic immune cell infiltration; this further induces DNA damage through reactive oxygen and nitrogen species (RONS). I discuss the possibility that DNA damage (or the response to it, including epigenetic changes) is the fundamental counting unit of cell cycles and cellular senescence, that ultimately accounts for cell composition changes and functional decline in tissues, as well as the key intervention points.

Spatial Reconstruction of Oligo and Single Cells by De Novo Coalescent Embedding of Transcriptomic Networks.

Single cell RNA-seq (scRNA-seq) profiles conceal temporal and spatial tissue developmental information. De novo reconstruction of single cell temporal trajectory has been fairly addressed, but reverse engineering single cell 3D spatial tissue organization is hitherto landmark based, and de novo spatial reconstruction is a compelling computational open problem. Here it is shown that a proposed algorithm for de novo coalescent embedding (D-CE) of oligo/single cell transcriptomic networks can help to address this problem. Relying on the spatial information encoded in the expression patterns of genes, it is found that D-CE of cell-cell association transcriptomic networks, by preserving mesoscale network organization, captures spatial domains, identifies spatially expressed genes, reconstructs cell samples' 3D spatial distribution, and uncovers spatial domains and markers necessary for understanding the design principles on spatial organization and pattern formation. Comparison to the novoSpaRC and CSOmap (the only available de novo 3D spatial reconstruction methods) on 14 datasets and 497 reconstructions, reveals a significantly superior performance of D-CE.

Human PBMC scRNA-seq-based aging clocks reveal ribosome to inflammation balance as a single-cell aging hallmark and super longevity.

Quantifying aging rate is important for evaluating age-associated decline and mortality. A blood single-cell RNA sequencing dataset for seven supercentenarians (SCs) was recently generated. Here, we generate a reference 28-sample aging cohort to compute a single-cell level aging clock and to determine the biological age of SCs. Our clock model placed the SCs at a blood biological age to between 80.43 and 102.67 years. Compared to the model-expected aging trajectory, SCs display increased naive CD8+ T cells, decreased cytotoxic CD8+ T cells, memory CD4+ T cells, and megakaryocytes. As the most prominent molecular hallmarks at the single-cell level, SCs contain more cells and cell types with high ribosome level, which is associated with and, according to Bayesian network inference, contributes to a low inflammation state and slow aging of SCs. Inhibiting ribosomal activity or translation in monocytes validates such translation against inflammation balance revealed by our single-cell aging clock.

Nuclear lamina erosion-induced resurrection of endogenous retroviruses underlies neuronal aging.

The primate frontal lobe (FL) is sensitive to aging-related neurocognitive decline. However, the aging-associated molecular mechanisms remain unclear. Here, using physiologically aged non-human primates (NHPs), we depicted a comprehensive landscape of FL aging with multidimensional profiling encompassing bulk and single-nucleus transcriptomes, quantitative proteome, and DNA methylome. Conjoint analysis across these molecular and neuropathological layers underscores nuclear lamina and heterochromatin erosion, resurrection of endogenous retroviruses (ERVs), activated pro-inflammatory cyclic GMP-AMP synthase (cGAS) signaling, and cellular senescence in post-mitotic neurons of aged NHP and human FL. Using human embryonic stem-cell-derived neurons recapitulating cellular aging in vitro, we verified the loss of B-type lamins inducing resurrection of ERVs as an initiating event of the aging-bound cascade in post-mitotic neurons. Of significance, these aging-related cellular and molecular changes can be alleviated by abacavir, a nucleoside reverse transcriptase inhibitor, either through direct treatment of senescent human neurons in vitro or oral administration to aged mice.

LncRNAs: the missing link to senescence nuclear architecture.

During cellular senescence and organismal aging, cells display various molecular and morphological changes. Although many aging-related long noncoding RNAs (lncRNAs) are highly associated with senescence-associated secretory phenotype, the roles of lncRNAs in senescence-associated nuclear architecture and morphological changes are just starting to emerge. Here I review lncRNAs associated with nuclear structure establishment and maintenance, their aging-related changes, and then focus on the pervasive, yet underappreciated, role of RNA double-strand DNA triplexes for lncRNAs to recognize targeted genomic regions, making lncRNAs the nexus between DNA and proteins to regulate nuclear structural changes. Finally, I discuss the future of deciphering direct links of lncRNA changes to various nuclear morphology changes assisted by artificial intelligence and genetic perturbations.

Transformer for one stop interpretable cell type annotation.

Consistent annotation transfer from reference dataset to query dataset is fundamental to the development and reproducibility of single-cell research. Compared with traditional annotation methods, deep learning based methods are faster and more automated. A series of useful single cell analysis tools based on autoencoder architecture have been developed but these struggle to strike a balance between depth and interpretability. Here, we present TOSICA, a multi-head self-attention deep learning model based on Transformer that enables interpretable cell type annotation using biologically understandable entities, such as pathways or regulons. We show that TOSICA achieves fast and accurate one-stop annotation and batch-insensitive integration while providing biologically interpretable insights for understanding cellular behavior during development and disease progressions. We demonstrate TOSICA's advantages by applying it to scRNA-seq data of tumor-infiltrating immune cells, and CD14+ monocytes in COVID-19 to reveal rare cell types, heterogeneity and dynamic trajectories associated with disease progression and severity.

KCNQ1OT1 promotes genome-wide transposon repression by guiding RNA-DNA triplexes and HP1 binding.

