Multiscale topology classifies cells in subcellular spatial transcriptomics.

Spatial transcriptomics measures in situ gene expression at millions of locations within a tissue1, hitherto with some trade-off between transcriptome depth, spatial resolution and sample size2. Although integration of image-based segmentation has enabled impactful work in this context, it is limited by imaging quality and tissue heterogeneity. By contrast, recent array-based technologies offer the ability to measure the entire transcriptome at subcellular resolution across large samples3-6. Presently, there exist no approaches for cell type identification that directly leverage this information to annotate individual cells. Here we propose a multiscale approach to automatically classify cell types at this subcellular level, using both transcriptomic information and spatial context. We showcase this on both targeted and whole-transcriptome spatial platforms, improving cell classification and morphology for human kidney tissue and pinpointing individual sparsely distributed renal mouse immune cells without reliance on image data. By integrating these predictions into a topological pipeline based on multiparameter persistent homology7-9, we identify cell spatial relationships characteristic of a mouse model of lupus nephritis, which we validate experimentally by immunofluorescence. The proposed framework readily generalizes to new platforms, providing a comprehensive pipeline bridging different levels of biological organization from genes through to tissues.

Cell-type-directed design of synthetic enhancers.

Transcriptional enhancers act as docking stations for combinations of transcription factors and thereby regulate spatiotemporal activation of their target genes1. It has been a long-standing goal in the field to decode the regulatory logic of an enhancer and to understand the details of how spatiotemporal gene expression is encoded in an enhancer sequence. Here we show that deep learning models2-6, can be used to efficiently design synthetic, cell-type-specific enhancers, starting from random sequences, and that this optimization process allows detailed tracing of enhancer features at single-nucleotide resolution. We evaluate the function of fully synthetic enhancers to specifically target Kenyon cells or glial cells in the fruit fly brain using transgenic animals. We further exploit enhancer design to create 'dual-code' enhancers that target two cell types and minimal enhancers smaller than 50 base pairs that are fully functional. By examining the state space searches towards local optima, we characterize enhancer codes through the strength, combination and arrangement of transcription factor activator and transcription factor repressor motifs. Finally, we apply the same strategies to successfully design human enhancers, which adhere to enhancer rules similar to those of Drosophila enhancers. Enhancer design guided by deep learning leads to better understanding of how enhancers work and shows that their code can be exploited to manipulate cell states.

Aplicação de ferramentas de quimioinformática na comparação da estrutura tridimensional da dihidrofolato redutase expressa em diferentes organismos / Application of cheminformatics tools for comparing dihydrofolate reductase's tridimensional structure expressed by different organisms

A quimioinformática, definida como o emprego de técnicas informáticas na solução de problemas da química, evolui em conjunto com o desenvolvimento de ferramentas computacionais e é de grande relevância para o planejamento racional de fármacos ao otimizar etapas do desenvolvimento de novas moléculas e economizar recursos e tempo. Dentre as técnicas disponíveis destacam-se o planejamento de fármacos baseado na estrutura e no ligante, que quando combinadas auxiliam na identificação e otimização de moléculas ativas frente a alvos farmacológicos. A Dihidrofolato Redutase (DHFR) é uma importante enzima da via dos folatos que catalisa a redução do dihidrofolato em tetrahidrofolato, utilizando NADPH como cofator, reação essencial para a replicação celular, visto que este ciclo resulta na síntese de precursores das bases nitrogenadas que compõem o DNA, consequentemente, inibidores de DHFR são utilizados no tratamento de infecções bacterianas e alguns tipos de câncer. Trypanosoma cruzi, protozoário causador da doença de chagas, é um dos organismos que expressam a DHFR, além do próprio Homo sapiens. Analisaram-se ligantes conhecidos e as estruturas da proteína expressa pelos dois organismos, visando identificar pontos de divergência que possam ser explorados no planejamento de moléculas seletivas para o tratamento da doença de Chagas. Os 6 modelos cristalográficos de T. cruzi e 2 de H. sapiens foram obtidos do banco de dados de proteínas (PDB) após aplicação de filtros de qualidade. Foram analisadas as sequências de aminoácidos dos modelos, com o uso do Cluster Ômega, sua estrutura tridimensional com os programas Pymol e Chimera X, além da análise das cavidades proteicas com o CavityPlus, que também gerou os farmacóforos de ambos alvos. A análise de estrutura primária identificou mutações em três aminoácidos nos cristais do parasita, que podem ser explicados por diferentes caminhos evolutivos de grupos segregados, embora nenhuma mutação observada esteja em regiões de sítio ativo. A análise dos modelos permitiu que fossem identificados os 25 aminoácidos que estão a menos de 5 Å de distância dos ligantes de T. cruzi, sendo 5 aminoácidos responsáveis por interações de hidrogênio com pelo menos um dos ligantes analisados. Destes, 18 se repetem na proteína humana ou são substituídos por outro aminoácido que mantém a mesma interação. Quanto às diferenças observadas, destacam-se a asparagina 44 substituída por uma prolina na proteína humana e a prolina 92, substituída por uma lisina. A análise de cavidades identificou três cavidades em cada proteína, embora somente as cavidades correspondentes ao sítio ativo sejam druggables. A cavidade da proteína humana é maior e mais alongada, além de apresentar o aspecto de um túnel, enquanto a cavidade da proteína parasita é mais aberta, tal abertura permite que ligantes com o anel benzeno meta substituídos explorem uma região existente na cavidade de T. cruzi que é fechada na humana. O farmacóforo de ambas proteínas foi identificado, apresentando diferenças no tamanho e angulação que também podem ser explorados no planejamento de fármacos seletivos

