VEGA is an interpretable generative model for inferring biological network activity in single-cell transcriptomics

IRnet: Immunotherapy response prediction using pathway knowledge-informed graph neural network

INTRODUCTION

Immune checkpoint inhibitors (ICIs) are potent and precise therapies for various cancer types, significantly improving survival rates in patients who respond positively to them. However, only a minority of patients benefit from ICI treatments.

OBJECTIVES

Identifying ICI responders before treatment could greatly conserve medical resources, minimize potential drug side effects, and expedite the search for alternative therapies. Our goal is to introduce a novel deep-learning method to predict ICI treatment responses in cancer patients.

METHODS

The proposed deep-learning framework leverages graph neural network and biological pathway knowledge. We trained and tested our method using ICI-treated patients' data from several clinical trials covering melanoma, gastric cancer, and bladder cancer.

RESULTS

Our results demonstrate that this predictive model outperforms current state-of-the-art methods and tumor microenvironment-based predictors. Additionally, the model quantifies the importance of pathways, pathway interactions, and genes in its predictions. A web server for IRnet has been developed and deployed, providing broad accessibility to users at https://irnet.missouri.edu.

CONCLUSION

IRnet is a competitive tool for predicting patient responses to immunotherapy, specifically ICIs. Its interpretability also offers valuable insights into the mechanisms underlying ICI treatments.

A Transfer Learning Framework for Predicting and Interpreting Drug Responses via Single-Cell RNA-Seq Data

Chemotherapy is a fundamental therapy in cancer treatment, yet its effectiveness is often undermined by drug resistance. Understanding the molecular mechanisms underlying drug response remains a major challenge due to tumor heterogeneity, complex cellular interactions, and limited access to clinical samples, which also hinder the performance and interpretability of existing predictive models. Meanwhile, single-cell RNA sequencing (scRNA-seq) has emerged as a powerful tool for uncovering resistance mechanisms, but the systematic collection and utilization of scRNA-seq drug response data remain limited. In this study, we collected scRNA-seq drug response datasets from publicly available web sources and proposed a transfer learning-based framework to align bulk and single cell sequencing data. A shared encoder was designed to project both bulk and single-cell sequencing data into a unified latent space for drug response prediction, while a sparse decoder guided by prior biological knowledge enhanced interpretability by mapping latent features to predefined pathways. The proposed model achieved superior performance across five curated scRNA-seq datasets and yielded biologically meaningful insights through integrated gradient analysis. This work demonstrates the potential of deep learning to advance drug response prediction and underscores the value of scRNA-seq data in supporting related research.

Advances and applications in single-cell and spatial genomics

The applications of single-cell and spatial technologies in recent times have revolutionized the present understanding of cellular states and the cellular heterogeneity inherent in complex biological systems. These advancements offer unprecedented resolution in the examination of the functional genomics of individual cells and their spatial context within tissues. In this review, we have comprehensively discussed the historical development and recent progress in the field of single-cell and spatial genomics. We have reviewed the breakthroughs in single-cell multi-omics technologies, spatial genomics methods, and the computational strategies employed toward the analyses of single-cell atlas data. Furthermore, we have highlighted the advances made in constructing cellular atlases and their clinical applications, particularly in the context of disease. Finally, we have discussed the emerging trends, challenges, and opportunities in this rapidly evolving field.

Quantitative characterization of cell niches in spatially resolved omics data

Spatial omics enable the characterization of colocalized cell communities that coordinate specific functions within tissues. These communities, or niches, are shaped by interactions between neighboring cells, yet existing computational methods rarely leverage such interactions for their identification and characterization. To address this gap, here we introduce NicheCompass, a graph deep-learning method that models cellular communication to learn interpretable cell embeddings that encode signaling events, enabling the identification of niches and their underlying processes. Unlike existing methods, NicheCompass quantitatively characterizes niches based on communication pathways and consistently outperforms alternatives. We show its versatility by mapping tissue architecture during mouse embryonic development and delineating tumor niches in human cancers, including a spatial reference mapping application. Finally, we extend its capabilities to spatial multi-omics, demonstrate cross-technology integration with datasets from different sequencing platforms and construct a whole mouse brain spatial atlas comprising 8.4 million cells, highlighting NicheCompass' scalability. Overall, NicheCompass provides a scalable framework for identifying and analyzing niches through signaling events.

scVAEDer: integrating deep diffusion models and variational autoencoders for single-cell transcriptomics analysis

Discovering a lower-dimensional embedding of single-cell data can improve downstream analysis. The embedding should encapsulate both the high-level features and low-level variations. While existing generative models attempt to learn such low-dimensional representations, they have limitations. Here, we introduce scVAEDer, a scalable deep-learning model that combines the power of variational autoencoders and deep diffusion models to learn a meaningful representation that retains both global structure and local variations. Using the learned embeddings, scVAEDer can generate novel scRNA-seq data, predict perturbation response on various cell types, identify changes in gene expression during dedifferentiation, and detect master regulators in biological processes.

Strategies to include prior knowledge in omics analysis with deep neural networks

High-throughput molecular profiling technologies have revolutionized molecular biology research in the past decades. One important use of molecular data is to make predictions of phenotypes and other features of the organisms using machine learning algorithms. Deep learning models have become increasingly popular for this task due to their ability to learn complex non-linear patterns. Applying deep learning to molecular profiles, however, is challenging due to the very high dimensionality of the data and relatively small sample sizes, causing models to overfit. A solution is to incorporate biological prior knowledge to guide the learning algorithm for processing the functionally related input together. This helps regularize the models and improve their generalizability and interpretability. Here, we describe three major strategies proposed to use prior knowledge in deep learning models to make predictions based on molecular profiles. We review the related deep learning architectures, including the major ideas in relatively new graph neural networks.

