Mapping information-rich genotype-phenotype landscapes with genome-scale Perturb-seq

CODEX: COunterfactual Deep learning for the in silico EXploration of cancer cell line perturbations

MOTIVATION

High-throughput screens (HTS) provide a powerful tool to decipher the causal effects of chemical and genetic perturbations on cancer cell lines. Their ability to evaluate a wide spectrum of interventions, from single drugs to intricate drug combinations and CRISPR-interference, has established them as an invaluable resource for the development of novel therapeutic approaches. Nevertheless, the combinatorial complexity of potential interventions makes a comprehensive exploration intractable. Hence, prioritizing interventions for further experimental investigation becomes of utmost importance.

RESULTS

We propose CODEX (COunterfactual Deep learning for the in silico EXploration of cancer cell line perturbations) as a general framework for the causal modeling of HTS data, linking perturbations to their downstream consequences. CODEX relies on a stringent causal modeling strategy based on counterfactual reasoning. As such, CODEX predicts drug-specific cellular responses, comprising cell survival and molecular alterations, and facilitates the in silico exploration of drug combinations. This is achieved for both bulk and single-cell HTS. We further show that CODEX provides a rationale to explore complex genetic modifications from CRISPR-interference in silico in single cells.

AVAILABILITY AND IMPLEMENTATION

Our implementation of CODEX is publicly available at https://github.com/sschrod/CODEX. All data used in this article are publicly available.

AttentionPert: accurately modeling multiplexed genetic perturbations with multi-scale effects

MOTIVATION

Genetic perturbations (e.g. knockouts, variants) have laid the foundation for our understanding of many diseases, implicating pathogenic mechanisms and indicating therapeutic targets. However, experimental assays are fundamentally limited by the number of measurable perturbations. Computational methods can fill this gap by predicting perturbation effects under novel conditions, but accurately predicting the transcriptional responses of cells to unseen perturbations remains a significant challenge.

RESULTS

We address this by developing a novel attention-based neural network, AttentionPert, which accurately predicts gene expression under multiplexed perturbations and generalizes to unseen conditions. AttentionPert integrates global and local effects in a multi-scale model, representing both the nonuniform system-wide impact of the genetic perturbation and the localized disturbance in a network of gene-gene similarities, enhancing its ability to predict nuanced transcriptional responses to both single and multi-gene perturbations. In comprehensive experiments, AttentionPert demonstrates superior performance across multiple datasets outperforming the state-of-the-art method in predicting differential gene expressions and revealing novel gene regulations. AttentionPert marks a significant improvement over current methods, particularly in handling the diversity of gene perturbations and in predicting out-of-distribution scenarios.

AVAILABILITY AND IMPLEMENTATION

Code is available at https://github.com/BaiDing1234/AttentionPert.

Predicting single-cell cellular responses to perturbations using cycle consistency learning

SUMMARY

Phenotype-based drug screening emerges as a powerful approach for identifying compounds that actively interact with cells. Transcriptional and proteomic profiling of cell lines and individual cells provide insights into the cellular state alterations that occur at the molecular level in response to external perturbations, such as drugs or genetic manipulations. In this paper, we propose cycleCDR, a novel deep learning framework to predict cellular response to external perturbations. We leverage the autoencoder to map the unperturbed cellular states to a latent space, in which we postulate the effects of drug perturbations on cellular states follow a linear additive model. Next, we introduce the cycle consistency constraints to ensure that unperturbed cellular state subjected to drug perturbation in the latent space would produces the perturbed cellular state through the decoder. Conversely, removal of perturbations from the perturbed cellular states can restore the unperturbed cellular state. The cycle consistency constraints and linear modeling in the latent space enable to learn transferable representations of external perturbations, so that our model can generalize well to unseen drugs during training stage. We validate our model on four different types of datasets, including bulk transcriptional responses, bulk proteomic responses, and single-cell transcriptional responses to drug/gene perturbations. The experimental results demonstrate that our model consistently outperforms existing state-of-the-art methods, indicating our method is highly versatile and applicable to a wide range of scenarios.

AVAILABILITY AND IMPLEMENTATION

The source code is available at: https://github.com/hliulab/cycleCDR.

Multiplexed single-cell characterization of alternative polyadenylation regulators

Most mammalian genes have multiple polyA sites, representing a substantial source of transcript diversity regulated by the cleavage and polyadenylation (CPA) machinery. To better understand how these proteins govern polyA site choice, we introduce CPA-Perturb-seq, a multiplexed perturbation screen dataset of 42 CPA regulators with a 3' scRNA-seq readout that enables transcriptome-wide inference of polyA site usage. We develop a framework to detect perturbation-dependent changes in polyadenylation and characterize modules of co-regulated polyA sites. We find groups of intronic polyA sites regulated by distinct components of the nuclear RNA life cycle, including elongation, splicing, termination, and surveillance. We train and validate a deep neural network (APARENT-Perturb) for tandem polyA site usage, delineating a cis-regulatory code that predicts perturbation response and reveals interactions between regulatory complexes. Our work highlights the potential for multiplexed single-cell perturbation screens to further our understanding of post-transcriptional regulation.

On knowing a gene: A distributional hypothesis of gene function

As words can have multiple meanings that depend on sentence context, genes can have various functions that depend on the surrounding biological system. This pleiotropic nature of gene function is limited by ontologies, which annotate gene functions without considering biological contexts. We contend that the gene function problem in genetics may be informed by recent technological leaps in natural language processing, in which representations of word semantics can be automatically learned from diverse language contexts. In contrast to efforts to model semantics as "is-a" relationships in the 1990s, modern distributional semantics represents words as vectors in a learned semantic space and fuels current advances in transformer-based models such as large language models and generative pre-trained transformers. A similar shift in thinking of gene functions as distributions over cellular contexts may enable a similar breakthrough in data-driven learning from large biological datasets to inform gene function.

Basis of gene-specific transcription regulation by the Integrator complex

The Integrator complex attenuates gene expression via the premature termination of RNA polymerase II (RNAP2) at promoter-proximal pausing sites. It is required for stimulus response, cell differentiation, and neurodevelopment, but how gene-specific and adaptive regulation by Integrator is achieved remains unclear. Here, we identify two sites on human Integrator subunits 13/14 that serve as binding hubs for sequence-specific transcription factors (TFs) and other transcription effector complexes. When Integrator is attached to paused RNAP2, these hubs are positioned upstream of the transcription bubble, consistent with simultaneous TF-promoter tethering. The TFs co-localize with Integrator genome-wide, increase Integrator abundance on target genes, and co-regulate responsive transcriptional programs. For instance, sensory cilia formation induced by glucose starvation depends on Integrator-TF contacts. Our data suggest TF-mediated promoter recruitment of Integrator as a widespread mechanism for targeted transcription regulation.