Chemoinformatic, defined as the use of informatic techniques to solve chemical problems, has evolved together with new computational tools and it is quite important for rational drug designing, by optimizing different steps on the development pipeline of new molecules, saving resources and time. From all the available tools, structure and ligand based drug design shall be highlighted, when combined, they support the identification and optimization of active molecules from pharmaceutical targets. Dihydrofolate reductase (DHFR) is an important enzyme of the folate pathway that catalyzes the reduction of dihydrofolate to tetrahydrofolate, by using NADPH as cofactor. This reaction is essential for cell replication, as this pathway results in the synthesis of nucleobases that build the DNA. That's the reason why DHFR inhibitors are used for treating bacterial infections and some types of cancer. Trypanosoma cruzi, a protozoa that causes Chagas disease, is one of the organisms that express DHFR, besides Homo sapiens itself. This work analyzed known ligands and the structure of the protein expressed by both organisms, aiming to identify divergence points that could be explored for designing selective drugs for Chagas disease treatment. The 6 proteins crystallographic models from T. cruzi and 2 from H. sapiens were obtained from protein data bank (PDB) after the application of quality filters. The amino acid sequence of each model was analyzed by Clustal Omega, its tridimensional structure by Pymol and Chimera X and the cavity analysis by CavityPlus, that also generated the pharmacophore from both targets. The primary structure analysis identified mutations on three amino acids on the parasite christal, which may be explained by different evolutive paths from segregated groups, although none of the observed mutations are on the active site region. The model's analysis allowed the identification of 24 amino acids that are closer than 5 Å from the T. cruzi ligands, 5 of them responsible for hydrogen interactions on at least one of the ligands analyzed. 18 of them are repeated on the human protein or are replaced by another amino acid that preserves the same interaction. As by the differences observed that shall be highlighted, asparagine 44 is replaced by a proline on the human protein, and proline 92 by a lysin. The cavity analysis identified three cavities on each protein, although only the cavities of the active site are druggables. The human protein cavity is bigger and longer, besides it looks like its a tunnel, when the parasite protein is open, that opening allows ligands with benzene ring meta substituted to explore the existing regions of the T. cruzi protein that is closed on the human protein. Lastly, the pharmacophore from both proteins was identified, it shows differences on size and angulation that also could be explored in the designing of selective drugs

nteraction between FAK/αB-crystalline is important for viability of the glioblastoma cells

Abstract Glioblastoma multiforme is a tumor of the central nervous system. Focal Adhesion Kinase (FAK) and αB-crystalline are two proteins involved in glioblastoma development. In this study, we investigated whether the FAK/αB-crystalline interaction is important for glioblastoma cells, we aimed to investigate the interaction of these two proteins in the glioblastoma multiforme cell line U87-MG. Two peptides named FP01 peptide (derived from αB-crystalline) and FP02 peptide (derived from FAK) were synthesized for this study. Treatment of U87-MG with the peptides FP01 and FP02 in the concentration at 50 µM reduced the viability cellular to around 41% and 51%, respectively. Morphological alterations in the cells treated with the peptides when compared to the control were observed. This study suggests that the interaction between FAK and αB-crystalline is important for the viability of glioblastoma cells

What is a cell type?