MOSim: bulk and single-cell multilayer regulatory network simulator

As multi-omics sequencing technologies advance, the need for simulation tools capable of generating realistic and diverse (bulk and single-cell) multi-omics datasets for method testing and benchmarking becomes increasingly important. We present MOSim, an R package that simulates both bulk (via mosim function) and single-cell (via sc_mosim function) multi-omics data. The mosim function generates bulk transcriptomics data (RNA-seq) and additional regulatory omics layers (ATAC-seq, miRNA-seq, ChIP-seq, Methyl-seq, and transcription factors), while sc_mosim simulates single-cell transcriptomics data (scRNA-seq) with scATAC-seq and transcription factors as regulatory layers. The tool supports various experimental designs, including simulation of gene co-expression patterns, biological replicates, and differential expression between conditions. MOSim enables users to generate quantification matrices for each simulated omics data type, capturing the heterogeneity and complexity of bulk and single-cell multi-omics datasets. Furthermore, MOSim provides differentially abundant features within each omics layer and elucidates the active regulatory relationships between regulatory omics and gene expression data at both bulk and single-cell levels. By leveraging MOSim, researchers will be able to generate realistic and customizable bulk and single-cell multi-omics datasets to benchmark and validate analytical methods specifically designed for the integrative analysis of diverse regulatory omics data.

Towards the Next Generation of Data-Driven Therapeutics Using Spatially Resolved Single-Cell Technologies and Generative AI

Recent advances in multi-omics and spatially resolved single-cell technologies have revolutionised our ability to profile millions of cellular states, offering unprecedented opportunities to understand the complex molecular landscapes of human tissues in both health and disease. These developments hold immense potential for precision medicine, particularly in the rational design of novel therapeutics for treating inflammatory and autoimmune diseases. However, the vast, high-dimensional data generated by these technologies present significant analytical challenges, such as distinguishing technical variation from biological variation or defining relevant questions that leverage the added spatial dimension to improve our understanding of tissue organisation. Generative artificial intelligence (AI), specifically variational autoencoder- or transformer-based latent variable models, provides a powerful and flexible approach to addressing these challenges. These models make inferences about a cell's intrinsic state by effectively identifying complex patterns, reducing data dimensionality and modelling the biological variability in single-cell datasets. This review explores the current landscape of single-cell and spatial multi-omics technologies, the application of generative AI in data analysis and modelling and their transformative impact on our understanding of autoimmune diseases. By combining spatial and single-cell data with advanced AI methodologies, we highlight novel insights into the pathogenesis of autoimmune disorders and outline future directions for leveraging these technologies to achieve the goal of AI-powered personalised medicine.

BuDDI: Bulk Deconvolution with Domain Invariance to predict cell-type-specific perturbations from bulk

While single-cell experiments provide deep cellular resolution within a single sample, some single-cell experiments are inherently more challenging than bulk experiments due to dissociation difficulties, cost, or limited tissue availability. This creates a situation where we have deep cellular profiles of one sample or condition, and bulk profiles across multiple samples and conditions. To bridge this gap, we propose BuDDI (BUlk Deconvolution with Domain Invariance). BuDDI utilizes domain adaptation techniques to effectively integrate available corpora of case-control bulk and reference scRNA-seq observations to infer cell-type-specific perturbation effects. BuDDI achieves this by learning independent latent spaces within a single variational autoencoder (VAE) encompassing at least four sources of variability: 1) cell type proportion, 2) perturbation effect, 3) structured experimental variability, and 4) remaining variability. Since each latent space is encouraged to be independent, we simulate perturbation responses by independently composing each latent space to simulate cell-type-specific perturbation responses. We evaluated BuDDI's performance on simulated and real data with experimental designs of increasing complexity. We first validated that BuDDI could learn domain invariant latent spaces on data with matched samples across each source of variability. Then we validated that BuDDI could accurately predict cell-type-specific perturbation response when no single-cell perturbed profiles were used during training; instead, only bulk samples had both perturbed and non-perturbed observations. Finally, we validated BuDDI on predicting sex-specific differences, an experimental design where it is not possible to have matched samples. In each experiment, BuDDI outperformed all other comparative methods and baselines. As more reference atlases are completed, BuDDI provides a path to combine these resources with bulk-profiled treatment or disease signatures to study perturbations, sex differences, or other factors at single-cell resolution.

Deep learning models reveal the link between dynamic brain connectivity patterns and states of consciousness

Decoding states of consciousness from brain activity is a central challenge in neuroscience. Dynamic functional connectivity (dFC) allows the study of short-term temporal changes in functional connectivity (FC) between distributed brain areas. By clustering dFC matrices from resting-state fMRI, we previously described "brain patterns" that underlie different functional configurations of the brain at rest. The networks associated with these patterns have been extensively analyzed. However, the overall dynamic organization and how it relates to consciousness remains unclear. We hypothesized that deep learning networks would help to model this relationship. Recent studies have used low-dimensional variational autoencoders (VAE) to learn meaningful representations that can help explaining consciousness. Here, we investigated the complexity of selecting such a generative model to study brain dynamics, and extended the available methods for latent space characterization and modeling. Therefore, our contributions are threefold. First, compared with probabilistic principal component analysis and sparse VAE, we showed that the selected low-dimensional VAE exhibits balanced performance in reconstructing dFCs and classifying brain patterns. We then explored the organization of the obtained low-dimensional dFC latent representations. We showed how these representations stratify the dynamic organization of the brain patterns as well as the experimental conditions. Finally, we proposed to delve into the proposed brain computational model. We first applied a receptive field analysis to identify preferred directions in the latent space to move from one brain pattern to another. Then, an ablation study was achieved where we virtually inactivated specific brain areas. We demonstrated the model's efficiency in summarizing consciousness-specific information encoded in key inter-areal connections, as described in the global neuronal workspace theory of consciousness. The proposed framework advocates the possibility of developing an interpretable computational brain model of interest for disorders of consciousness, paving the way for a dynamic diagnostic support tool.

A mini-review on perturbation modelling across single-cell omic modalities

Recent advances in single-cell omics technology have transformed the landscape of cellular and molecular research, enriching the scope and intricacy of cellular characterisation. Perturbation modelling seeks to comprehensively grasp the effects of external influences like disease onset or molecular knock-outs or external stimulants on cellular physiology, specifically on transcription factors, signal transducers, biological pathways, and dynamic cell states. Machine and deep learning tools transform complex perturbational phenomena in algorithmically tractable tasks to formulate predictions based on various types of single-cell datasets. However, the recent surge in tools and datasets makes it challenging for experimental biologists and computational scientists to keep track of the recent advances in this rapidly expanding filed of single-cell modelling. Here, we recapitulate the main objectives of perturbation modelling and summarise novel single-cell perturbation technologies based on genetic manipulation like CRISPR or compounds, spanning across omic modalities. We then concisely review a burgeoning group of computational methods extending from classical statistical inference methodologies to various machine and deep learning architectures like shallow models or autoencoders, to biologically informed approaches based on gene regulatory networks, and to combinatorial efforts reminiscent of ensemble learning. We also discuss the rising trend of large foundational models in single-cell perturbation modelling inspired by large language models. Lastly, we critically assess the challenges that underline single-cell perturbation modelling while pointing towards relevant future perspectives like perturbation atlases, multi-omics and spatial datasets, causal machine learning for interpretability, multi-task learning for performance and explainability as well as prospects for solving interoperability and benchmarking pitfalls.