SpRY-mediated screens facilitate functional dissection of non-coding sequences at single-base resolution

CRISPR mutagenesis screens conducted with SpCas9 and other nucleases have identified certain cis-regulatory elements and genetic variants but at a limited resolution due to the absence of protospacer adjacent motif (PAM) sequences. Here, leveraging the broad targeting scope of the near-PAMless SpRY variant, we have demonstrated that saturated SpRY mutagenesis and base editing screens can faithfully identify functional regulatory elements and essential genetic variants for target gene expression at single-base resolution. We further extended this methodology to investigate a genome-wide association study (GWAS) locus at 10q22.1 associated with a red blood cell trait, where we identified potential enhancers regulating HK1 gene expression, despite not all of these enhancers exhibiting typical chromatin signatures. More importantly, our saturated base editing screens pinpoint multiple causal variants within this locus that would otherwise be missed by Bayesian statistical fine-mapping. Our approach is generally applicable to functional interrogation of all non-coding genomic elements while complementing other high-coverage CRISPR screens.

Predicting transcriptional outcomes of novel multigene perturbations with GEARS

Understanding cellular responses to genetic perturbation is central to numerous biomedical applications, from identifying genetic interactions involved in cancer to developing methods for regenerative medicine. However, the combinatorial explosion in the number of possible multigene perturbations severely limits experimental interrogation. Here, we present graph-enhanced gene activation and repression simulator (GEARS), a method that integrates deep learning with a knowledge graph of gene-gene relationships to predict transcriptional responses to both single and multigene perturbations using single-cell RNA-sequencing data from perturbational screens. GEARS is able to predict outcomes of perturbing combinations consisting of genes that were never experimentally perturbed. GEARS exhibited 40% higher precision than existing approaches in predicting four distinct genetic interaction subtypes in a combinatorial perturbation screen and identified the strongest interactions twice as well as prior approaches. Overall, GEARS can predict phenotypically distinct effects of multigene perturbations and thus guide the design of perturbational experiments.

Comparative cofactor screens show the influence of transactivation domains and core promoters on the mechanisms of transcription

Eukaryotic transcription factors (TFs) activate gene expression by recruiting cofactors to promoters. However, the relationships between TFs, promoters and their associated cofactors remain poorly understood. Here we combine GAL4-transactivation assays with comparative CRISPR-Cas9 screens to identify the cofactors used by nine different TFs and core promoters in human cells. Using this dataset, we associate TFs with cofactors, classify cofactors as ubiquitous or specific and discover transcriptional co-dependencies. Through a reductionistic, comparative approach, we demonstrate that TFs do not display discrete mechanisms of activation. Instead, each TF depends on a unique combination of cofactors, which influences distinct steps in transcription. By contrast, the influence of core promoters appears relatively discrete. Different promoter classes are constrained by either initiation or pause-release, which influences their dynamic range and compatibility with cofactors. Overall, our comparative cofactor screens characterize the interplay between TFs, cofactors and core promoters, identifying general principles by which they influence transcription.

Murine cochlear damage models in the context of hair cell regeneration research

Understanding the complex pathologies associated with hearing loss is a significant motivation for conducting inner ear research. Lifelong exposure to loud noise, ototoxic drugs, genetic diversity, sex, and aging collectively contribute to human hearing loss. Replicating this pathology in research animals is challenging because hearing impairment has varied causes and different manifestations. A central aspect, however, is the loss of sensory hair cells and the inability of the mammalian cochlea to replace them. Researching therapeutic strategies to rekindle regenerative cochlear capacity, therefore, requires the generation of animal models in which cochlear hair cells are eliminated. This review discusses different approaches to ablate cochlear hair cells in adult mice. We inventoried the cochlear cyto- and histo-pathology caused by acoustic overstimulation, systemic and locally applied drugs, and various genetic tools. The focus is not to prescribe a perfect damage model but to highlight the limitations and advantages of existing approaches and identify areas for further refinement of damage models for use in regenerative studies.

CRISPR technologies for genome, epigenome and transcriptome editing

Our ability to edit genomes lags behind our capacity to sequence them, but the growing understanding of CRISPR biology and its application to genome, epigenome and transcriptome engineering is narrowing this gap. In this Review, we discuss recent developments of various CRISPR-based systems that can transiently or permanently modify the genome and the transcriptome. The discovery of further CRISPR enzymes and systems through functional metagenomics has meaningfully broadened the applicability of CRISPR-based editing. Engineered Cas variants offer diverse capabilities such as base editing, prime editing, gene insertion and gene regulation, thereby providing a panoply of tools for the scientific community. We highlight the strengths and weaknesses of current CRISPR tools, considering their efficiency, precision, specificity, reliance on cellular DNA repair mechanisms and their applications in both fundamental biology and therapeutics. Finally, we discuss ongoing clinical trials that illustrate the potential impact of CRISPR systems on human health.

Nucleotide depletion promotes cell fate transitions by inducing DNA replication stress

Control of cellular identity requires coordination of developmental programs with environmental factors such as nutrient availability, suggesting that perturbing metabolism can alter cell state. Here, we find that nucleotide depletion and DNA replication stress drive differentiation in human and murine normal and transformed hematopoietic systems, including patient-derived acute myeloid leukemia (AML) xenografts. These cell state transitions begin during S phase and are independent of ATR/ATM checkpoint signaling, double-stranded DNA break formation, and changes in cell cycle length. In systems where differentiation is blocked by oncogenic transcription factor expression, replication stress activates primed regulatory loci and induces lineage-appropriate maturation genes despite the persistence of progenitor programs. Altering the baseline cell state by manipulating transcription factor expression causes replication stress to induce genes specific for alternative lineages. The ability of replication stress to selectively activate primed maturation programs across different contexts suggests a general mechanism by which changes in metabolism can promote lineage-appropriate cell state transitions.