A next step for cell atlases should be to chart perturbations in human model systems.

A DNA methylation atlas of normal human cell types.

DNA methylation is a fundamental epigenetic mark that governs gene expression and chromatin organization, thus providing a window into cellular identity and developmental processes1. Current datasets typically include only a fraction of methylation sites and are often based either on cell lines that underwent massive changes in culture or on tissues containing unspecified mixtures of cells2-5. Here we describe a human methylome atlas, based on deep whole-genome bisulfite sequencing, allowing fragment-level analysis across thousands of unique markers for 39 cell types sorted from 205 healthy tissue samples. Replicates of the same cell type are more than 99.5% identical, demonstrating the robustness of cell identity programmes to environmental perturbation. Unsupervised clustering of the atlas recapitulates key elements of tissue ontogeny and identifies methylation patterns retained since embryonic development. Loci uniquely unmethylated in an individual cell type often reside in transcriptional enhancers and contain DNA binding sites for tissue-specific transcriptional regulators. Uniquely hypermethylated loci are rare and are enriched for CpG islands, Polycomb targets and CTCF binding sites, suggesting a new role in shaping cell-type-specific chromatin looping. The atlas provides an essential resource for study of gene regulation and disease-associated genetic variants, and a wealth of potential tissue-specific biomarkers for use in liquid biopsies.

Automated human cell classification in sparse datasets using few-shot learning.

Classifying and analyzing human cells is a lengthy procedure, often involving a trained professional. In an attempt to expedite this process, an active area of research involves automating cell classification through use of deep learning-based techniques. In practice, a large amount of data is required to accurately train these deep learning models. However, due to the sparse human cell datasets currently available, the performance of these models is typically low. This study investigates the feasibility of using few-shot learning-based techniques to mitigate the data requirements for accurate training. The study is comprised of three parts: First, current state-of-the-art few-shot learning techniques are evaluated on human cell classification. The selected techniques are trained on a non-medical dataset and then tested on two out-of-domain, human cell datasets. The results indicate that, overall, the test accuracy of state-of-the-art techniques decreased by at least 30% when transitioning from a non-medical dataset to a medical dataset. Reptile and EPNet were the top performing techniques tested on the BCCD dataset and HEp-2 dataset respectively. Second, this study evaluates the potential benefits, if any, to varying the backbone architecture and training schemes in current state-of-the-art few-shot learning techniques when used in human cell classification. To this end, the best technique identified in the first part of this study, EPNet, is used for experimentation. In particular, the study used 6 different network backbones, 5 data augmentation methodologies, and 2 model training schemes. Even with these additions, the overall test accuracy of EPNet decreased from 88.66% on non-medical datasets to 44.13% at best on the medical datasets. Third, this study presents future directions for using few-shot learning in human cell classification. In general, few-shot learning in its current state performs poorly on human cell classification. The study proves that attempts to modify existing network architectures are not effective and concludes that future research effort should be focused on improving robustness towards out-of-domain testing using optimization-based or self-supervised few-shot learning techniques.

In vitro evaluation of the cytotoxicity and eye irritation potential of preservatives widely used in cosmetics

Abstract The consumption of cosmetics has been increasing every year and is expected to reach $675 billion by 2020 at an estimated growth rate of 6.4% per year. Exposure to skin irritants is the major cause of non-immunological inflammation of the skin. Therefore, the safety evaluation of cosmetic preservatives should be increased. Thus, the present work aimed to evaluate the cytotoxicity as the viability endpoint and the eye irritation potential of preservatives used in cosmetics. Cytotoxicity assays were performed using MTT and NRU in human keratinocytes (HaCaT), human dermal fibroblasts, adult (HDFa), and human hepatoma cells (HepG2). The eye irritation potential was evaluated using the Hen's Egg Test-chorioallantoic membrane (HET-CAM). The evaluated preservatives were methylparaben (MP), propylparaben (PP), phenoxyethanol (PE), and a mixture of methylchloroisothiazolinone and methylisothiazolinone (CMI/MI). All preservatives showed cytotoxic potential within the permitted concentrations for use in cosmetic products. In the HET-CAM test, PE and CMI/MI, MP, and PP were classified as severe, moderate, and poor irritants, respectively. Our results indicate that proper safety evaluations are required to ensure the beneficial properties of preservatives on cosmetic products without exceeding exposure levels that would result in adverse health effects for consumers.