Deep Learning of radiology-genomics integration for computational oncology: A mini review

In the field of computational oncology, patient status is often assessed using radiology-genomics, which includes two key technologies and data, such as radiology and genomics. Recent advances in deep learning have facilitated the integration of radiology-genomics data, and even new omics data, significantly improving the robustness and accuracy of clinical predictions. These factors are driving artificial intelligence (AI) closer to practical clinical applications. In particular, deep learning models are crucial in identifying new radiology-genomics biomarkers and therapeutic targets, supported by explainable AI (xAI) methods. This review focuses on recent developments in deep learning for radiology-genomics integration, highlights current challenges, and outlines some research directions for multimodal integration and biomarker discovery of radiology-genomics or radiology-omics that are urgently needed in computational oncology.

iGTP: Learning interpretable cellular embedding for inferring biological mechanisms underlying single-cell transcriptomics

Deep-learning models like Variational AutoEncoder have enabled low dimensional cellular embedding representation for large-scale single-cell transcriptomes and shown great flexibility in downstream tasks. However, biologically meaningful latent space is usually missing if no specific structure is designed. Here, we engineered a novel interpretable generative transcriptional program (iGTP) framework that could model the importance of transcriptional program (TP) space and protein-protein interactions (PPI) between different biological states. We demonstrated the performance of iGTP in a diverse biological context using gene ontology, canonical pathway, and different PPI curation. iGTP not only elucidated the ground truth of cellular responses but also surpassed other deep learning models and traditional bioinformatics methods in functional enrichment tasks. By integrating the latent layer with a graph neural network framework, iGTP could effectively infer cellular responses to perturbations. Lastly, we applied iGTP TP embeddings with a latent diffusion model to accurately generate cell embeddings for specific cell types and states. We anticipate that iGTP will offer insights at both PPI and TP levels and holds promise for predicting responses to novel perturbations.

Latent space arithmetic on data embeddings from healthy multi-tissue human RNA-seq decodes disease modules

Computational analyses of transcriptomic data have dramatically improved our understanding of complex diseases. However, such approaches are limited by small sample sets of disease-affected material. We asked if a variational autoencoder trained on large groups of healthy human RNA sequencing (RNA-seq) data can capture the fundamental gene regulation system and generalize to unseen disease changes. Importantly, we found this model to successfully compress unseen transcriptomic changes from 25 independent disease datasets. We decoded disease-specific signals from the latent space and found them to contain more disease-specific genes than the corresponding differential expression analysis in 20 of 25 cases. Finally, we matched these disease signals with known drug targets and extracted sets of known and potential pharmaceutical candidates. In summary, our study demonstrates how data-driven representation learning enables the arithmetic deconstruction of the latent space, facilitating the dissection of disease mechanisms and drug targets.

Inferring pattern-driving intercellular flows from single-cell and spatial transcriptomics

From single-cell RNA-sequencing (scRNA-seq) and spatial transcriptomics (ST), one can extract high-dimensional gene expression patterns that can be described by intercellular communication networks or decoupled gene modules. These two descriptions of information flow are often assumed to occur independently. However, intercellular communication drives directed flows of information that are mediated by intracellular gene modules, in turn triggering outflows of other signals. Methodologies to describe such intercellular flows are lacking. We present FlowSig, a method that infers communication-driven intercellular flows from scRNA-seq or ST data using graphical causal modeling and conditional independence. We benchmark FlowSig using newly generated experimental cortical organoid data and synthetic data generated from mathematical modeling. We demonstrate FlowSig's utility by applying it to various studies, showing that FlowSig can capture stimulation-induced changes to paracrine signaling in pancreatic islets, demonstrate shifts in intercellular flows due to increasing COVID-19 severity and reconstruct morphogen-driven activator-inhibitor patterns in mouse embryogenesis.

Activity-dependent transcriptional programs in memory regulate motor recovery after stroke

Stroke causes death of brain tissue leading to long-term deficits. Behavioral evidence from neurorehabilitative therapies suggest learning-induced neuroplasticity can lead to beneficial outcomes. However, molecular and cellular mechanisms that link learning and stroke recovery are unknown. We show that in a mouse model of stroke, which exhibits enhanced recovery of function due to genetic perturbations of learning and memory genes, animals display activity-dependent transcriptional programs that are normally active during formation or storage of new memories. The expression of neuronal activity-dependent genes are predictive of recovery and occupy a molecular latent space unique to motor recovery. With motor recovery, networks of activity-dependent genes are co-expressed with their transcription factor targets forming gene regulatory networks that support activity-dependent transcription, that are normally diminished after stroke. Neuronal activity-dependent changes at the circuit level are influenced by interactions with microglia. At the molecular level, we show that enrichment of activity-dependent programs in neurons lead to transcriptional changes in microglia where they differentially interact to support intercellular signaling pathways for axon guidance, growth and synaptogenesis. Together, these studies identify activity-dependent transcriptional programs as a fundamental mechanism for neural repair post-stroke.

Meta graphical lasso: uncovering hidden interactions among latent mechanisms

In complex systems, it's crucial to uncover latent mechanisms and their context-dependent relationships. This is especially true in medical research, where identifying unknown cancer mechanisms and their impact on phenomena like drug resistance is vital. Directly observing these mechanisms is challenging due to measurement complexities, leading to an approach that infers latent mechanisms from observed variable distributions. Despite machine learning advancements enabling sophisticated generative models, their black-box nature complicates the interpretation of complex latent mechanisms. A promising method for understanding these mechanisms involves estimating latent factors through linear projection, though there's no assurance that inferences made under specific conditions will remain valid across contexts. We propose a novel solution, suggesting data, even from systems appearing complex, can often be explained by sparse dependencies among a few common latent factors, regardless of the situation. This simplification allows for modeling that yields significant insights across diverse fields. We demonstrate this with datasets from finance, where we capture societal trends from stock price movements, and medicine, where we uncover new insights into cancer drug resistance through gene expression analysis.