Building and analyzing metacells in single-cell genomics data

The advent of high-throughput single-cell genomics technologies has fundamentally transformed biological sciences. Currently, millions of cells from complex biological tissues can be phenotypically profiled across multiple modalities. The scaling of computational methods to analyze and visualize such data is a constant challenge, and tools need to be regularly updated, if not redesigned, to cope with ever-growing numbers of cells. Over the last few years, metacells have been introduced to reduce the size and complexity of single-cell genomics data while preserving biologically relevant information and improving interpretability. Here, we review recent studies that capitalize on the concept of metacells-and the many variants in nomenclature that have been used. We further outline how and when metacells should (or should not) be used to analyze single-cell genomics data and what should be considered when analyzing such data at the metacell level. To facilitate the exploration of metacells, we provide a comprehensive tutorial on the construction and analysis of metacells from single-cell RNA-seq data ( https://github.com/GfellerLab/MetacellAnalysisTutorial ) as well as a fully integrated pipeline to rapidly build, visualize and evaluate metacells with different methods ( https://github.com/GfellerLab/MetacellAnalysisToolkit ).

High-resolution genome-wide mapping of chromosome-arm-scale truncations induced by CRISPR-Cas9 editing

Clustered regularly interspaced short palindromic repeats (CRISPR)-CRISPR-associated protein 9 (Cas9) is a powerful tool for introducing targeted mutations in DNA, but recent studies have shown that it can have unintended effects such as structural changes. However, these studies have not yet looked genome wide or across data types. Here we performed a phenotypic CRISPR-Cas9 scan targeting 17,065 genes in primary human cells, revealing a 'proximity bias' in which CRISPR knockouts show unexpected similarities to unrelated genes on the same chromosome arm. This bias was found to be consistent across cell types, laboratories, Cas9 delivery methods and assay modalities, and the data suggest that it is caused by telomeric truncations of chromosome arms, with cell cycle and apoptotic pathways playing a mediating role. Additionally, a simple correction is demonstrated to mitigate this pervasive bias while preserving biological relationships. This previously uncharacterized effect has implications for functional genomic studies using CRISPR-Cas9, with applications in discovery biology, drug-target identification, cell therapies and genetic therapeutics.

Structural basis of the Integrator complex assembly and association with transcription factors

Integrator is a multi-subunit protein complex responsible for premature transcription termination of coding and non-coding RNAs. This is achieved via two enzymatic activities, RNA endonuclease and protein phosphatase, acting on the promoter-proximally paused RNA polymerase Ⅱ (RNAPⅡ). Yet, it remains unclear how Integrator assembly and recruitment are regulated and what the functions of many of its core subunits are. Here, we report the structures of two human Integrator sub-complexes: INTS10/13/14/15 and INTS5/8/10/15, and an integrative model of the fully assembled Integrator bound to the RNAPⅡ paused elongating complex (PEC). An in silico protein-protein interaction screen of over 1,500 human transcription factors (TFs) identified ZNF655 as a direct interacting partner of INTS13 within the fully assembled Integrator. We propose a model wherein INTS13 acts as a platform for the recruitment of TFs that could modulate the stability of the Integrator's association at specific loci and regulate transcription attenuation of the target genes.

Morphological profiling for drug discovery in the era of deep learning

Morphological profiling is a valuable tool in phenotypic drug discovery. The advent of high-throughput automated imaging has enabled the capturing of a wide range of morphological features of cells or organisms in response to perturbations at the single-cell resolution. Concurrently, significant advances in machine learning and deep learning, especially in computer vision, have led to substantial improvements in analyzing large-scale high-content images at high throughput. These efforts have facilitated understanding of compound mechanism of action, drug repurposing, characterization of cell morphodynamics under perturbation, and ultimately contributing to the development of novel therapeutics. In this review, we provide a comprehensive overview of the recent advances in the field of morphological profiling. We summarize the image profiling analysis workflow, survey a broad spectrum of analysis strategies encompassing feature engineering- and deep learning-based approaches, and introduce publicly available benchmark datasets. We place a particular emphasis on the application of deep learning in this pipeline, covering cell segmentation, image representation learning, and multimodal learning. Additionally, we illuminate the application of morphological profiling in phenotypic drug discovery and highlight potential challenges and opportunities in this field.

Establishing a conceptual framework for holistic cell states and state transitions

Cell states were traditionally defined by how they looked, where they were located, and what functions they performed. In this post-genomic era, the field is largely focused on a molecular view of cell state. Moving forward, we anticipate that the observables used to define cell states will evolve again as single-cell imaging and analytics are advancing at a breakneck pace via the collection of large-scale, systematic cell image datasets and the application of quantitative image-based data science methods. This is, therefore, a key moment in the arc of cell biological research to develop approaches that integrate the spatiotemporal observables of the physical structure and organization of the cell with molecular observables toward the concept of a holistic cell state. In this perspective, we propose a conceptual framework for holistic cell states and state transitions that is data-driven, practical, and useful to enable integrative analyses and modeling across many data types.

Central dogma rates in human mitochondria

In human cells, the nuclear and mitochondrial genomes engage in a complex interplay to produce dual-encoded oxidative phosphorylation (OXPHOS) complexes. The coordination of these dynamic gene expression processes is essential for producing matched amounts of OXPHOS protein subunits. This review focuses on our current understanding of the mitochondrial central dogma rates, highlighting the striking differences in gene expression rates between mitochondrial and nuclear genes. We synthesize a coherent model of mitochondrial gene expression kinetics, highlighting the emerging principles and emphasizing where more precise measurements would be beneficial. Such an understanding is pivotal for grasping the unique aspects of mitochondrial function and its role in cellular energetics, and it has profound implications for aging, metabolic disorders, and neurodegenerative diseases.

Analyzing the functional effects of DNA variants with gene editing

Continual advancements in genomics have led to an ever-widening disparity between the rate of discovery of genetic variants and our current understanding of their functions and potential roles in disease. Systematic methods for phenotyping DNA variants are required to effectively translate genomics data into improved outcomes for patients with genetic diseases. To make the biggest impact, these approaches must be scalable and accurate, faithfully reflect disease biology, and define complex disease mechanisms. We compare current methods to analyze the function of variants in their endogenous DNA context using genome editing strategies, such as saturation genome editing, base editing and prime editing. We discuss how these technologies can be linked to high-content readouts to gain deep mechanistic insights into variant effects. Finally, we highlight key challenges that need to be addressed to bridge the genotype to phenotype gap, and ultimately improve the diagnosis and treatment of genetic diseases.

