Search

1AbstractWe introduce a framework for end-to-end integrative modeling of 3D single-cell multi-channel fluorescent image data of diverse subcellular structures. We employ stacked conditional β-variational autoencoders to first learn a latent representation of cell morphology, and then learn a latent representation of subcellular structure localization which is conditioned on the learned cell morphology. Our model is flexible and can be trained on images of arbitrary subcellular structures and at varying degrees of sparsity and reconstruction fidelity. We train our full model on 3D cell image data and explore design trade-offs in the 2D setting. Once trained, our model can be used to impute structures in cells where they were not imaged and to quantify the variation in the location of all subcellular structures by generating plausible instantiations of each structure in arbitrary cell geometries. We apply our trained model to a small drug perturbation screen to demonstrate its applicability to new data. We show how the latent representations of drugged cells differ from unperturbed cells as expected by on-target effects of the drugs.2Author summaryIt’s impossible to acquire all the information we want about every cell we’re interested in in a single experiment. Even just limiting ourselves to imaging, we can only image a small set of subcellular structures in each cell. If we are interested in integrating those images into a holistic picture of cellular organization directly from data, there are a number of approaches one might take. Here, we leverage the fact that of the three channels we image in each cell, two stay the same across the data set; these two channels assess the cell’s shape and nuclear morphology. Given these two reference channels, we learn a model of cell and nuclear morphology, and then use this as a reference frame in which to learn a representation of the localization of each subcellular structure as measured by the third channel. We use β-variational autoencoders to learn representations of both the reference channels and representations of each subcellular structure (conditioned on the reference channels of the cell in which it was imaged). Since these models are both probabilistic and generative, we can use them to understand the variation in the data from which they were trained, to generate instantiations of new cell morphologies, and to generate imputations of structures in real cell images to create an integrated model of subcellular organization.

We performed a comprehensive analysis of the transcriptional changes within and across cell populations during human induced pluripotent stem cell (hiPSC) differentiation to cardiomyocytes. Using the single cell RNA-seq combinatorial barcoding method SPLiT-seq, we sequenced >20,000 single cells from 55 independent samples representing two differentiation protocols and multiple hiPSC lines. Samples included experimental replicates ranging from undifferentiated hiPSCs to mixed populations of cells at D90 post-differentiation. As expected, differentiated cell populations clustered by time point, with differential expression analysis revealing markers of cardiomyocyte differentiation and maturation changing from D12 to D90. We next performed a complementary cluster-independent sparse regression analysis to identify and rank genes that best assigned cells to differentiation time points. The two highest ranked genes between D12 and D24 (MYH7 and MYH6) resulted in an accuracy of 0.84, and the three highest ranked genes between D24 and D90 (A2M, H19, IGF2) resulted in an accuracy of 0.94, revealing that low dimensional gene features can identify differentiation or maturation stages in differentiating cardiomyocytes. Expression levels of select genes were validated using RNA FISH. Finally, we interrogated differences in differentiation population composition and cardiac gene expression resulting from two differentiation protocols, experimental replicates, and three hiPSC lines in the WTC-11 background to identify sources of variation across these experimental variables.

SummaryWe present a quantitative co-analysis of RNA abundance and sarcomere organization in single cells and an integrated framework to predict subcellular organization states from gene expression. We used human induced pluripotent stem cell (hiPSC)-derived cardiomyocytes expressing mEGFP-tagged alpha-actinin-2 to develop quantitative image analysis tools for systematic and automated classification of subcellular organization. This captured a wide range of sarcomeric organization states within cell populations that were previously difficult to quantify. We performed RNA FISH targeting genes identified by single cell RNA sequencing to simultaneously assess the relationship between transcript abundance and structural states in single cells. Co-analysis of gene expression and sarcomeric patterns in the same cells revealed biologically meaningful correlations that could be used to predict organizational states. This study establishes a framework for multi-dimensional analysis of single cells to study the relationships between gene expression and subcellular organization and to develop a more nuanced description of cell states.Graphical AbstractTranscriptional profiling and structural classification was performed on human induced pluripotent stem cell-derived cardiomyocytes to characterize the relationship between transcript abundance and subcellular organization.

Although some cell types may be defined anatomically or by physiological function, a rigorous definition of cell state remains elusive. Here, we develop a quantitative, imaging-based platform for the systematic and automated classification of subcellular organization in single cells. We use this platform to quantify subcellular organization and gene expression in >30,000 individual human induced pluripotent stem cell-derived cardiomyocytes, producing a publicly available dataset that describes the population distributions of local and global sarcomere organization, mRNA abundance, and correlations between these traits. While the mRNA abundance of some phenotypically important genes correlates with subcellular organization (e.g., the beta-myosin heavy chain, MYH7), these two cellular metrics are heterogeneous and often uncorrelated, which suggests that gene expression alone is not sufficient to classify cell states. Instead, we posit that cell state should be defined by observing full distributions of quantitative, multidimensional traits in single cells that also account for space, time, and function.