Designing interpretable deep learning applications for functional genomics: a quantitative analysis

Deep learning applications have had a profound impact on many scientific fields, including functional genomics. Deep learning models can learn complex interactions between and within omics data; however, interpreting and explaining these models can be challenging. Interpretability is essential not only to help progress our understanding of the biological mechanisms underlying traits and diseases but also for establishing trust in these model's efficacy for healthcare applications. Recognizing this importance, recent years have seen the development of numerous diverse interpretability strategies, making it increasingly difficult to navigate the field. In this review, we present a quantitative analysis of the challenges arising when designing interpretable deep learning solutions in functional genomics. We explore design choices related to the characteristics of genomics data, the neural network architectures applied, and strategies for interpretation. By quantifying the current state of the field with a predefined set of criteria, we find the most frequent solutions, highlight exceptional examples, and identify unexplored opportunities for developing interpretable deep learning models in genomics.

Interpretable deep learning in single-cell omics

MOTIVATION

Single-cell omics technologies have enabled the quantification of molecular profiles in individual cells at an unparalleled resolution. Deep learning, a rapidly evolving sub-field of machine learning, has instilled a significant interest in single-cell omics research due to its remarkable success in analysing heterogeneous high-dimensional single-cell omics data. Nevertheless, the inherent multi-layer nonlinear architecture of deep learning models often makes them 'black boxes' as the reasoning behind predictions is often unknown and not transparent to the user. This has stimulated an increasing body of research for addressing the lack of interpretability in deep learning models, especially in single-cell omics data analyses, where the identification and understanding of molecular regulators are crucial for interpreting model predictions and directing downstream experimental validations.

RESULTS

In this work, we introduce the basics of single-cell omics technologies and the concept of interpretable deep learning. This is followed by a review of the recent interpretable deep learning models applied to various single-cell omics research. Lastly, we highlight the current limitations and discuss potential future directions.

Deep generative models in single-cell omics

Deep Generative Models (DGMs) are becoming instrumental for inferring probability distributions inherent to complex processes, such as most questions in biomedical research. For many years, there was a lack of mathematical methods that would allow this inference in the scarce data scenario of biomedical research. The advent of single-cell omics has finally made square the so-called "skinny matrix", allowing to apply mathematical methods already extensively used in other areas. Moreover, it is now possible to integrate data at different molecular levels in thousands or even millions of samples, thanks to the number of single-cell atlases being collaboratively generated. Additionally, DGMs have proven useful in other frequent tasks in single-cell analysis pipelines, from dimensionality reduction, cell type annotation to RNA velocity inference. In spite of its promise, DGMs need to be used with caution in biomedical research, paying special attention to its use to answer the right questions and the definition of appropriate error metrics and validation check points that confirm not only its correct use but also its relevance. All in all, DGMs provide an exciting tool that opens a bright future for the integrative analysis of single-cell -omics to understand health and disease.

The manatee variational autoencoder model for predicting gene expression alterations caused by transcription factor perturbations

We present the Manatee variational autoencoder model to predict transcription factor (TF) perturbation-induced transcriptomes. We demonstrate that the Manatee in silico perturbation analysis recapitulates target transcriptomic phenotypes in diverse cellular lineage transitions. We further propose the Manatee in silico screening analysis for prioritizing TF combinations targeting desired transcriptomic phenotypes.

Comprehensive single-cell RNA-seq analysis using deep interpretable generative modeling guided by biological hierarchy knowledge

Recent advances in microfluidics and sequencing technologies allow researchers to explore cellular heterogeneity at single-cell resolution. In recent years, deep learning frameworks, such as generative models, have brought great changes to the analysis of transcriptomic data. Nevertheless, relying on the potential space of these generative models alone is insufficient to generate biological explanations. In addition, most of the previous work based on generative models is limited to shallow neural networks with one to three layers of latent variables, which may limit the capabilities of the models. Here, we propose a deep interpretable generative model called d-scIGM for single-cell data analysis. d-scIGM combines sawtooth connectivity techniques and residual networks, thereby constructing a deep generative framework. In addition, d-scIGM incorporates hierarchical prior knowledge of biological domains to enhance the interpretability of the model. We show that d-scIGM achieves excellent performance in a variety of fundamental tasks, including clustering, visualization, and pseudo-temporal inference. Through topic pathway studies, we found that d-scIGM-learned topics are better enriched for biologically meaningful pathways compared to the baseline models. Furthermore, the analysis of drug response data shows that d-scIGM can capture drug response patterns in large-scale experiments, which provides a promising way to elucidate the underlying biological mechanisms. Lastly, in the melanoma dataset, d-scIGM accurately identified different cell types and revealed multiple melanin-related driver genes and key pathways, which are critical for understanding disease mechanisms and drug development.

Pathway-based analyses of gene expression profiles at low doses of ionizing radiation

Radiation exposure poses a significant threat to human health. Emerging research indicates that even low-dose radiation once believed to be safe, may have harmful effects. This perception has spurred a growing interest in investigating the potential risks associated with low-dose radiation exposure across various scenarios. To comprehensively explore the health consequences of low-dose radiation, our study employs a robust statistical framework that examines whether specific groups of genes, belonging to known pathways, exhibit coordinated expression patterns that align with the radiation levels. Notably, our findings reveal the existence of intricate yet consistent signatures that reflect the molecular response to radiation exposure, distinguishing between low-dose and high-dose radiation. Moreover, we leverage a pathway-constrained variational autoencoder to capture the nonlinear interactions within gene expression data. By comparing these two analytical approaches, our study aims to gain valuable insights into the impact of low-dose radiation on gene expression patterns, identify pathways that are differentially affected, and harness the potential of machine learning to uncover hidden activity within biological networks. This comparative analysis contributes to a deeper understanding of the molecular consequences of low-dose radiation exposure.

MUSE-XAE: MUtational Signature Extraction with eXplainable AutoEncoder enhances tumour types classification

MOTIVATION

Mutational signatures are a critical component in deciphering the genetic alterations that underlie cancer development and have become a valuable resource to understand the genomic changes during tumorigenesis. Therefore, it is essential to employ precise and accurate methods for their extraction to ensure that the underlying patterns are reliably identified and can be effectively utilized in new strategies for diagnosis, prognosis, and treatment of cancer patients.