Transposon (de)repression and heterochromatin reorganization are dynamically regulated during cell fate determination and are hallmarks of cellular senescence. However, whether they are sequence specifically regulated remains unknown. Here we uncover that the KCNQ1OT1 lncRNA, by sequence-specific Hoogsteen base pairing with double-stranded genomic DNA via its repeat-rich region and binding to the heterochromatin protein HP1α, guides, induces and maintains epigenetic silencing at specific repetitive DNA elements. Repressing KCNQ1OT1 or deleting its repeat-rich region reduces DNA methylation and H3K9me3 on KCNQ1OT1-targeted transposons. Engineering a fusion KCNQ1OT1 with an ectopically targeting guiding triplex sequence induces de novo DNA methylation at the target site. Phenotypically, repressing KCNQ1OT1 induces senescence-associated heterochromatin foci, transposon activation and retrotransposition as well as cellular senescence, demonstrating an essential role of KCNQ1OT1 to safeguard against genome instability and senescence.

3D chromatin structure changes during spermatogenesis and oogenesis.

Gametogenesis, including spermatogenesis and oogenesis, are unique differentiation processes involving extraordinarily complex and precise regulatory mechanisms that require the interactions of multiple cell types, hormones, paracrine factors, genes and epigenetic regulators, and extensive chromatin 3D structure re-organization. In recent years, the development of 3D genome technology represented by Hi-C, enabled mapping of the 3D re-organization of chromosomes during zygogenesis at an unprecedented resolution. The 3D remodeling is achieved by folding chromatin into loops, topologically associating domains (TADs), and compartments (A and B), which ultimately affect transcriptional activity. In this review, we summarize the research progresses and findings on chromatin 3D structure changes during spermatogenesis and oogenesis.

Distinct biological ages of organs and systems identified from a multi-omics study.

Biological age (BA) has been proposed to evaluate the aging status instead of chronological age (CA). Our study shows evidence that there might be multiple "clocks" within the whole-body system: systemic aging drivers/clocks overlaid with organ/tissue-specific counterparts. We utilize multi-omics data, including clinical tests, immune repertoire, targeted metabolomic molecules, gut microbiomes, physical fitness examinations, and facial skin examinations, to estimate the BA of different organs (e.g., liver, kidney) and systems (immune and metabolic system). The aging rates of organs/systems are diverse. People's aging patterns are different. We also demonstrate several applications of organs/systems BA in two independent datasets. Mortality predictions are compared among organs' BA in the dataset of the United States National Health and Nutrition Examination Survey. Polygenic risk score of BAs constructed in the Chinese Longitudinal Healthy Longevity Survey cohort can predict the possibility of becoming centenarian.

Ageing induces tissue-specific transcriptomic changes in Caenorhabditis elegans.

Ageing is a complex process with common and distinct features across tissues. Unveiling the underlying processes driving ageing in individual tissues is indispensable to decipher the mechanisms of organismal longevity. Caenorhabditis elegans is a well-established model organism that has spearheaded ageing research with the discovery of numerous genetic pathways controlling its lifespan. However, it remains challenging to dissect the ageing of worm tissues due to the limited description of tissue pathology and access to tissue-specific molecular changes during ageing. In this study, we isolated cells from five major tissues in young and old worms and profiled the age-induced transcriptomic changes within these tissues. We observed a striking diversity of ageing across tissues and identified different sets of longevity regulators therein. In addition, we found novel tissue-specific factors, including irx-1 and myrf-2, which control the integrity of the intestinal barrier and sarcomere structure during ageing respectively. This study demonstrates the complexity of ageing across worm tissues and highlights the power of tissue-specific transcriptomic profiling during ageing, which can serve as a resource to the field.

Cross-species metabolomic analysis identifies uridine as a potent regeneration promoting factor.

Regenerative capacity declines throughout evolution and with age. In this study, we asked whether metabolic programs underlying regenerative capability might be conserved across species, and if so, whether such metabolic drivers might be harnessed to promote tissue repair. To this end, we conducted metabolomic analyses in two vertebrate organ regeneration models: the axolotl limb blastema and antler stem cells. To further reveal why young individuals have higher regenerative capacity than the elderly, we also constructed metabolic profiles for primate juvenile and aged tissues, as well as young and aged human stem cells. In joint analyses, we uncovered that active pyrimidine metabolism and fatty acid metabolism correlated with higher regenerative capacity. Furthermore, we identified a set of regeneration-related metabolite effectors conserved across species. One such metabolite is uridine, a pyrimidine nucleoside, which can rejuvenate aged human stem cells and promote regeneration of various tissues in vivo. These observations will open new avenues for metabolic intervention in tissue repair and regeneration.

Immunosuppression Induced by Brain-Specific HDAC6 Knockdown Improves Aging Performance in Drosophila melanogaster.

HDAC6 is involved in several biological processes related to aging-associated diseases. However, it was unknown whether HDAC6 could directly regulate lifespan and healthspan. We found that HDAC6 knockdown induced transcriptome changes to attenuate the aging changes in the Drosophila head, particularly on the inflammation and innate immunity-related genes. Whole-body knockdown of HDAC6 extended lifespan in the fly, furthermore brain-specific knockdown of HDAC6 extended both lifespan and healthspan in the fly. Our results established HDAC6 as a lifespan regulator and provided a potential anti-aging target. Supplementary Information: The online version contains supplementary material available at 10.1007/s43657-022-00045-2.