Motivated by the extremely high computing costs associated with estimates of free energies for biological systems using molecular simulations, we further the exploration of existing "belief propagation" (BP) algorithms for fixed-backbone peptide and protein systems. The precalculation of pairwise interactions among discretized libraries of side-chain conformations, along with representation of protein side chains as nodes in a graphical model, enables direct application of the BP approach, which requires only ∼1 s of single-processor run time after the precalculation stage. We use a "loopy BP" algorithm, which can be seen as an approximate generalization of the transfer-matrix approach to highly connected (i.e., loopy) graphs, and it has previously been applied to protein calculations. We examine the application of loopy BP to several peptides as well as the binding site of the T4 lysozyme L99A mutant. The present study reports on (i) the comparison of the approximate BP results with estimates from unbiased estimators based on the Amber99SB force field; (ii) investigation of the effects of varying library size on BP predictions; and (iii) a theoretical discussion of the discretization effects that can arise in BP calculations. The data suggest that, despite their approximate nature, BP free-energy estimates are highly accurate-indeed, they never fall outside confidence intervals from unbiased estimators for the systems where independent results could be obtained. Furthermore, we find that libraries of sufficiently fine discretization (which diminish library-size sensitivity) can be obtained with standard computing resources in most cases. Altogether, the extremely low computing times and accurate results suggest the BP approach warrants further study.

There is intense interest in mapping the tissue-specific binding sites of transcription factors in the human genome to reconstruct gene regulatory networks and predict functions for non-coding genetic variation. DNase-seq footprinting provides a means to predict genome-wide binding sites for hundreds of transcription factors (TFs) simultaneously. However, despite the public availability of DNase-seq data for hundreds of samples, there is neither a unified analytical workflow nor a publicly accessible database providing the locations of footprints across all available samples. Here, we implemented a workflow for uniform processing of footprints using two state-of-the-art footprinting algorithms: Wellington and HINT. Our workflow scans the footprints generated by these algorithms for 1,530 sequence motifs to predict binding sites for 1,515 human transcription factors. We applied our workflow to detect footprints in 192 DNase-seq experiments from ENCODE spanning 27 human tissues. This collection of footprints describes an expansive landscape of potential TF occupancy. At thresholds optimized through machine learning, we report high-quality footprints covering 9.8% of the human genome. These footprints were enriched for true positive TF binding sites as defined by ChIP-seq peaks, as well as for genetic variants associated with changes in gene expression. Integrating our footprint atlas with summary statistics from genome-wide association studies revealed that risk for neuropsychiatric traits was enriched specifically at highly-scoring footprints in human brain, while risk for immune traits was enriched specifically at highly-scoring footprints in human lymphoblasts. Our cloud-based workflow is available at github.com/globusgenomics/genomics-footprint and a database with all footprints and TF binding site predictions are publicly available at http://data.nemoarchive.org/other/grant/sament/sament/footprint_atlas.

We present a conditional generative model for learning variation in cell and nuclear morphology and predicting the location of subcellular structures from 3D microscopy images. The model generalizes well to a wide array of structures and allows for a probabilistic interpretation of cell and nuclear morphology and structure localization from fluorescence images. We demonstrate the effectiveness of the approach by producing and evaluating photo-realistic 3D cell images using the generative model, and show that the conditional nature of the model provides the ability to predict the localization of unobserved structures, given cell and nuclear morphology. We additionally explore the model’s utility in a number of applications, including cellular integration from multiple experiments and exploration of variation in structure localization. Finally, we discuss the model in the context of foundational and contemporary work and suggest forthcoming extensions.

The ATP synthase (F-ATPase) is a highly complex rotary machine that synthesizes ATP, powered by a proton electrochemical gradient. Why did evolution select such an elaborate mechanism over arguably simpler alternating-access processes that can be reversed to perform ATP synthesis? We studied a systematic enumeration of alternative mechanisms, using numerical and theoretical means. When the alternative models are optimized subject to fundamental thermodynamic constraints, they fail to match the kinetic ability of the rotary mechanism over a wide range of conditions, particularly under low-energy conditions. We used a physically interpretable, closed-form solution for the steady-state rate for an arbitrary chemical cycle, which clarifies kinetic effects of complex free-energy landscapes. Our analysis also yields insights into the debated “kinetic equivalence” of ATP synthesis driven by transmembrane pH and potential difference. Overall, our study suggests that the complexity of the F-ATPase may have resulted from positive selection for its kinetic advantage.