RESULTS

We present MUSE-XAE, a novel method for mutational signature extraction from cancer genomes using an explainable autoencoder. Our approach employs a hybrid architecture consisting of a nonlinear encoder that can capture nonlinear interactions among features, and a linear decoder which ensures the interpretability of the active signatures. We evaluated and compared MUSE-XAE with other available tools on both synthetic and real cancer datasets and demonstrated that it achieves superior performance in terms of precision and sensitivity in recovering mutational signature profiles. MUSE-XAE extracts highly discriminative mutational signature profiles by enhancing the classification of primary tumour types and subtypes in real world settings. This approach could facilitate further research in this area, with neural networks playing a critical role in advancing our understanding of cancer genomics.

AVAILABILITY AND IMPLEMENTATION

MUSE-XAE software is freely available at https://github.com/compbiomed-unito/MUSE-XAE.

BuDDI: BulkDeconvolution withDomainInvariance to predict cell-type-specific perturbations from bulk

While single-cell experiments provide deep cellular resolution within a single sample, some single-cell experiments are inherently more challenging than bulk experiments due to dissociation difficulties, cost, or limited tissue availability. This creates a situation where we have deep cellular profiles of one sample or condition, and bulk profiles across multiple samples and conditions. To bridge this gap, we propose BuDDI (BUlk Deconvolution with Domain Invariance). BuDDI utilizes domain adaptation techniques to effectively integrate available corpora of case-control bulk and reference scRNA-seq observations to infer cell-type-specific perturbation effects. BuDDI achieves this by learning independent latent spaces within a single variational autoencoder (VAE) encompassing at least four sources of variability: 1) cell type proportion, 2) perturbation effect, 3) structured experimental variability, and 4) remaining variability. Since each latent space is encouraged to be independent, we simulate perturbation responses by independently composing each latent space to simulate cell-type-specific perturbation responses. We evaluated BuDDI's performance on simulated and real data with experimental designs of increasing complexity. We first validated that BuDDI could learn domain invariant latent spaces on data with matched samples across each source of variability. Then we validated that BuDDI could accurately predict cell-type-specific perturbation response when no single-cell perturbed profiles were used during training; instead, only bulk samples had both perturbed and non-perturbed observations. Finally, we validated BuDDI on predicting sex-specific differences, an experimental design where it is not possible to have matched samples. In each experiment, BuDDI outperformed all other comparative methods and baselines. As more reference atlases are completed, BuDDI provides a path to combine these resources with bulk-profiled treatment or disease signatures to study perturbations, sex differences, or other factors at single-cell resolution.

scHybridBERT: integrating gene regulation and cell graph for spatiotemporal dynamics in single-cell clustering

Graph learning models have received increasing attention in the computational analysis of single-cell RNA sequencing (scRNA-seq) data. Compared with conventional deep neural networks, graph neural networks and language models have exhibited superior performance by extracting graph-structured data from raw gene count matrices. Established deep neural network-based clustering approaches generally focus on temporal expression patterns while ignoring inherent interactions at gene-level as well as cell-level, which could be regarded as spatial dynamics in single-cell data. Both gene-gene and cell-cell interactions are able to boost the performance of cell type detection, under the framework of multi-view modeling. In this study, spatiotemporal embedding and cell graphs are extracted to capture spatial dynamics at the molecular level. In order to enhance the accuracy of cell type detection, this study proposes the scHybridBERT architecture to conduct multi-view modeling of scRNA-seq data using extracted spatiotemporal patterns. In this scHybridBERT method, graph learning models are employed to deal with cell graphs and the Performer model employs spatiotemporal embeddings. Experimental outcomes about benchmark scRNA-seq datasets indicate that the proposed scHybridBERT method is able to enhance the accuracy of single-cell clustering tasks by integrating spatiotemporal embeddings and cell graphs.

Single-cell spatial metabolomics with cell-type specific protein profiling for tissue systems biology

Metabolic reprogramming in cancer and immune cells occurs to support their increasing energy needs in biological tissues. Here we propose Single Cell Spatially resolved Metabolic (scSpaMet) framework for joint protein-metabolite profiling of single immune and cancer cells in male human tissues by incorporating untargeted spatial metabolomics and targeted multiplexed protein imaging in a single pipeline. We utilized the scSpaMet to profile cell types and spatial metabolomic maps of 19507, 31156, and 8215 single cells in human lung cancer, tonsil, and endometrium tissues, respectively. The scSpaMet analysis revealed cell type-dependent metabolite profiles and local metabolite competition of neighboring single cells in human tissues. Deep learning-based joint embedding revealed unique metabolite states within cell types. Trajectory inference showed metabolic patterns along cell differentiation paths. Here we show scSpaMet's ability to quantify and visualize the cell-type specific and spatially resolved metabolic-protein mapping as an emerging tool for systems-level understanding of tissue biology.

Autoencoder neural networks enable low dimensional structure analyses of microbial growth dynamics

The ability to effectively represent microbiome dynamics is a crucial challenge in their quantitative analysis and engineering. By using autoencoder neural networks, we show that microbial growth dynamics can be compressed into low-dimensional representations and reconstructed with high fidelity. These low-dimensional embeddings are just as effective, if not better, than raw data for tasks such as identifying bacterial strains, predicting traits like antibiotic resistance, and predicting community dynamics. Additionally, we demonstrate that essential dynamical information of these systems can be captured using far fewer variables than traditional mechanistic models. Our work suggests that machine learning can enable the creation of concise representations of high-dimensional microbiome dynamics to facilitate data analysis and gain new biological insights.