Single-cell analyses reveal evolution mimicry during the specification of breast cancer subtype

Background: The stem or progenitor antecedents confer developmental plasticity and unique cell identities to cancer cells via genetic and epigenetic programs. A comprehensive characterization and mapping of the cell-of-origin of breast cancer using novel technologies to unveil novel subtype-specific therapeutic targets is still absent. Methods: We integrated 195,144 high-quality cells from normal breast tissues and 406,501 high-quality cells from primary breast cancer samples to create a large-scale single-cell atlas of human normal and cancerous breasts. Potential heterogeneous origin of malignant cells was explored by contrasting cancer cells against reference normal epithelial cells. Multi-omics analyses and both in vitro and in vivo experiments were performed to screen and validate potential subtype-specific treatment targets. Novel biomarkers of identified immune and stromal cell subpopulations were validated by immunohistochemistry in our cohort. Results: Tumor stratification based on cancer cell-of-origin patterns correlated with clinical outcomes, genomic aberrations and diverse microenvironment constitutions. We found that the luminal progenitor (LP) subtype was robustly associated with poor prognosis, genomic instability and dysfunctional immune microenvironment. However, the LP subtype patients were sensitive to neoadjuvant chemotherapy (NAC), PARP inhibitors (PARPi) and immunotherapy. The LP subtype-specific target PLK1 was investigated by both in vitro and in vivo experiments. Besides, large-scale single-cell profiling of breast cancer inspired us to identify a range of clinically relevant immune and stromal cell subpopulations, including subsets of innate lymphoid cells (ILCs), macrophages and endothelial cells. Conclusion: The present single-cell study revealed the cellular repertoire and cell-of-origin patterns of breast cancer. Combining single-cell and bulk transcriptome data, we elucidated the evolution mimicry from normal to malignant subtypes and expounded the LP subtype with vital clinical implications. Novel immune and stromal cell subpopulations of breast cancer identified in our study could be potential therapeutic targets. Taken together, Our findings lay the foundation for the precise prognostic and therapeutic stratification of breast cancer.

Bidirectional epigenetic editing reveals hierarchies in gene regulation

CRISPR perturbation methods are limited in their ability to study non-coding elements and genetic interactions. In this study, we developed a system for bidirectional epigenetic editing, called CRISPRai, in which we apply activating (CRISPRa) and repressive (CRISPRi) perturbations to two loci simultaneously in the same cell. We developed CRISPRai Perturb-seq by coupling dual perturbation gRNA detection with single-cell RNA sequencing, enabling study of pooled perturbations in a mixed single-cell population. We applied this platform to study the genetic interaction between two hematopoietic lineage transcription factors, SPI1 and GATA1, and discovered novel characteristics of their co-regulation on downstream target genes, including differences in SPI1 and GATA1 occupancy at genes that are regulated through different modes. We also studied the regulatory landscape of IL2 (interleukin-2) in Jurkat T cells, primary T cells and chimeric antigen receptor (CAR) T cells and elucidated mechanisms of enhancer-mediated IL2 gene regulation. CRISPRai facilitates investigation of context-specific genetic interactions, provides new insights into gene regulation and will enable exploration of non-coding disease-associated variants.

Robust differential expression testing for single-cell CRISPR screens at low multiplicity of infection

Single-cell CRISPR screens (perturb-seq) link genetic perturbations to phenotypic changes in individual cells. The most fundamental task in perturb-seq analysis is to test for association between a perturbation and a count outcome, such as gene expression. We conduct the first-ever comprehensive benchmarking study of association testing methods for low multiplicity-of-infection (MOI) perturb-seq data, finding that existing methods produce excess false positives. We conduct an extensive empirical investigation of the data, identifying three core analysis challenges: sparsity, confounding, and model misspecification. Finally, we develop an association testing method - SCEPTRE low-MOI - that resolves these analysis challenges and demonstrates improved calibration and power.

Multimodal Learning for Mapping the Genotype-Phenotype Dynamics

How complex phenotypes emerge from intricate gene expression patterns is a fundamental question in biology. Quantitative characterization of this relationship, however, is challenging due to the vast combinatorial possibilities and dynamic interplay between genotype and phenotype landscapes. Integrating high-content genotyping approaches such as single-cell RNA sequencing and advanced learning methods such as language models offers an opportunity for dissecting this complex relationship. Here, we present a computational integrated genetics framework designed to analyze and interpret the high-dimensional landscape of genotypes and their associated phenotypes simultaneously. We applied this approach to develop a multimodal foundation model to explore the genotype-phenotype relationship manifold for human transcriptomics at the cellular level. Analyzing this joint manifold showed a refined resolution of cellular heterogeneity, enhanced precision in phenotype annotating, and uncovered potential cross-tissue biomarkers that are undetectable through conventional gene expression analysis alone. Moreover, our results revealed that the gene networks are characterized by scale-free patterns and show context-dependent gene-gene interactions, both of which result in significant variations in the topology of the gene network, particularly evident during aging. Finally, utilizing contextualized embeddings, we investigated gene polyfunctionality which illustrates the multifaceted roles that genes play in different biological processes, and demonstrated that for VWF gene in endothelial cells. Overall, this study advances our understanding of the dynamic interplay between gene expression and phenotypic manifestation and demonstrates the potential of integrated genetics in uncovering new dimensions of cellular function and complexity.

Quantifying 3'UTR length from scRNA-seq data reveals changes independent of gene expression

Although more than half of all genes generate transcripts that differ in 3'UTR length, current analysis pipelines only quantify the amount but not the length of mRNA transcripts. 3'UTR length is determined by 3' end cleavage sites (CS). We map CS in more than 200 primary human and mouse cell types and increase CS annotations relative to the GENCODE database by 40%. Approximately half of all CS are used in few cell types, revealing that most genes only have one or two major 3' ends. We incorporate the CS annotations into a computational pipeline, called scUTRquant, for rapid, accurate, and simultaneous quantification of gene and 3'UTR isoform expression from single-cell RNA sequencing (scRNA-seq) data. When applying scUTRquant to data from 474 cell types and 2134 perturbations, we discover extensive 3'UTR length changes across cell types that are as widespread and coordinately regulated as gene expression changes but affect mostly different genes. Our data indicate that mRNA abundance and mRNA length are two largely independent axes of gene regulation that together determine the amount and spatial organization of protein synthesis.

The future of rapid and automated single-cell data analysis using reference mapping

As the number of single-cell datasets continues to grow rapidly, workflows that map new data to well-curated reference atlases offer enormous promise for the biological community. In this perspective, we discuss key computational challenges and opportunities for single-cell reference-mapping algorithms. We discuss how mapping algorithms will enable the integration of diverse datasets across disease states, molecular modalities, genetic perturbations, and diverse species and will eventually replace manual and laborious unsupervised clustering pipelines.