Role of cytochrome P450 1A2 and N-acetyltransferase 2 in 2, 6-dimethylaniline induced genotoxicity

Abstract The purpose of the current work was to assess a possible role of cytochrome P450 1A2 (CYP1A2) and N-acetyltransferase 2 (NAT2) in the metabolic activation of 2,6-dimethylaniline (2,6-DMA) and also clarify the function of DNA repair in affecting the ultimate mutagenic potency. Two cell lines, nucleotide excision repair (NER)-deficient 5P3NAT2 and proficient 5P3NAT2R9 both expressing CYP1A2 and NAT2, were treated with 2,6-DMA for 48 h or its metabolites for 1 h. Cell survival determined by trypan blue exclusion and MTT assays, and 8-azaadenine-resistant mutants at the adenine phosphoribosyltransferase (aprt) gene locus were evaluated. 5P3NAT2 and 5P3NAT2R9 cells treated with 2,6-DMA and its metabolites showed a dose-dependent increase in cytotoxicity and mutant fraction; N-OH-2,6-DMA and 2,6-DMAP in serum-free α-minimal essential medium (MEM) are more potent than 2,6-DMA in complete MEM. 5P3NAT2 cells was more sensitive to the cytotoxic and mutagenic action than 5P3NAT2R9 cells. H2DCFH-DA assay showed dose-dependent ROS production under 2,6- DMAP treatment. These findings indicate that the genotoxic effects of 2,6-DMA are mediated by CYP1A2 activation via N-hydroxylation and the subsequent esterification by the phase II conjugation enzyme NAT2, and through the generation of ROS by hydroxylamine and/or aminophenol metabolites. NER status is also an important contributor

Análise proteômica de células BEAS-2B expostas a benzo[α]pireno e nicotinamida ribosídeo / Proteomic Analysis of BEAS-2B cells exposed to benzo[α]pyrene and nicotinamide riboside

Responsável por milhões de óbitos anuais e um grande custo para a saúde pública, o câncer é a segunda maior causa de mortes no mundo. Dentre seus diversos tipos, o câncer de pulmão, além da alta incidência, é um dos mais letais. A exposição a substâncias tóxicas provenientes da combustão de matéria orgânica, assim como o consumo de cigarro, são os principais responsáveis pela alta incidência de câncer de pulmão. Dentre estas substâncias, está o benzo[α]pireno (B[α]P), um carcinógeno completo, ou seja, capaz de iniciar e promover o processo de carcinogênese. Resultados anteriores obtidos pelo grupo demonstraram que células BEAS-2B expostas a 1 µM de B[α]P apresentaram alterações das concentrações de metabólitos intracelulares, indução de estresse redox e hipermetilação do DNA. A exposição a 1 µM de nicotinamida ribosídeo (NR), um dos precursores de NAD+, foi capaz de proteger as células BEAS-2B contra a transformação induzida por B[α]P, além de impedir totalmente que células não expostas a B[α]P formassem colônias em soft-agar. A utilização da proteômica neste trabalho permitiu verificar a abundância das proteínas nos quatro diferentes grupos de exposição: Controle, B[α]P, B[α]P + NR e NR. Após 120 h de exposição as células foram coletadas, as proteínas extraídas e preparadas para análise. Foram descobertas 3024 proteínas posteriormente analisadas com o objetivo de elucidar vias possivelmente envolvidas na proteção contra o processo de transfomação maligna. Os grupos NR e Controle demonstram ser mais parecidos em relação ao seu conteúdo, enquanto os grupos B[α]P e B[α]P + NR foram mais semelhantes entre si. A análise de proteínas exclusivas revelou menos processos relacionados ao reparo de DNA no grupo tratado apenas com B[α]P quando comparado com B[α]P + NR. A análise estatística do total de proteínas utilizando o teste ANOVA (p < 0,05, N = 5) revelou 564 proteínas diferencialmente expressas entre os grupos. A clusterização nos permitiu observar a diferença na abundância de proteínas entre os quatro tratamentos. As proteínas estão envolvidas em funções como a regulação do metabolismo, resposta a estresse, transdução de sinal, regulação de expressão gênica e morte celular. Um dos clusters (cluster 1), contendo 59 proteínas, revelou poucos processos na análise de enriquecimento, mas as proteínas contidas nele apresentam funções como controle da divisão celular, apoptose e proteção ao estresse redox. Nele podemos observar que, no geral, o tratamento com B[α]P aumentou a abundância de algumas proteínas, o que foi revertido no grupo B[α]P + NR. O tratamento apenas com NR diminuiu a abundância das proteínas contidas nesse cluster. Outro cluster (cluster 4) apresentou 51 proteínas de abundância diminuída durante a exposição ao B[α]P, o que se reverteu no grupo B[α]P + NR. As proteínas desse cluster estão envolvidas em etapas importantes da via glicolítica, de crescimento, adesão, migração e invasão celular. Apesar de ser descrito que a exposição a NR pode aumentar a eficiência do reparo de DNA, os resultados apresentados nesse trabalho indicam que o efeito protetor pode estar relacionado com a modulação do ciclo celular ou alterações na adesão celular