Explainable artificial intelligence for omics data: a systematic mapping study

Researchers increasingly turn to explainable artificial intelligence (XAI) to analyze omics data and gain insights into the underlying biological processes. Yet, given the interdisciplinary nature of the field, many findings have only been shared in their respective research community. An overview of XAI for omics data is needed to highlight promising approaches and help detect common issues. Toward this end, we conducted a systematic mapping study. To identify relevant literature, we queried Scopus, PubMed, Web of Science, BioRxiv, MedRxiv and arXiv. Based on keywording, we developed a coding scheme with 10 facets regarding the studies' AI methods, explainability methods and omics data. Our mapping study resulted in 405 included papers published between 2010 and 2023. The inspected papers analyze DNA-based (mostly genomic), transcriptomic, proteomic or metabolomic data by means of neural networks, tree-based methods, statistical methods and further AI methods. The preferred post-hoc explainability methods are feature relevance (n = 166) and visual explanation (n = 52), while papers using interpretable approaches often resort to the use of transparent models (n = 83) or architecture modifications (n = 72). With many research gaps still apparent for XAI for omics data, we deduced eight research directions and discuss their potential for the field. We also provide exemplary research questions for each direction. Many problems with the adoption of XAI for omics data in clinical practice are yet to be resolved. This systematic mapping study outlines extant research on the topic and provides research directions for researchers and practitioners.

Attention-based deep clustering method for scRNA-seq cell type identification

Single-cell sequencing (scRNA-seq) technology provides higher resolution of cellular differences than bulk RNA sequencing and reveals the heterogeneity in biological research. The analysis of scRNA-seq datasets is premised on the subpopulation assignment. When an appropriate reference is not available, such as specific marker genes and single-cell reference atlas, unsupervised clustering approaches become the predominant option. However, the inherent sparsity and high-dimensionality of scRNA-seq datasets pose specific analytical challenges to traditional clustering methods. Therefore, a various deep learning-based methods have been proposed to address these challenges. As each method improves partially, a comprehensive method needs to be proposed. In this article, we propose a novel scRNA-seq data clustering method named AttentionAE-sc (Attention fusion AutoEncoder for single-cell). Two different scRNA-seq clustering strategies are combined through an attention mechanism, that include zero-inflated negative binomial (ZINB)-based methods dealing with the impact of dropout events and graph autoencoder (GAE)-based methods relying on information from neighbors to guide the dimension reduction. Based on an iterative fusion between denoising and topological embeddings, AttentionAE-sc can easily acquire clustering-friendly cell representations that similar cells are closer in the hidden embedding. Compared with several state-of-art baseline methods, AttentionAE-sc demonstrated excellent clustering performance on 16 real scRNA-seq datasets without the need to specify the number of groups. Additionally, AttentionAE-sc learned improved cell representations and exhibited enhanced stability and robustness. Furthermore, AttentionAE-sc achieved remarkable identification in a breast cancer single-cell atlas dataset and provided valuable insights into the heterogeneity among different cell subtypes.

A Spatiotemporal and Machine-Learning Platform Accelerates the Manufacturing of hPSC-derived Esophageal Mucosa

Human pluripotent stem cell-derived tissue engineering offers great promise in designer cell-based personalized therapeutics. To harness such potential, a broader approach requires a deeper understanding of tissue-level interactions. We previously developed a manufacturing system for the ectoderm-derived skin epithelium for cell replacement therapy. However, it remains challenging to manufacture the endoderm-derived esophageal epithelium, despite both possessing similar stratified structure. Here we employ single cell and spatial technologies to generate a spatiotemporal multi-omics cell atlas for human esophageal development. We illuminate the cellular diversity, dynamics and signal communications for the developing esophageal epithelium and stroma. Using the machine-learning based Manatee, we prioritize the combinations of candidate human developmental signals for in vitro derivation of esophageal basal cells. Functional validation of the Manatee predictions leads to a clinically-compatible system for manufacturing human esophageal mucosa. Our approach creates a versatile platform to accelerate human tissue manufacturing for future cell replacement therapies to treat human genetic defects and wounds.

Deep convolutional and conditional neural networks for large-scale genomic data generation

Applications of generative models for genomic data have gained significant momentum in the past few years, with scopes ranging from data characterization to generation of genomic segments and functional sequences. In our previous study, we demonstrated that generative adversarial networks (GANs) and restricted Boltzmann machines (RBMs) can be used to create novel high-quality artificial genomes (AGs) which can preserve the complex characteristics of real genomes such as population structure, linkage disequilibrium and selection signals. However, a major drawback of these models is scalability, since the large feature space of genome-wide data increases computational complexity vastly. To address this issue, we implemented a novel convolutional Wasserstein GAN (WGAN) model along with a novel conditional RBM (CRBM) framework for generating AGs with high SNP number. These networks implicitly learn the varying landscape of haplotypic structure in order to capture complex correlation patterns along the genome and generate a wide diversity of plausible haplotypes. We performed comparative analyses to assess both the quality of these generated haplotypes and the amount of possible privacy leakage from the training data. As the importance of genetic privacy becomes more prevalent, the need for effective privacy protection measures for genomic data increases. We used generative neural networks to create large artificial genome segments which possess many characteristics of real genomes without substantial privacy leakage from the training dataset. In the near future, with further improvements in haplotype quality and privacy preservation, large-scale artificial genome databases can be assembled to provide easily accessible surrogates of real databases, allowing researchers to conduct studies with diverse genomic data within a safe ethical framework in terms of donor privacy.

NCAE: data-driven representations using a deep network-coherent DNA methylation autoencoder identify robust disease and risk factor signatures

Precision medicine relies on the identification of robust disease and risk factor signatures from omics data. However, current knowledge-driven approaches may overlook novel or unexpected phenomena due to the inherent biases in biological knowledge. In this study, we present a data-driven signature discovery workflow for DNA methylation analysis utilizing network-coherent autoencoders (NCAEs) with biologically relevant latent embeddings. First, we explored the architecture space of autoencoders trained on a large-scale pan-tissue compendium (n = 75 272) of human epigenome-wide association studies. We observed the emergence of co-localized patterns in the deep autoencoder latent space representations that corresponded to biological network modules. We determined the NCAE configuration with the strongest co-localization and centrality signals in the human protein interactome. Leveraging the NCAE embeddings, we then trained interpretable deep neural networks for risk factor (aging, smoking) and disease (systemic lupus erythematosus) prediction and classification tasks. Remarkably, our NCAE embedding-based models outperformed existing predictors, revealing novel DNA methylation signatures enriched in gene sets and pathways associated with the studied condition in each case. Our data-driven biomarker discovery workflow provides a generally applicable pipeline to capture relevant risk factor and disease information. By surpassing the limitations of knowledge-driven methods, our approach enhances the understanding of complex epigenetic processes, facilitating the development of more effective diagnostic and therapeutic strategies.