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.

Mapping stress-responsive signaling pathways induced by mitochondrial proteostasis perturbations

Imbalances in mitochondrial proteostasis are associated with pathologic mitochondrial dysfunction implicated in etiologically diverse diseases. This has led to considerable interest in defining the mechanisms responsible for regulating mitochondria in response to mitochondrial stress. Numerous stress-responsive signaling pathways have been suggested to regulate mitochondria in response to proteotoxic stress. These include the integrated stress response (ISR), the heat shock response (HSR), and the oxidative stress response (OSR). Here, we define the stress signaling pathways activated in response to chronic mitochondrial proteostasis perturbations by monitoring the expression of sets of genes regulated downstream of each of these signaling pathways in published Perturb-seq datasets from K562 cells CRISPRi-depleted of mitochondrial proteostasis factors. Interestingly, we find that the ISR is preferentially activated in response to chronic, genetically-induced mitochondrial proteostasis stress, with no other pathway showing significant activation. Further, we demonstrate that CRISPRi depletion of other mitochondria-localized proteins similarly shows preferential activation of the ISR relative to other stress-responsive signaling pathways. These results both establish our gene set profiling approach as a viable strategy to probe stress responsive signaling pathways induced by perturbations to specific organelles and identify the ISR as the predominant stress-responsive signaling pathway activated in response to chronic disruption of mitochondrial proteostasis.

Structural basis of Integrator-dependent RNA polymerase II termination

The Integrator complex can terminate RNA polymerase II (Pol II) in the promoter-proximal region of genes. Previous work has shed light on how Integrator binds to the paused elongation complex consisting of Pol II, the DRB sensitivity-inducing factor (DSIF) and the negative elongation factor (NELF) and how it cleaves the nascent RNA transcript1, but has not explained how Integrator removes Pol II from the DNA template. Here we present three cryo-electron microscopy structures of the complete Integrator-PP2A complex in different functional states. The structure of the pre-termination complex reveals a previously unresolved, scorpion-tail-shaped INTS10-INTS13-INTS14-INTS15 module that may use its 'sting' to open the DSIF DNA clamp and facilitate termination. The structure of the post-termination complex shows that the previously unresolved subunit INTS3 and associated sensor of single-stranded DNA complex (SOSS) factors prevent Pol II rebinding to Integrator after termination. The structure of the free Integrator-PP2A complex in an inactive closed conformation2 reveals that INTS6 blocks the PP2A phosphatase active site. These results lead to a model for how Integrator terminates Pol II transcription in three steps that involve major rearrangements.

A logic-incorporated gene regulatory network deciphers principles in cell fate decisions

Organisms utilize gene regulatory networks (GRN) to make fate decisions, but the regulatory mechanisms of transcription factors (TF) in GRNs are exceedingly intricate. A longstanding question in this field is how these tangled interactions synergistically contribute to decision-making procedures. To comprehensively understand the role of regulatory logic in cell fate decisions, we constructed a logic-incorporated GRN model and examined its behavior under two distinct driving forces (noise-driven and signal-driven). Under the noise-driven mode, we distilled the relationship among fate bias, regulatory logic, and noise profile. Under the signal-driven mode, we bridged regulatory logic and progression-accuracy trade-off, and uncovered distinctive trajectories of reprogramming influenced by logic motifs. In differentiation, we characterized a special logic-dependent priming stage by the solution landscape. Finally, we applied our findings to decipher three biological instances: hematopoiesis, embryogenesis, and trans-differentiation. Orthogonal to the classical analysis of expression profile, we harnessed noise patterns to construct the GRN corresponding to fate transition. Our work presents a generalizable framework for top-down fate-decision studies and a practical approach to the taxonomy of cell fate decisions.

Discovering mechanisms of human genetic variation and controlling cell states at scale

Population-scale sequencing efforts have catalogued substantial genetic variation in humans such that variant discovery dramatically outpaces interpretation. We discuss how single-cell sequencing is poised to reveal genetic mechanisms at a rate that may soon approach that of variant discovery. The functional genomics toolkit is sufficiently modular to systematically profile almost any type of variation within increasingly diverse contexts and with molecularly comprehensive and unbiased readouts. As a result, we can construct deep phenotypic atlases of variant effects that span the entire regulatory cascade. The same conceptual approach to interpreting genetic variation should be applied to engineering therapeutic cell states. In this way, variant mechanism discovery and cell state engineering will become reciprocating and iterative processes towards genomic medicine.

Unveiling novel insights in acute myeloid leukemia through single-cell RNA sequencing

Acute myeloid leukemia (AML) is a complex and heterogeneous group of aggressive hematopoietic stem cell disease. The presence of diverse and functionally distinct populations of leukemia cells within the same patient's bone marrow or blood poses a significant challenge in diagnosing and treating AML. A substantial proportion of AML patients demonstrate resistance to induction chemotherapy and a grim prognosis upon relapse. The rapid advance in next generation sequencing technologies, such as single-cell RNA-sequencing (scRNA-seq), has revolutionized our understanding of AML pathogenesis by enabling high-resolution interrogation of the cellular heterogeneity in the AML ecosystem, and their transcriptional signatures at a single-cell level. New studies have successfully characterized the inextricably intertwined interactions among AML cells, immune cells and bone marrow microenvironment and their contributions to the AML development, therapeutic resistance and relapse. These findings have deepened and broadened our understanding the complexity and heterogeneity of AML, which are difficult to detect with bulk RNA-seq. This review encapsulates the burgeoning body of knowledge generated through scRNA-seq, providing the novel insights and discoveries it has unveiled in AML biology. Furthermore, we discuss the potential implications of scRNA-seq in therapeutic opportunities, focusing on immunotherapy. Finally, we highlight the current limitations and future direction of scRNA-seq in the field.

Robust differential expression testing for single-cell CRISPR screens at low multiplicity of infection

Single-cell CRISPR screens (perturb-seq) link genetic perturbations to phenotypic changes in individual cells. The most fundamental task in perturb-seq analysis is to test for association between a perturbation and a count outcome, such as gene expression. We conduct the first-ever comprehensive benchmarking study of association testing methods for low multiplicity-of-infection (MOI) perturb-seq data, finding that existing methods produce excess false positives. We conduct an extensive empirical investigation of the data, identifying three core analysis challenges: sparsity, confounding, and model misspecification. Finally, we develop an association testing method - SCEPTRE low-MOI - that resolves these analysis challenges and demonstrates improved calibration and power.