Responsible for millions of annual deaths and a great health expense, cancer is the second leading cause of death in the world. Among its many types, lung cancer, besides its high incidence, is also one of the most lethal. Exposure to toxic substances resulting from the combustion of organic matter, as well as cigarette consumption, are the mainly responsible for the high incidence of lung cancer. One of these substances is benzo[α]pyrene (B[α]P), a complete carcinogen, able to initiate and promote the carcinogenesis process. Results obtained previously demonstrated that BEAS-2B cells exposed to 1 µM BaP presented alterations in the levels of intracellular metabolites, induction of oxidative stress, and hypermethylation of DNA. The exposure to 1 µM nicotinamide riboside (NR), one of the precursors of NAD+, was able to protect BEAS-2B cells against the transformation induced by B[α]P, moreover, it also totally prevented the colonies formation on soft agar in cells not exposed to B[α]P. The use of proteomics allowed us to verify the abundance of proteins in the four different exposure groups: Control, B[α]P, B[α]P + NR e NR. After 120h of exposure, the cells were collected followed by the extraction of the proteins. A total of 3024 proteins were identified and analyzed aiming to elucidate possible pathways involved in the protective effect against the malignant transformation induced by B[α]P. The NR and Control groups showed to be more similar, while B[α]P and B[α]P + NR were more similar. The analysis of exclusive proteins revealed fewer processes related to DNA repair in B[α]P when compared with B[α]P + NR. The statistical analysis of the total proteins using the ANOVA test (p <0.5, N = 5) revealed 564 proteins differentially expressed between the groups. The heatmap showed the difference in protein abundance between the four treatments. Proteins are involved in functionssuch asthe regulation of metabolism, stress response, signal transduction, regulation of gene expression, and cell death. One of the clusters (cluster 1), containing 59 proteins, revealed a few processes in the enrichment analysis, but the proteins contained in it have functions such as control of cell division, apoptosis, and protection from redox stress. It is possible to observe, in general, treatment with B[α]P increased the abundance of some proteins, which was partially reversed in group B[α]P + NR. On the other hand, the NR treatment decreased the abundance of proteins contained in this cluster. Another cluster (cluster 4) showed 51 proteins of decreased abundance during exposure to B [α] P, which was partially reversed in group B[α]P + NR. The proteins in this cluster are involved in important stages of the glycolytic pathway, also in growth, adhesion, migration, and cell invasion. Although it has been described that exposure to NR can increase the efficiency of DNA repair, the results presented in this work indicate that the protective effect may be related to the modulation of the cell cycle or cell adehsion modifications

Cell-type specialization is encoded by specific chromatin topologies.