The Deep Generative Decoder: MAP estimation of representations improves modelling of single-cell RNA data

MOTIVATION

Learning low-dimensional representations of single-cell transcriptomics has become instrumental to its downstream analysis. The state of the art is currently represented by neural network models, such as variational autoencoders, which use a variational approximation of the likelihood for inference.

RESULTS

We here present the Deep Generative Decoder (DGD), a simple generative model that computes model parameters and representations directly via maximum a posteriori estimation. The DGD handles complex parameterized latent distributions naturally unlike variational autoencoders, which typically use a fixed Gaussian distribution, because of the complexity of adding other types. We first show its general functionality on a commonly used benchmark set, Fashion-MNIST. Secondly, we apply the model to multiple single-cell datasets. Here, the DGD learns low-dimensional, meaningful, and well-structured latent representations with sub-clustering beyond the provided labels. The advantages of this approach are its simplicity and its capability to provide representations of much smaller dimensionality than a comparable variational autoencoder.

AVAILABILITY AND IMPLEMENTATION

scDGD is available as a python package at https://github.com/Center-for-Health-Data-Science/scDGD. The remaining code is made available here: https://github.com/Center-for-Health-Data-Science/dgd.

A Roadmap for a Consensus Human Skin Cell Atlas and Single-Cell Data Standardization

Single-cell technologies have become essential to driving discovery in both basic and translational investigative dermatology. Despite the multitude of available datasets, a central reference atlas of normal human skin, which can serve as a reference resource for skin cell types, cell states, and their molecular signatures, is still lacking. For any such atlas to receive broad acceptance, participation by many investigators during atlas construction is an essential prerequisite. As part of the Human Cell Atlas project, we have assembled a Skin Biological Network to build a consensus Human Skin Cell Atlas and outline a roadmap toward that goal. We define the drivers of skin diversity to be considered when selecting sequencing datasets for the atlas and list practical hurdles during skin sampling that can result in data gaps and impede comprehensive representation and technical considerations for tissue processing and computational analysis, the accounting for which should minimize biases in cell type enrichments and exclusions and decrease batch effects. By outlining our goals for Atlas 1.0, we discuss how it will uncover new aspects of skin biology.

An Overview of Deep Generative Models in Functional and Evolutionary Genomics

Following the widespread use of deep learning for genomics, deep generative modeling is also becoming a viable methodology for the broad field. Deep generative models (DGMs) can learn the complex structure of genomic data and allow researchers to generate novel genomic instances that retain the real characteristics of the original dataset. Aside from data generation, DGMs can also be used for dimensionality reduction by mapping the data space to a latent space, as well as for prediction tasks via exploitation of this learned mapping or supervised/semi-supervised DGM designs. In this review, we briefly introduce generative modeling and two currently prevailing architectures, we present conceptual applications along with notable examples in functional and evolutionary genomics, and we provide our perspective on potential challenges and future directions.

A systematic review of biologically-informed deep learning models for cancer: fundamental trends for encoding and interpreting oncology data

BACKGROUND

There is an increasing interest in the use of Deep Learning (DL) based methods as a supporting analytical framework in oncology. However, most direct applications of DL will deliver models with limited transparency and explainability, which constrain their deployment in biomedical settings.

METHODS

This systematic review discusses DL models used to support inference in cancer biology with a particular emphasis on multi-omics analysis. It focuses on how existing models address the need for better dialogue with prior knowledge, biological plausibility and interpretability, fundamental properties in the biomedical domain. For this, we retrieved and analyzed 42 studies focusing on emerging architectural and methodological advances, the encoding of biological domain knowledge and the integration of explainability methods.

RESULTS

We discuss the recent evolutionary arch of DL models in the direction of integrating prior biological relational and network knowledge to support better generalisation (e.g. pathways or Protein-Protein-Interaction networks) and interpretability. This represents a fundamental functional shift towards models which can integrate mechanistic and statistical inference aspects. We introduce a concept of bio-centric interpretability and according to its taxonomy, we discuss representational methodologies for the integration of domain prior knowledge in such models.

CONCLUSIONS

The paper provides a critical outlook into contemporary methods for explainability and interpretability used in DL for cancer. The analysis points in the direction of a convergence between encoding prior knowledge and improved interpretability. We introduce bio-centric interpretability which is an important step towards formalisation of biological interpretability of DL models and developing methods that are less problem- or application-specific.

Utilization of Computer Classification Methods for Exposure Prediction and Gene Selection in Daphnia magna Toxicogenomics

Zinc (Zn) is an essential element that influences many cellular functions. Depending on bioavailability, Zn can cause both deficiency and toxicity. Zn bioavailability is influenced by water hardness. Therefore, water quality analysis for health-risk assessment should consider both Zn concentration and water hardness. However, exposure media selection for traditional toxicology tests are set to defined hardness levels and do not represent the diverse water chemistry compositions observed in nature. Moreover, these tests commonly use whole organism endpoints, such as survival and reproduction, which require high numbers of test animals and are labor intensive. Gene expression stands out as a promising alternative to provide insight into molecular events that can be used for risk assessment. In this work, we apply machine learning techniques to classify the Zn concentrations and water hardness from Daphnia magna gene expression by using quantitative PCR. A method for gene ranking was explored using techniques from game theory, namely, Shapley values. The results show that standard machine learning classifiers can classify both Zn concentration and water hardness simultaneously, and that Shapley values are a versatile and useful alternative for gene ranking that can provide insight about the importance of individual genes.

PAUSE: principled feature attribution for unsupervised gene expression analysis

As interest in using unsupervised deep learning models to analyze gene expression data has grown, an increasing number of methods have been developed to make these models more interpretable. These methods can be separated into two groups: post hoc analyses of black box models through feature attribution methods and approaches to build inherently interpretable models through biologically-constrained architectures. We argue that these approaches are not mutually exclusive, but can in fact be usefully combined. We propose PAUSE ( https://github.com/suinleelab/PAUSE ), an unsupervised pathway attribution method that identifies major sources of transcriptomic variation when combined with biologically-constrained neural network models.

clusterMaker2: a major update to clusterMaker, a multi-algorithm clustering app for Cytoscape

BACKGROUND

Since the initial publication of clusterMaker, the need for tools to analyze large biological datasets has only increased. New datasets are significantly larger than a decade ago, and new experimental techniques such as single-cell transcriptomics continue to drive the need for clustering or classification techniques to focus on portions of datasets of interest. While many libraries and packages exist that implement various algorithms, there remains the need for clustering packages that are easy to use, integrated with visualization of the results, and integrated with other commonly used tools for biological data analysis. clusterMaker2 has added several new algorithms, including two entirely new categories of analyses: node ranking and dimensionality reduction. Furthermore, many of the new algorithms have been implemented using the Cytoscape jobs API, which provides a mechanism for executing remote jobs from within Cytoscape. Together, these advances facilitate meaningful analyses of modern biological datasets despite their ever-increasing size and complexity.