Association between de novo variants of nuclear-encoded mitochondrial-related genes and undiagnosed developmental disorder and autism

BACKGROUND

Evidence suggests that mitochondrial abnormalities increase the risk of two neurodevelopmental disorders: undiagnosed developmental disorder (UDD) and autism spectrum disorder (ASD). However, which nuclear-encoded mitochondrial-related genes (NEMGs) were associated with UDD-ASD is unclear.

AIM

To explore the association between de novo variants (DNVs) of NEMGs and UDD-ASD.

DESIGN

Comprehensive analysis based on DNVs of NEMGs identified in patients (31 058 UDD probands and 10 318 ASD probands) and 4262 controls.

METHODS

By curating NEMGs and cataloging publicly published DNVs in NEMGs, we compared the frequency of DNVs in cases and controls. We also applied a TADA-denovo model to highlight disease-associated NEMGs and characterized them based on gene intolerance, functional networks and expression patterns.

RESULTS

Compared with levels in 4262 controls, an excess of protein-truncating variants and deleterious missense variants in 1421 cataloged NEMGs from 41 376 patients (31 058 UDD and 10 318 ASD probands) was observed. Overall, 3.23% of de novo deleterious missense variants and 3.20% of de novo protein-truncating variants contributed to 1.1% and 0.39% of UDD-ASD cases, respectively. We prioritized 130 disease-associated NEMGs and showed distinct expression patterns in the developing human brain. Disease-associated NEMGs expression was enriched in both excitatory and inhibitory neuronal lineages from the developing human cortex.

CONCLUSIONS

Rare genetic alterations of disease-associated NEMGs may play a role in UDD-ASD development and lay the groundwork for a better understanding of the biology of UDD-ASD.

Predictomes: A classifier-curated database of AlphaFold-modeled protein-protein interactions

Protein-protein interactions (PPIs) are ubiquitous in biology, yet a comprehensive structural characterization of the PPIs underlying biochemical processes is lacking. Although AlphaFold-Multimer (AF-M) has the potential to fill this knowledge gap, standard AF-M confidence metrics do not reliably separate relevant PPIs from an abundance of false positive predictions. To address this limitation, we used machine learning on well curated datasets to train a Structure Prediction and Omics informed Classifier called SPOC that shows excellent performance in separating true and false PPIs, including in proteome-wide screens. We applied SPOC to an all-by-all matrix of nearly 300 human genome maintenance proteins, generating ~40,000 predictions that can be viewed at predictomes.org, where users can also score their own predictions with SPOC. High confidence PPIs discovered using our approach suggest novel hypotheses in genome maintenance. Our results provide a framework for interpreting large scale AF-M screens and help lay the foundation for a proteome-wide structural interactome.

Heterozygous loss-of-function SMC3 variants are associated with variable growth and developmental features

Heterozygous missense variants and in-frame indels in SMC3 are a cause of Cornelia de Lange syndrome (CdLS), marked by intellectual disability, growth deficiency, and dysmorphism, via an apparent dominant-negative mechanism. However, the spectrum of manifestations associated with SMC3 loss-of-function variants has not been reported, leading to hypotheses of alternative phenotypes or even developmental lethality. We used matchmaking servers, patient registries, and other resources to identify individuals with heterozygous, predicted loss-of-function (pLoF) variants in SMC3, and analyzed population databases to characterize mutational intolerance in this gene. Here, we show that SMC3 behaves as an archetypal haploinsufficient gene: it is highly constrained against pLoF variants, strongly depleted for missense variants, and pLoF variants are associated with a range of developmental phenotypes. Among 14 individuals with SMC3 pLoF variants, phenotypes were variable but coalesced on low growth parameters, developmental delay/intellectual disability, and dysmorphism, reminiscent of atypical CdLS. Comparisons to individuals with SMC3 missense/in-frame indel variants demonstrated an overall milder presentation in pLoF carriers. Furthermore, several individuals harboring pLoF variants in SMC3 were nonpenetrant for growth, developmental, and/or dysmorphic features, and some had alternative symptomatologies with rational biological links to SMC3. Analyses of tumor and model system transcriptomic data and epigenetic data in a subset of cases suggest that SMC3 pLoF variants reduce SMC3 expression but do not strongly support clustering with functional genomic signatures of typical CdLS. Our finding of substantial population-scale LoF intolerance in concert with variable growth and developmental features in subjects with SMC3 pLoF variants expands the scope of cohesinopathies, informs on their allelic architecture, and suggests the existence of additional clearly LoF-constrained genes whose disease links will be confirmed only by multilayered genomic data paired with careful phenotyping.

Classification and functional characterization of regulators of intracellular STING trafficking identified by genome-wide optical pooled screening

STING is an innate immune sensor that traffics across many cellular compartments to carry out its function of detecting cyclic di-nucleotides and triggering defense processes. Mutations in factors that regulate this process are often linked to STING-dependent human inflammatory disorders. To systematically identify factors involved in STING trafficking, we performed a genome-wide optical pooled screen and examined the impact of genetic perturbations on intracellular STING localization. Based on subcellular imaging of STING protein and trafficking markers in 45 million cells perturbed with sgRNAs, we defined 464 clusters of gene perturbations with similar cellular phenotypes. A higher-dimensional focused optical pooled screen on 262 perturbed genes which assayed 11 imaging channels identified 73 finer phenotypic clusters. In a cluster containing USE1, a protein that mediates Golgi to ER transport, we found a gene of unknown function, C19orf25. Consistent with the known role of USE1, loss of C19orf25 enhanced STING signaling. Other clusters contained subunits of the HOPS, GARP and RIC1-RGP1 complexes. We show that HOPS deficiency delayed STING degradation and consequently increased signaling. Similarly, GARP/RIC1-RGP1 loss increased STING signaling by delaying STING exit from the Golgi. Our findings demonstrate that genome-wide genotype-phenotype maps based on high-content cell imaging outperform other screening approaches, and provide a community resource for mining for factors that impact STING trafficking as well as other cellular processes observable in our dataset.

