Biology subjects

Talmon, R.

Publications and source records attributed to Talmon, R..

4 recordsLinked to original sources

bioRxiv2024

SORBET: Automated cell-neighborhood analysis of spatial transcriptomics or proteomics for interpretable sample classification via GNN

Spatial cellular profiling technologies have revolutionized our understanding of complex biological processes, from development and disease progression to immunity and aging. Despite their promise, integrating spatial information with multiplexed molecular data to accurately predict phenotypes poses significant challenges, especially in clinical settings. Here, we present SORBET, a geometric deep learning framework that directly analyzes complete spatial profiling data, eliminating the need to compress complete cell profiles into a limited set of annotations, such as cell types. SORBET models tissues as graphs of adjacent cells and applies graph convolutional networks to infer emergent phenotypes, such as responses to immunotherapy. The model leverages a novel data augmentation technique to ensure robust predictions, complemented by tailored interpretability analyses to identify the molecular and spatial patterns underlying the models phenotype inferences. We apply our method to a CosMx spatial transcriptomics dataset of pre-treatment metastatic melanoma samples annotated with response to immunotherapy; we show that spatial information significantly improves clinical endpoint, or phenotype, prediction and identifies important biological patterns. To our knowledge, SORBET is the first example of phenotype prediction on spatial transcriptomics data. We further validated our method using two spatial proteomics datasets, Imaging Mass Cytometry (IMC) and Co-detection by indexing (CODEX), obtained from Non-Small Cell Lung Cancer and Colorectal Cancer samples, respectively. SORBET demonstrates superior accuracy in phenotype prediction over leading spatial and non-spatial methods across various datasets of different observed phenotypes and technologies. SORBET sets a new benchmark for predictive analysis in spatial omics, promising to advance personalized medicine through refined patient treatment stratification, grounded in molecular and spatial tissue profiling.

bioinformatics↗

bioRxiv2023

VTA projections to M1 are essential for reorganization of layer 2-3 network dynamics underlying motor learning

The primary motor cortex (M1) is crucial for motor skill learning. Previous studies demonstrated that skill acquisition requires dopaminergic VTA (ventral-tegmental area) signaling in M1, however little is known regarding the effect of these inputs at the neuronal and network levels. Using dexterity task, calcium imaging, chemogenetic silencing, and geometric data analysis, we demonstrate VTA-dependent reorganization of M1 layer 2-3 during motor learning. While average activity and average functional connectivity of layer 2-3 network remain stable during learning, the activity kinetics, the correlational configuration of functional connectivity, and average connectivity strength of layer 2-3 neurons gradually transform towards an expert configuration. In addition, task success-failure outcome signaling gradually emerges. Silencing VTA dopaminergic inputs to M1 during learning, prevents all these changes. Our findings demonstrate dopaminergic VTA-dependent formation of outcome signaling and new connectivity configuration of the layer 2-3 network, supporting reorganization of the M1 network for storing new motor skills.

neuroscience↗

bioRxiv2020

Graph of graphs analysis for multiplexed data with application to imaging mass cytometry

Hyper spectral imaging, sensor networks, spatial multiplexed proteomics, and spatial transcriptomics assays is a representative subset of distinct technologies from diverse domains of science and engineering that share common data structures. The data in all these modalities consist of high-dimensional multivariate observations (m-dimensional feature space) collected at different spatial positions and therefore can be analyzed using similar computational methodologies. Furthermore, in many studies practitioners collect datasets consisting of multiple spatial assays of this type, each capturing such data from a single biological sample, patient, or hyper spectral image, etc. Each of these spatial assays could be characterized by several regions of interest (ROIs). The focus of this paper is on a particular application, imaging mass cytometry (IMC), which falls into this problem setup. To extract meaningful information from the multi-dimensional observations recorded at different ROIs across different assays, we propose to analyze such datasets using a two-step graph-based approach. We first construct for each ROI a graph representing the interactions between the m covariates and compute an m dimensional vector characterizing the steady state distribution among features. We then use all these m-dimensional vectors to construct a graph between the ROIs from all assays. This second graph is subjected to a nonlinear dimension reduction analysis, retrieving the intrinsic geometric representation of the ROIs. Such a representation provides the foundation for efficient and accurate organization of the different ROIs that correlates with their phenotypes. Theoretically, we show that when the ROIs have a particular bi-modal distribution, the new representation gives rise to a better distinction between the two modalities compared to the maximum a posteriori (MAP) estimator. We applied our method to predict the sensitivity to PD-1 axis blockers treatment of lung cancer subjects based on IMC data, achieving 92% accuracy. This serves as empirical evidence that the graph of graphs approach enables us to integrate multiple ROIs and the intra-relationships between the features at each ROI, giving rise to an informative representation that is strongly associated with the phenotypic state of the entire image. Importantly, this approach is applicable to other modalities such as spatial transcriptomics. Author summaryWe propose a two-step graph-based analyses for high-dimensional multiplexed datasets characterizing ROIs and their inter-relationships. The first step consists of extracting the steady state distribution of the random walk on the graph, which captures the mutual relations between the covariates of each ROI. The second step employs a nonlinear dimensionality reduction on the steady state distributions to construct a map that unravels the intrinsic geometric structure of the ROIs. We show theoretically that when the ROIs have a two-class structure, our method accentuates the distinction between the classes. Particularly, in a setting with Gaussian distribution it outperforms the MAP estimator, implying that the mutual relations between the covariates and spatial coordinates are well captured by the steady state distributions. We apply our method to imaging mass cytometry (IMC). Our analysis provides a representation that facilitates prediction of the sensitivity to PD-1 axis blockers treatment of lung cancer subjects. Particularly, our approach achieves state of the art results with accuracy of 92%.

bioinformatics↗

bioRxiv2020

Cell-type specific outcome representation in primary motor cortex

Adaptive movements are critical to animal survival. To guide future actions, the brain monitors different outcomes, including achievement of movement and appetitive goals. The nature of outcome signals and their neuronal and network realization in motor cortex (M1), which commands the performance of skilled movements, is largely unknown. Using a dexterity task, calcium imaging, optogenetic perturbations, and behavioral manipulations, we studied outcome signals in murine M1. We find two populations of layer 2-3 neurons, "success"- and "failure" related neurons that develop with training and report end-result of trials. In these neurons, prolonged responses were recorded after success or failure trials, independent of reward and kinematics. In contrast, the initial state of layer-5 pyramidal tract neurons contains a memory trace of the previous trials outcome. Inter-trial cortical activity was needed to learn new task requirements. These M1 reflective layer-specific performance outcome signals, can support reinforcement motor learning of skilled behavior.

neuroscience↗