The three-dimensional (3D) structure of chromatin is intrinsically associated with gene regulation and cell function1-3. Methods based on chromatin conformation capture have mapped chromatin structures in neuronal systems such as in vitro differentiated neurons, neurons isolated through fluorescence-activated cell sorting from cortical tissues pooled from different animals and from dissociated whole hippocampi4-6. However, changes in chromatin organization captured by imaging, such as the relocation of Bdnf away from the nuclear periphery after activation7, are invisible with such approaches8. Here we developed immunoGAM, an extension of genome architecture mapping (GAM)2,9, to map 3D chromatin topology genome-wide in specific brain cell types, without tissue disruption, from single animals. GAM is a ligation-free technology that maps genome topology by sequencing the DNA content from thin (about 220 nm) nuclear cryosections. Chromatin interactions are identified from the increased probability of co-segregation of contacting loci across a collection of nuclear slices. ImmunoGAM expands the scope of GAM to enable the selection of specific cell types using low cell numbers (approximately 1,000 cells) within a complex tissue and avoids tissue dissociation2,10. We report cell-type specialized 3D chromatin structures at multiple genomic scales that relate to patterns of gene expression. We discover extensive 'melting' of long genes when they are highly expressed and/or have high chromatin accessibility. The contacts most specific of neuron subtypes contain genes associated with specialized processes, such as addiction and synaptic plasticity, which harbour putative binding sites for neuronal transcription factors within accessible chromatin regions. Moreover, sensory receptor genes are preferentially found in heterochromatic compartments in brain cells, which establish strong contacts across tens of megabases. Our results demonstrate that highly specific chromatin conformations in brain cells are tightly related to gene regulation mechanisms and specialized functions.

Cell type ontologies of the Human Cell Atlas.

Massive single-cell profiling efforts have accelerated our discovery of the cellular composition of the human body while at the same time raising the need to formalize this new knowledge. Here, we discuss current efforts to harmonize and integrate different sources of annotations of cell types and states into a reference cell ontology. We illustrate with examples how a unified ontology can consolidate and advance our understanding of cell types across scientific communities and biological domains.

Anatomical structures, cell types and biomarkers of the Human Reference Atlas.

The Human Reference Atlas (HRA) aims to map all of the cells of the human body to advance biomedical research and clinical practice. This Perspective presents collaborative work by members of 16 international consortia on two essential and interlinked parts of the HRA: (1) three-dimensional representations of anatomy that are linked to (2) tables that name and interlink major anatomical structures, cell types, plus biomarkers (ASCT+B). We discuss four examples that demonstrate the practical utility of the HRA.

Impact of Senescent Cell Subtypes on Tissue Dysfunction and Repair: Importance and Research Questions.

Cellular senescence, first observed and defined through cell culture studies, is a cell fate associated with essentially permanent cell cycle arrest and that can be triggered by a variety of inducers. Emerging evidence suggests senescence is a dynamic process with diverse functional characteristics. Depending on the tissue, type of inducer, and time since induction, senescent cells can promote tissue repair and re-modeling, prevent tumor development, or contribute to age-related disorders and chronic diseases, including cancers. Senescent cell characteristics appear to depend on multiple factors and be influenced by the milieu and other senescent cells locally and at a distance. We review diverse phenotypes of senescent cells originating from different cell types, senescence inducers over time since induction of senescence, and across conditions and diseases. This background is essential to inform further understanding about senescent cell subtypes and will point towards rational senescence-modulating strategies for achieving therapeutic benefit.

Evolutionary cell type mapping with single-cell genomics.

A fundamental characteristic of animal multicellularity is the spatial coexistence of functionally specialized cell types that are all encoded by a single genome sequence. Cell type transcriptional programs are deployed and maintained by regulatory mechanisms that control the asymmetric, differential access to genomic information in each cell. This genome regulation ultimately results in specific cellular phenotypes. However, the emergence, diversity, and evolutionary dynamics of animal cell types remain almost completely unexplored beyond a few species. Single-cell genomics is emerging as a powerful tool to build comprehensive catalogs of cell types and their associated gene regulatory programs in non-traditional model species. We review the current state of sampling efforts across the animal tree of life and challenges ahead for the comparative study of cell type programs. We also discuss how the phylogenetic integration of cell atlases can lead to the development of models of cell type evolution and a phylogenetic taxonomy of cells.

Hierarchical progressive learning of cell identities in single-cell data.

Supervised methods are increasingly used to identify cell populations in single-cell data. Yet, current methods are limited in their ability to learn from multiple datasets simultaneously, are hampered by the annotation of datasets at different resolutions, and do not preserve annotations when retrained on new datasets. The latter point is especially important as researchers cannot rely on downstream analysis performed using earlier versions of the dataset. Here, we present scHPL, a hierarchical progressive learning method which allows continuous learning from single-cell data by leveraging the different resolutions of annotations across multiple datasets to learn and continuously update a classification tree. We evaluate the classification and tree learning performance using simulated as well as real datasets and show that scHPL can successfully learn known cellular hierarchies from multiple datasets while preserving the original annotations. scHPL is available at https://github.com/lcmmichielsen/scHPL .