Single-cell image-based genetic screens systematically identify regulators of Ebola virus subcellular infection dynamics

Ebola virus (EBOV) is a high-consequence filovirus that gives rise to frequent epidemics with high case fatality rates and few therapeutic options. Here, we applied image-based screening of a genome-wide CRISPR library to systematically identify host cell regulators of Ebola virus infection in 39,085,093 million single cells. Measuring viral RNA and protein levels together with their localization in cells identified over 998 related host factors and provided detailed information about the role of each gene across the virus replication cycle. We trained a deep learning model on single-cell images to associate each host factor with predicted replication steps, and confirmed the predicted relationship for select host factors. Among the findings, we showed that the mitochondrial complex III subunit UQCRB is a post-entry regulator of Ebola virus RNA replication, and demonstrated that UQCRB inhibition with a small molecule reduced overall Ebola virus infection with an IC50 of 5 μM. Using a random forest model, we also identified perturbations that reduced infection by disrupting the equilibrium between viral RNA and protein. One such protein, STRAP, is a spliceosome-associated factor that was found to be closely associated with VP35, a viral protein required for RNA processing. Loss of STRAP expression resulted in a reduction in full-length viral genome production and subsequent production of non-infectious virus particles. Overall, the data produced in this genome-wide high-content single-cell screen and secondary screens in additional cell lines and related filoviruses (MARV and SUDV) revealed new insights about the role of host factors in virus replication and potential new targets for therapeutic intervention.

An in vitro neurogenetics platform for precision disease modeling in the mouse

The power and scope of disease modeling can be markedly enhanced through the incorporation of broad genetic diversity. The introduction of pathogenic mutations into a single inbred mouse strain sometimes fails to mimic human disease. We describe a cross-species precision disease modeling platform that exploits mouse genetic diversity to bridge cell-based modeling with whole organism analysis. We developed a universal protocol that permitted robust and reproducible neural differentiation of genetically diverse human and mouse pluripotent stem cell lines and then carried out a proof-of-concept study of the neurodevelopmental gene DYRK1A. Results in vitro reliably predicted the effects of genetic background on Dyrk1a loss-of-function phenotypes in vivo. Transcriptomic comparison of responsive and unresponsive strains identified molecular pathways conferring sensitivity or resilience to Dyrk1a1A loss and highlighted differential messenger RNA isoform usage as an important determinant of response. This cross-species strategy provides a powerful tool in the functional analysis of candidate disease variants identified through human genetic studies.

Pleiotropy, epistasis and the genetic architecture of quantitative traits

Pleiotropy (whereby one genetic polymorphism affects multiple traits) and epistasis (whereby non-linear interactions between genetic polymorphisms affect the same trait) are fundamental aspects of the genetic architecture of quantitative traits. Recent advances in the ability to characterize the effects of polymorphic variants on molecular and organismal phenotypes in human and model organism populations have revealed the prevalence of pleiotropy and unexpected shared molecular genetic bases among quantitative traits, including diseases. By contrast, epistasis is common between polymorphic loci associated with quantitative traits in model organisms, such that alleles at one locus have different effects in different genetic backgrounds, but is rarely observed for human quantitative traits and common diseases. Here, we review the concepts and recent inferences about pleiotropy and epistasis, and discuss factors that contribute to similarities and differences between the genetic architecture of quantitative traits in model organisms and humans.

The Role of Endothelial Cells in Atherosclerosis: Insights from Genetic Association Studies

Endothelial cells (ECs) mediate several biological functions that are relevant to atherosclerosis and coronary artery disease (CAD), regulating an array of vital processes including vascular tone, wound healing, reactive oxygen species, shear stress response, and inflammation. Although which of these functions is linked causally with CAD development and/or progression is not yet known, genome-wide association studies have implicated more than 400 loci associated with CAD risk, among which several have shown EC-relevant functions. Given the arduous process of mechanistically interrogating single loci to CAD, high-throughput variant characterization methods, including pooled Clustered Regularly Interspaced Short Palindromic Repeats screens, offer exciting potential to rapidly accelerate the discovery of bona fide EC-relevant genetic loci. These discoveries in turn will broaden the therapeutic avenues for CAD beyond lipid lowering and behavioral risk modification to include EC-centric modalities of risk prevention and treatment.

Acute multi-level response to defective de novo chromatin assembly in S-phase

Long-term perturbation of de novo chromatin assembly during DNA replication has profound effects on epigenome maintenance and cell fate. The early mechanistic origin of these defects is unknown. Here, we combine acute degradation of Chromatin Assembly Factor 1 (CAF-1), a key player in de novo chromatin assembly, with single-cell genomics, quantitative proteomics, and live-microscopy to uncover these initiating mechanisms in human cells. CAF-1 loss immediately slows down DNA replication speed and renders nascent DNA hyperaccessible. A rapid cellular response, distinct from canonical DNA damage signaling, is triggered and lowers histone mRNAs. As a result, histone variants usage and their modifications are altered, limiting transcriptional fidelity and delaying chromatin maturation within a single S-phase. This multi-level response induces a cell-cycle arrest after mitosis. Our work reveals the immediate consequences of defective de novo chromatin assembly during DNA replication, explaining how at later times the epigenome and cell fate can be altered.

Triaging of α-helical proteins to the mitochondrial outer membrane by distinct chaperone machinery based on substrate topology

Mitochondrial outer membrane ⍺-helical proteins play critical roles in mitochondrial-cytoplasmic communication, but the rules governing the targeting and insertion of these biophysically diverse proteins remain unknown. Here, we first defined the complement of required mammalian biogenesis machinery through genome-wide CRISPRi screens using topologically distinct membrane proteins. Systematic analysis of nine identified factors across 21 diverse ⍺-helical substrates reveals that these components are organized into distinct targeting pathways that act on substrates based on their topology. NAC is required for the efficient targeting of polytopic proteins, whereas signal-anchored proteins require TTC1, a cytosolic chaperone that physically engages substrates. Biochemical and mutational studies reveal that TTC1 employs a conserved TPR domain and a hydrophobic groove in its C-terminal domain to support substrate solubilization and insertion into mitochondria. Thus, the targeting of diverse mitochondrial membrane proteins is achieved through topological triaging in the cytosol using principles with similarities to ER membrane protein biogenesis systems.