RESULTS

The use of clusterMaker2 is exemplified by reanalyzing the yeast heat shock expression experiment that was included in our original paper; however, here we explored this dataset in significantly more detail. Combining this dataset with the yeast protein-protein interaction network from STRING, we were able to perform a variety of analyses and visualizations from within clusterMaker2, including Leiden clustering to break the entire network into smaller clusters, hierarchical clustering to look at the overall expression dataset, dimensionality reduction using UMAP to find correlations between our hierarchical visualization and the UMAP plot, fuzzy clustering, and cluster ranking. Using these techniques, we were able to explore the highest-ranking cluster and determine that it represents a strong contender for proteins working together in response to heat shock. We found a series of clusters that, when re-explored as fuzzy clusters, provide a better presentation of mitochondrial processes.

CONCLUSIONS

clusterMaker2 represents a significant advance over the previously published version, and most importantly, provides an easy-to-use tool to perform clustering and to visualize clusters within the Cytoscape network context. The new algorithms should be welcome to the large population of Cytoscape users, particularly the new dimensionality reduction and fuzzy clustering techniques.

siVAE: interpretable deep generative models for single-cell transcriptomes

Neural networks such as variational autoencoders (VAE) perform dimensionality reduction for the visualization and analysis of genomic data, but are limited in their interpretability: it is unknown which data features are represented by each embedding dimension. We present siVAE, a VAE that is interpretable by design, thereby enhancing downstream analysis tasks. Through interpretation, siVAE also identifies gene modules and hubs without explicit gene network inference. We use siVAE to identify gene modules whose connectivity is associated with diverse phenotypes such as iPSC neuronal differentiation efficiency and dementia, showcasing the wide applicability of interpretable generative models for genomic data analysis.

Biologically informed deep learning to query gene programs in single-cell atlases

The increasing availability of large-scale single-cell atlases has enabled the detailed description of cell states. In parallel, advances in deep learning allow rapid analysis of newly generated query datasets by mapping them into reference atlases. However, existing data transformations learned to map query data are not easily explainable using biologically known concepts such as genes or pathways. Here we propose expiMap, a biologically informed deep-learning architecture that enables single-cell reference mapping. ExpiMap learns to map cells into biologically understandable components representing known 'gene programs'. The activity of each cell for a gene program is learned while simultaneously refining them and learning de novo programs. We show that expiMap compares favourably to existing methods while bringing an additional layer of interpretability to integrative single-cell analysis. Furthermore, we demonstrate its applicability to analyse single-cell perturbation responses in different tissues and species and resolve responses of patients who have coronavirus disease 2019 to different treatments across cell types.

MetaSEM: Gene Regulatory Network Inference from Single-Cell RNA Data by Meta-Learning

Regulators in gene regulatory networks (GRNs) are crucial for identifying cell states. However, GRN inference based on scRNA-seq data has several problems, including high dimensionality and sparsity, and requires more label data. Therefore, we propose a meta-learning GRN inference framework to identify regulatory factors. Specifically, meta-learning solves the parameter optimization problem caused by high-dimensional sparse data features. In addition, a few-shot solution was used to solve the problem of lack of label data. A structural equation model (SEM) was embedded in the model to identify important regulators. We integrated the parameter optimization strategy into the bi-level optimization to extract the feature consistent with GRN reasoning. This unique design makes our model robust to small-scale data. By studying the GRN inference task, we confirmed that the selected regulators were closely related to gene expression specificity. We further analyzed the GRN inferred to find the important regulators in cell type identification. Extensive experimental results showed that our model effectively captured the regulator in single-cell GRN inference. Finally, the visualization results verified the importance of the selected regulators for cell type recognition.

scCapsNet-mask: an updated version of scCapsNet with extended applicability in functional analysis related to scRNA-seq data

BACKGROUND

With the rapid accumulation of scRNA-seq data, more and more automatic cell type identification methods have been developed, especially those based on deep learning. Although these methods have reached relatively high prediction accuracy, many issues still exist. One is the interpretability. The second is how to deal with the non-standard test samples that are not encountered in the training process.

RESULTS

Here we introduce scCapsNet-mask, an updated version of scCapsNet. The scCapsNet-mask provides a reasonable solution to the issues of interpretability and non-standard test samples. Firstly, the scCapsNet-mask utilizes a mask to ease the task of model interpretation in the original scCapsNet. The results show that scCapsNet-mask could constrain the coupling coefficients, and make a one-to-one correspondence between the primary capsules and type capsules. Secondly, the scCapsNet-mask can process non-standard samples more reasonably. In one example, the scCapsNet-mask was trained on the committed cells, and then tested on less differentiated cells as the non-standard samples. It could not only estimate the lineage bias of less differentiated cells, but also distinguish the development stages more accurately than traditional machine learning models. Therefore, the pseudo-temporal order of cells for each lineage could be established. Following these pseudo-temporal order, lineage specific genes exhibit a gradual increase expression pattern and stem cell associated genes exhibit a gradual decrease expression pattern. In another example, the scCapsNet-mask was trained on scRNA-seq data, and then used to assign cell type in spatial transcriptomics that may contain non-standard sample of doublets. The results show that the scCapsNet-mask not only restored the spatial map but also identified several non-standard samples of doublet.

CONCLUSIONS

The scCapsNet-mask offers a suitable solution to the challenge of interpretability and non-standard test samples. By adding a mask, it has the advantages of automatic processing and easy interpretation compared with the original scCapsNet. In addition, the scCapsNet-mask could more accurately reflect the composition of non-standard test samples than traditional machine learning methods. Therefore, it can extend its applicability in functional analysis, such as fate bias prediction in less differentiated cells and cell type assignment in spatial transcriptomics.