Targeting PIKfyve-driven lipid homeostasis as a metabolic vulnerability in pancreatic cancer

Pancreatic ductal adenocarcinoma (PDAC) subsists in a nutrient-deregulated microenvironment, making it particularly susceptible to treatments that interfere with cancer metabolism12. For example, PDAC utilizes and is dependent on high levels of autophagy and other lysosomal processes3-5. Although targeting these pathways has shown potential in preclinical studies, progress has been hampered by the challenge of identifying and characterizing favorable targets for drug development6. Here, we characterize PIKfyve, a lipid kinase integral to lysosomal functioning7, as a novel and targetable vulnerability in PDAC. In human patient and murine PDAC samples, we discovered that PIKFYVE is overexpressed in PDAC cells compared to adjacent normal cells. Employing a genetically engineered mouse model, we established the essential role of PIKfyve in PDAC progression. Further, through comprehensive metabolic analyses, we found that PIKfyve inhibition obligated PDAC to upregulate de novo lipid synthesis, a relationship previously undescribed. PIKfyve inhibition triggered a distinct lipogenic gene expression and metabolic program, creating a dependency on de novo lipid metabolism pathways, by upregulating genes such as FASN and ACACA. In PDAC, the KRAS-MAPK signaling pathway is a primary driver of de novo lipid synthesis, specifically enhancing FASN and ACACA levels. Accordingly, the simultaneous targeting of PIKfyve and KRAS-MAPK resulted in the elimination of tumor burden in a syngeneic orthotopic model and tumor regression in a xenograft model of PDAC. Taken together, these studies suggest that disrupting lipid metabolism through PIKfyve inhibition induces synthetic lethality in conjunction with KRAS-MAPK-directed therapies for PDAC.

CROPseq-multi: a versatile solution for multiplexed perturbation and decoding in pooled CRISPR screens

Forward genetic screens seek to dissect complex biological systems by systematically perturbing genetic elements and observing the resulting phenotypes. While standard screening methodologies introduce individual perturbations, multiplexing perturbations improves the performance of single-target screens and enables combinatorial screens for the study of genetic interactions. Current tools for multiplexing perturbations are incompatible with pooled screening methodologies that require mRNA-embedded barcodes, including some microscopy and single cell sequencing approaches. Here, we report the development of CROPseq-multi, a CROPseq1-inspired lentiviral system to multiplex Streptococcus pyogenes (Sp) Cas9-based perturbations with mRNA-embedded barcodes. CROPseq-multi has equivalent per-guide activity to CROPseq and low lentiviral recombination frequencies. CROPseq-multi is compatible with enrichment screening methodologies and optical pooled screens, and is extensible to screens with single-cell sequencing readouts. For optical pooled screens, an optimized and multiplexed in situ detection protocol improves barcode detection efficiency 10-fold, enables detection of recombination events, and increases decoding efficiency 3-fold relative to CROPseq. CROPseq-multi is a widely applicable multiplexing solution for diverse SpCas9-based genetic screening approaches.

Discovering Root Causal Genes with High Throughput Perturbations

Root causal gene expression levels - or root causal genes for short - correspond to the initial changes to gene expression that generate patient symptoms as a downstream effect. Identifying root causal genes is critical towards developing treatments that modify disease near its onset, but no existing algorithms attempt to identify root causal genes from data. RNA-sequencing (RNA-seq) data introduces challenges such as measurement error, high dimensionality and non-linearity that compromise accurate estimation of root causal effects even with state-of-the-art approaches. We therefore instead leverage Perturb-seq, or high throughput perturbations with single cell RNA-seq readout, to learn the causal order between the genes. We then transfer the causal order to bulk RNA-seq and identify root causal genes specific to a given patient for the first time using a novel statistic. Experiments demonstrate large improvements in performance. Applications to macular degeneration and multiple sclerosis also reveal root causal genes that lie on known pathogenic pathways, delineate patient subgroups and implicate a newly defined omnigenic root causal model.

Non-linear transcriptional responses to gradual modulation of transcription factor dosage

Genomic loci associated with common traits and diseases are typically non-coding and likely impact gene expression, sometimes coinciding with rare loss-of-function variants in the target gene. However, our understanding of how gradual changes in gene dosage affect molecular, cellular, and organismal traits is currently limited. To address this gap, we induced gradual changes in gene expression of four genes using CRISPR activation and inactivation. Downstream transcriptional consequences of dosage modulation of three master trans-regulators associated with blood cell traits (GFI1B, NFE2, and MYB) were examined using targeted single-cell multimodal sequencing. We showed that guide tiling around the TSS is the most effective way to modulate cis gene expression across a wide range of fold-changes, with further effects from chromatin accessibility and histone marks that differ between the inhibition and activation systems. Our single-cell data allowed us to precisely detect subtle to large gene expression changes in dozens of trans genes, revealing that many responses to dosage changes of these three TFs are non-linear, including non-monotonic behaviours, even when constraining the fold-changes of the master regulators to a copy number gain or loss. We found that the dosage properties are linked to gene constraint and that some of these non-linear responses are enriched for disease and GWAS genes. Overall, our study provides a straightforward and scalable method to precisely modulate gene expression and gain insights into its downstream consequences at high resolution.

scPerturb: harmonized single-cell perturbation data

Analysis across a growing number of single-cell perturbation datasets is hampered by poor data interoperability. To facilitate development and benchmarking of computational methods, we collect a set of 44 publicly available single-cell perturbation-response datasets with molecular readouts, including transcriptomics, proteomics and epigenomics. We apply uniform quality control pipelines and harmonize feature annotations. The resulting information resource, scPerturb, enables development and testing of computational methods, and facilitates comparison and integration across datasets. We describe energy statistics (E-statistics) for quantification of perturbation effects and significance testing, and demonstrate E-distance as a general distance measure between sets of single-cell expression profiles. We illustrate the application of E-statistics for quantifying similarity and efficacy of perturbations. The perturbation-response datasets and E-statistics computation software are publicly available at scperturb.org. This work provides an information resource for researchers working with single-cell perturbation data and recommendations for experimental design, including optimal cell counts and read depth.

scGPT: toward building a foundation model for single-cell multi-omics using generative AI

Generative pretrained models have achieved remarkable success in various domains such as language and computer vision. Specifically, the combination of large-scale diverse datasets and pretrained transformers has emerged as a promising approach for developing foundation models. Drawing parallels between language and cellular biology (in which texts comprise words; similarly, cells are defined by genes), our study probes the applicability of foundation models to advance cellular biology and genetic research. Using burgeoning single-cell sequencing data, we have constructed a foundation model for single-cell biology, scGPT, based on a generative pretrained transformer across a repository of over 33 million cells. Our findings illustrate that scGPT effectively distills critical biological insights concerning genes and cells. Through further adaptation of transfer learning, scGPT can be optimized to achieve superior performance across diverse downstream applications. This includes tasks such as cell type annotation, multi-batch integration, multi-omic integration, perturbation response prediction and gene network inference.